Varicose veins : Epidemiology and Outcornes
Xavier Kurz Deparbnent of Epidemiology and Statistics McGill University, Montreal
December, 1999
A Thesis submitted to the Faculty of Graduate Studies and research in partial fulfilment of the requirements of the Degree of Doctor in Philosophy
@Xavier Kurz 1999
Librafy I*loNational fCmda
Bibliothèque nationale du Canada
Acquisitions and Bibliographie Services
Acquisitions et services bibliographiques
395 Wollirigton Sbwt OcraumON K t A W
395. rue Wellington -ON KlAON4
Canado
canada
The author has granteci a nonexclusive licence ailowing the National Library of Canada to reproduce, loan, distribute or sell copies of this thesis in microform, paper or electronic formats.
L'auteur a accordé une licence non exclusive pexmettant à la Bibliothèque nationale du Canada de reproduire, prêter, distribuer ou vendre des copies de cette thèse sous la forme de microfiche/nim, de reproduction sur papier ou sur format électronique.
The author retains ownership of the copyright in this thesis. Neither the thesis nor substantial extraçts fiom it may be printed or otherwise reproduced without the author's permission.
L'auteur conserve la propriété du droit d'auteur qui protège cette thèse. Ni la thèse ni des extraits substantiels de celle-ci ne doivent être imprimés ou autrement reproduits sans son autorisation.
Statement of originality The primary goal of this thesis is to identify risk indicators and outcomes specifically related to varicose veins. The originality of the thesis arises from its major contribution to the knowledge of one of the most prevalent medical conditions in western countries. One original aspect of this study is to propose a classification of subjects which singles out a large group of subjects with only varicose veins. It allows appraisal of specific risk indicators for this condition. and identifies outcomes associated with early stages of venous disease. This study also provides original information on the validity of the diagnosis of varicose veins made by general practitioners in cornparison to specialists. More importantly. it identifies variables associated with the probability of misclassification. In order to investigate risk indicators. a multivariate risk mode! in women is fitted taking into account previously suspected risk factors. The relationship between varicose veins and deep or superficial venous incornpetence is specifically investigated. This projed has also contributed to the developrnent of an unique disease-specific quality of life measure that has been fully validated in several languages. It allows the study of the impact of varicose veins on symptoms and quality of life taking into account other stages of venous disorders and quality of life deteminants. which had never been done before. The utilisation of health resource by patients with varicose veins in Belgium is assessed. and estimates of the annual cost of variwse veins from the society and patient perspectives are provided. The originality of this thesis in this respect is in the use of both patient-based and physician-based prospective infomation collected independently from each other. and of retrospective data to supplement this information.
The data analysed in this thesis were collected in the ccntext of an international research programme on chronic venous disorders of the leg w-ordinated at the Centre for Clinical Epidemiology and Community Studies, Jewish General Hospital, McGill University. Montreal. This programme included an International Task Force on the Management of Chronic Venous Disorders of the Leg and the VEnous INsufficiency Epidemioiogical and
economical Study (VEINES). a one-year cohort study of 1.531 subjects recruited in Belgium. France, ltaly and Quebec. The programme thus involved the collaboration of many persons in several countries. The core of the comprehensive epidemiological review on varicose veins presented in this document is the result of the collaborative work of the Task Force. whose report is to be published as a Supplement to Phlebology in June 1999. My role in that project was a central one. I contributed to the development of the protocol for literature search. identification of relevant studies and evaluation of scientific evidence. I organised and carried out part of the work of the search and data extraction. I reviewed al1 articles prior to their distribution to members of the Task Force and synthesised the scientific evidence on the basis of scores provided by reviewers. I wrote several chapters of the final report. and I finalised draft versions of chapters written by other Task Force rnembers. The literature review on epidemiological aspects of varicose veins has been updated and considerably expanded for this thesis. specifically with regard to the identification of risk indicators.
My role in the development and conduct of the VEnous INsufficiency Epidemiological and economical Study was also essential and represents an original contribution to the epidemiology of varicose veins.
I was a member of the scientific cornmittee that
developed the study design and I wrote al1 versions of the protocol. I constructed the eleven questionnaires used in the study. I CO-ordinatedthe launch and the conduct of the study in Belgium. France and Italy. a task that cornprised the recruitrnent, training and supervision of local staff. clinical research assistants and investigators, the monitoring of study progress over the two year period of data collection. and the data management. After consultation with specialists in the field. I wrote the protocol to a parallel study carried out to assess the prevalence of venous incornpetence based on an echo-Doppler test (the 'Duplex study') and the validity of general practitioners'
diagnosis of varicose veins.
I CO-ordinatedits implementation in 150 patients in
Belgium. France and Italy, and organised the data management.
The disease-specific quality-of-life
(QOL)scales used in this thesis were developed
and validated in the course of the VEINES study by a member of rny thesis committee.
1 nevertheless contributed to the development of scale items and to the design of the questionnaire. I organised the pre-testing of the questionnaire and the back-to-back translations. I prepared clean data files used for the computation of QOL scores and for the psychometric evaluation. I also wrote the protocol of the test-retest study of the
QOL scale carried out in connection with the VEINES study. organised its irnplernentation in Belgium. France and ltaly and took over the data management. All statistical analyses presented in this thesis. including those based on QOL scores. are my own contribution.
Acknowledgements
Many people have contributed to the accomplishment of this thesis. l am particularly indebted to rny supervisor. Prof. Lucien Abenhaim. for teaching me how epidemiological principles can be translated into practice. His insight into the theory of epidemiology and his experience of field work have tumed each step of the realisation of this thesis. from protocol development to data analysis, into a challenging learning experience. His confidence in my ability to participate in the developrnant and coordination of the VEINES study has been an important contribution to my professional attitude. I am also indebted to Prof. Jean-Paul Collet. Donna Lamping and Louis Eeckhoudt for
their guidance and constant encouragement. Their scientific advice and their active support for several years have largely contributed to the fuifilment of this thesis. I wish also to thank Dr. François Zuccareili for sharing his clinical expertise anytime I called him on the phone. This work could not have been accomplished without the assistance provided by Prof. Albert Dresse. He hosted part of the coordination of the VEINES study in the
Department of Pharmacology of the University of Liège and provided logistical support for the conduct of the study, the data analysis and the writing of the thesis. His backing
has been essential for the completion of this project. The VEINES study involved the collaboration of many people in several countries. I wish to gratefully acknowledge al1 general practitioners and specialists who accepted to participate in the VEINES study. I am especially grateful to persons who were key actors in its organisation and implementation in Europe: Ugo Baccaglini. Michel Boutsen, Jacqueline Brau, Didier Bry, Jean-Jacques Busso. Agnès Claes, Hector Falkoff. Michel Roland. Patrick Serey and Giorgio Spreafico. In Montreal. Anick Bérard. Susan Khan, Kader Kourad and Luc Lalonde played an active role in the coordination
a
and data management of the VEINES study
I thank Carole Bohbot, Linda Carfagnini, Francesca Lipari and Mima Mancini for their
administrative and secretarial assistance at various stages of progress of the thesis. This thesis owes much to Thierry Roisin, my colleague at the Belgian Centre for
Pharmacovigilance. who took over part of my workload during the last weeks of my writing. 1 also wish to thank Dr. Sebastian Moss who fan the linguistic and spelling checks in a
short period of time. l am much indebted to my wife Marie-Françoise for her support. understanding and patience during the last years. My thoughts also go to my daughters Amélie. Adele and Olivia who grew up at the same time as the VEINES study and quickly learned the meaning of the word "thesis".
I wish to thank my mother. brother and sisters for their continuous support. This work is
dedicated to rny father who taught me to carry on my projects to their end but did not see the completion of this one.
Abstract Varicose veins : Epiderniology and Outcomes Varicose veins are among the rnost prevalent medical conditions in western populations. with a prevalence estimated at 25-35% in women and 10-2056 in men. Until now. few studies have regarded varicose veins as a distinct clinical entity and have investigated specific risk factors. Their consequences for the patients have not been adequately investigated. The main objective of this thesis was to examine the association between varicose veins and specific risk indicators and outcomes. taking into account the effects of more severe venous disorders often found in combination with varicose veins. This work is based on the VEnous INsufficiency Epidemiological and econornical Study (VEINES), a one-year cohort study on venous disorders carried out in Belgium. France, ltaly and Quebec. It included 1531 patients sampled among 5688 consecutive patients consulting a physician for a venous disorder. A sub-sample
of 150 patients were referred to specialists for clinical examination and duplex investigation of venous incornpetence.
This study illustrated the probtems of the
diagnosis and classification of varicose veins. with a specificity of 45% for the diagnosis made by general practitioners. In a case-control analysis, the strongest risk indicators of varicose veins were pregnancy, age and farnily history. No association was found with other hypothesised determinants (obesity, smoking, history of thrornbophlebitis. blood group A).
Results of duplex studies support the hypothesis of a distal onset of
venous reflux and varicose veins. Using a classification of varicose veins proposed to take account the concomitant presence of other signs of venous disease. varicose veins alone had no impact on a symptom score and on generic (SF-36) and disease-specific quality of life scores. The results suggest that symptoms and presence of varicose veins are independent outcomes. which has implications for clinical practice. A detailed analysis of health service utilisation performed in Belgium also showed that both are independent
a
predictors of resource use.
Veine variqueuses : épidémiologie et conséquences
Les veines variqueuses sont parmi les problemes médicaux les plus fréquents dans les pays occidentaux. avec un prévalence de 25-35% chez les femmes et 10-20% chez les hommes. Jusqu'à présent. peu d'études. ont considéré cette pathologie comme une entitié clinique spécifique. notamment concernant leurs facteurs de risque et leurs conséquences pour les patients. L'objectif de la thèse a été d'examiner Itassociation entre les veines variqueuses. leurs facteurs de risque et leurs conséquences pour les patients, tenant compte des effects d'autres déterminants. Ce travail est basé sur l'étude VEnous INsufficiency Epidemiological and economical Study (VEINES). une étude de cohorte d'un an réalisée en Belgique, en France. en Italie et au Québec. Elle
a inclus 1531 sujets recrutés parmi 5688 patients consécutifs consultant un médecin pour des problèmes veineux.
L'étude a illustré les problemes de diagnostic et de
classification des veines variqueuses. avec une spécificité de 45% pour le diagnostic réalisé par des médecins généralistes. Dans une analyse de type cas-témoin. les indicateurs de risque les plus importants ont été la grossesse. l'âge et une histoire familiale de varices. Une association avec d'autres facteurs de risque n' a pas été retrouvée. Les résultats de l'étude duplex sont en faveur d'une origine distale du reflux veineux et des varices. Une classification des veines variqueuses qui tient compte de la présence concomittante d'autres signes de la pathologie veineuse est proposée. La
présence de varices seules n'a pas eu d'effet sur un score de symptômes ni sur des scores de qualité de vie générique et spécifique.
Les résultats suggèrent que
symptômes et aspects cosmétiques sont indépendants; ceci a des conséquences en pratique clinique. Une analyse détaillée de l'utilisation des soins de santé en Belgique a également montré que les symptômes et la présence de varices sont des prédicteurs indépendant de l'utilisation des soins de santé.
viii
Table of contents 1. Scope of thesis ................................................................................................. 1
.
2 Objectives ......................................................................................................... 5
.
3 Literature Review
3.7 Definition and classification of varicose veins...............................................7 3.1.1 3.1.2 3.1.3 3.1.4 3.1.5
Pathophysiology ...........................................................................................7 Definition of varicose veins ...........................................................................9 Methods used for the diagnosis of varicose veins ...................................... 10 Assessrnent of varicose veins in patients' relatives ....................................13 Classification of varicose veins by type and severity .................................15
3.2.1 3.2.2 3.2.3 3.2.4 3.2.5
Prevalence .................................................................................................22 Incidence....................................................................................................27 Variables associated with varicose veins.................................................. 28 Associated signs and clinical complications ............................................... 38 Conclusions................................................................................................40
3.3 Symptoms and quality of life related to varicose veins ............................... 41 3.3.1 Symptoms ..................................................................................................41 3.3.2 Quality of life .............................................................................................. 43 3.3.3 Conclusions................................................................................................ 47
3.4 Health resources utilisation ....................................................................... 4 8 3.4.1 Introduction ............................................................................................... 4 8 3.4.2 Costs associated with the treatment of varicose veins ...............................49
.
4 Patients and Methods :the VEINES study 4.1 Introduction................................................................................................. 52
4.2 Scope and objectives ........................................................................... 5
2
4.3 Developmenf and initiation of study............................................................53 4.3.1 Protocol development................................................................................. 53 4.3.2 Initiation of study ...................................................................................... 54
4.4 Study design............................................................................................... 4.4.1 4.4.2 4.4.3 4.4.4
54
Rationale for the choice of a cohort population.........................................54 Definition of source population ................................................................... 57 Definition of study population ................................................................. 6 0 Size of study population ............................................................................. 62
4.5 Sampling procedures................................................................................. 6 3 4.5.1 4.5.2 4.5.3 4.5.4 4.5.5
....................................... 63 Recruitment of physicians ............................... . Registration phase ..................................................................................... 65 Sampling procedure ................................................................................... 68 Assessment of exclusion criteria ...........................................................69 Inclusion.....................................................................................................70
4.6 Data collection ............................................................................................ 72 4.6.1 Schedule of data collection.........................................................................72 4.6.2 Algorithm of data collection ....................................................................... 74 4.6.3 Socio-demographic information..................................................................75 4.6.4 History of venous disease and associated factors ......................................75 4.6.5 Symptoms .................................................................................................. 76 4.6.6 Signs .......................................................................................................... 76 4.6.7 Laboratory measures .................................................................................77 4.6.8 Economic evaluations ................................................................................ 78
4.7 The Duplex study........................................................................................ 78 4.7.1 4.7.2 4.7.3 4.7.4
Objectives ..................................................................................................78 Rationate ................................................................................................. 79 Patients and methods................................................................................. 79 Study population ........................................................................................ 81
4.8 Quality control...........................................................................................8 7 4.8.1 4.8.2 4.8.3 4.8.4 4.8.5
Single dossier per patient ........................................................................... 81 Patient identification.............................. . .................................................82 Check visits and data control...................................................................... 82 Newsletters .............................................................................................. 83 Data audit................................................................................................... 84
4.9 Ethical considerations................................................................................ 8 4
.
5 Source and study populations 5.1 Source population ...................................................................................... 85 5.1.1 introduction ................................................................................................85 5.1-2 Results ....................................................................................................... 85 5.1.3 Discussion..................................................................................................87
5.2 Study population........................................................................................ 8 8 5.2.1 5.2.2 5.2.3 5.2.4
.
Introduction................................................................................................ 83 Results ....................................................................................................... 89 Follow-up rate and observation time........................................................... 90 Discussion.................................................................................................. 91
6 Classification and clinical description of varicose veins
8.3.5 Controls.................................................................................................... 141 8.3.6 Sample size.............................................................................................. 143 8.3.7 Statistical power ...................................................................................... 143 8.3.8 Exposure variables ................................................................................... 144 8.3.9 Confounders.............................................................................................149 8.3.10 Statistical analysis ..................................................................................151
8.4 Results ..................................................................................................... 153 8.4.1 Unadjusted analyses ................................................................................ 153 8.4.2 Multivariate analyses .............................................................................. 157
8.5 Discussion ................................................................................................ 165 8.5.1 8.5.2 8.5.3 8.5.4
.
Methodological issues ..............................................................................165 Comparison with other studies .................................................................174 Discussion of results ................................................................................176 Conclusion ............................................................................................... 184
9 Varicose veins and venous reflux
9.7 Introduction............................................................................................... 185 9.2 Objectives................................................................................................. 186 9.3 Patients and methods............................................................................... 787 9.3.1 9.3.2 9.3.3 9.3.4 9.3.5
Overview ..................................................................................................187 Study population ...................................................................................... 188 Assessment of venous reflux...................................................................188 Assessment of clinical status .................................................................190 Data analysis ............................................................................................191
9.4 Results ..................................................................................................... 9.4.1 9.4.2 9.4.3 9.4.4 9.4.5
192
Prevalence of venous reflux .................................................................... 192 Distribution of reflux. by leg ...................................................................... 195 Laterality of reflux ..................................................................................... 196 Laterality of varicose veins and reflux according to pregnancy................. 198 Variables associated with venous reflux ................................................... 199
9.5 Discussion ............................................................................................ 2 0 1 9.5.1 Prevalence of venous reflux .................................................................. 201 9.5.2 Distribution of venous reflux ............*..............*.......... ..--.....................202 9.5.3 Variables associated with reflux ............................................................ 204
.
10 Complications 10.1 Introduction..........~.................................................................................2 0 5 10.2 Objectives..............................................................................................
206
10.3 Methods................................ .................................................................2 0 6 10.3.1 Study population........................~............................................................ 206
A 4
A A
lu
A
+a
pY
oi_ 3 ez O
0,3 - U)
-. .. .. ... ... . .... ..... .... .... ... ... .. .. ...,. ... ... .. ..... .. *.. ., ... ., ,. ... .
ip" i
.. ... ... ... ... ... .... ... .. .. .. .... . .... ... ... .... .... .... . ..... .. . ..... .,., .....
,
,
a
.
7
.
8
.
IUN
P
cd
w ru
xiii 12.3.4 Design options for cost valuation ........................... .... ........................... 264 12.3.5 Data analyses ........................................................................................268
7 2.4 Results .................................................................................................. -277 12.4.1 12.4.2 12.4.3 12.4.4 12.4.5 12.4.6 12.4.7 12.4.8 12.4.9
Observation time .................................................................................... 277 Consultations ....................................................................................... 278 Diagnostic tests ...................................................................................... 279 Dnig utilisation ....................................................................................... 280 Use of compression material ............................................................. ..... 282 Use of sclerotherapy and surgery ......................................................... 283 Hospitalisations ...................................................................................... 284 Cost estimates in general-practitioner-based sample ............................. 284 Variables associated with total costs ...................................................... 292
12.5 Discussion .............................................................................................-294 12.5.1 Methodological issues .....................................................................294 12.5.2 Total costs associated with varicose veins ............................................. 299 12.5.3 Variables associated with costs .............................................................. 300
Appendices ...................................................................................................... 303 Appendix 1.........................................................................................................303 Appendix 2 ....................................................................................................... 3û4 Appendix 3 ........................................................................................................ 314
13. Conclusions 73.7 Methodological aspects .......................................................................... 323 13.1-1 Strengths of the VEINES cohort study design ........................................ 323 13.1.2 Weaknesses of the VEl NES cohort study design ...................................325
13.2 Main findings .......................................................................................... 329
List of references ............................................................................................. 335
1. Scope of thesis
Varicose veins have been recognised as a pathological condition since very ancient times. Hippocrates (-460. -377 BC). Jean Fernel (1497-1558). Marianus Sanctus of Barletta (16th century) and Ambroise Pare (1509-1590) have, among others. proposed various aetiologies and treatments for varicose veins (Madar et al.. 1986). The first known monograph devoted to this subject dates from 1837. With foresight. Piorry wrote: 'It is rather difficult to understand why the investigation of veins has been passed over almost in silence. while such great diagnostic value has been attached to the investigation of arteries'. This statement was reiterated eighty years later by Nobl in his work on venous diseases (Nobl. 1918). It was only in the 1930's that data about the public health importance of varicose veins becarne available. In the United-States National Health Survey of 1935-1936.varicose
veins were estimated to be present in 1.75 million of the population and ranked 7th in the list of the most important causes of medical referrals (Callam. 1994).
The
importance of varicose veins among other medical conditions was later confirmed in large-scale sickness surveys carried out in the United Kingdorn (the Survey of Sickness of England and Wales). in Denmark (the Sickness Survey of Denmark), and in a further US National Health Survey. Other data came from anecdotal evidence or from case series published in the SOIS, but it is only in the last thirty years that formal epidemiological studies on prevalence, incidence and risk factors of varicose veins have been carried out.
Landmark studies have included the population cohorts of
Tecumseh (Coon et al.. 1973) and Framingham (Brand et al.. 1988),and a follow-up of several thousand workers of the chemical industry in Basle (Widmer et al.. 1978; Widmer et al., 1992). In recent years. collegiate meetings and conferences have reached consensus on specific aspects of the management of venous disorders. such as their classification (Porter et al.. 1995). their investigation (Nicolaides et al.. 1999) and the use of
sclerotherapy (Baccaglini et al., 1995). An international Task Force made an in-depth evaluation of the published evidence regarding the epidemiology. outcornes, diagnosis and treatrnents of chronic venous disorders (Abenhaim et al., 1999). These consensus documents and several epidemiological reviews (Beaglehole. 1986; Callam. 1994; Evans et al., 1994; De Backer. 1997) have consistently emphasised methodological difficulties found in many studies.
A major issue is the proper definition. diagnosis and classification of varicose veins.
The general term "varicose veins" refers to a very wide range of pathology. At one extreme. the main concern is cosmetic; at the other extreme, incornpetence of both the deep and superficial venous systems is associated with problerns of skin changes and
ulceration. A classification of varicose veins which reflects the range in severity of the condition. and which is of operational use to define health care needs. is therefore clearly important (Robbins et al., 1992). Widmer et al. (1978) made a pioneering work by classifying subjects according to two criteria: the size of the varicosities and the severity of the venous disease: they found that the sub-groups so defined differed in terms of complications and intensity of treatment. However. perhaps due to the relative complexity of this dual classification. the full system proposed by Widmer has been rarely used in subsequent epidemiological studies. The majority of available data on risk factors, symptoms. quality of life and health care costs of varicose veins are therefore confounded by severity of venous disease. A recent classification system (Porter et al., 1995) has isolated a group of subjects with varicose veins alone. but does not allow dissociation, among individuals with more severe venous disease, of those who have and those who do not have varicose veins. In addition, the validity and reliability of the clinical diagnosis of varicose veins have often been questioned.
The methodological difficulties found in many epidemiological studies have several consequences.
Evidence regarding many aspects of the condition is currently
inadequate and this uncertainty applies to the etiology of varicuse veins. Although their multifactorial origin has been hypothesised for a long time. it is only reœntly that
multivariate analyses have been undertaken to attempt to answer the hypothesis (Scott et al., 1995; Sisto et al., 1995). Despite of this. available results have only been adjusted for a limited number of potential confounding variables.
Although studies have provided
scientific evidence regarding the association between varicose veins and variables such as pregnancy and obesity, the size of the risk and its relationship to levels of exposure is not adequately characterised. Many other risk indicators have been studied in univariate analysis only. Data are also lacking for patient-based outcornes. and especially quality of life. Although they are needed in a wmprehensive evaluation of the burden of a chronic disease. the data currently available do not allow separation of quality of life impairment due to varicose
veins from that ascribed to age. CO-morbidityor signs of more severe venous disease. This information is nevertheless of critical importance. as it has been suggested that a substantial proportion of the high costs expended on the treatment of varicose veins could be due to wnsumption of health care resources for non venous conditions or non medically significant venous disorders (Bradbury et al. 1999: Robbins et al.. 1992). Several economic studies have been carried out in hospital patients adrnitted for treatment of varicose veins. using aggregate statistics from health insurance companies or reimbursement schemes but the fraction of the costs specifically related to that pathology has not been adequately investigated. Neither has the association between health semice utilisation and disease-specific quality of life been thoroughly examined. The goal of this thesis is to investigate the association between varicose veins and specific risk indicators and outcornes. taking into account the effect of more severe venous disease.
The term "outcorne" is used in this context as per the Oxford
Dictionary of Current English broad definition: the effect or result of an event, or of circumstances. In order to achieve this goal. an operational classification of varicose veins into six categories based on the CEAP classification (Porter et al.. 1995) is proposed. Risk indicators for presenting varicose veins alone are identified and the association between varicose veins and patterns of venous incornpetence is examined. The effects of different clinical categories of varicose veins on symptoms. quality of life.
complications and health care utilisation are measured and compared. Since intensity of trealrnent is rnostly delermined at the prirnary care levei. the vaiidity of the diagnosis of varicose veins by family physicians is also investigated.
2.1 General objective
To investigate the association between varicose veins and specific risk indicators and outcomes, taking into account the effects of more severe venous disorders often combined with varicose veins.
2.2. Specific objectives To assess the validity of the clinical diagnosis of varicose veins made by general practitioners, and to identify variables associated with misclassification.
0
To identify risk indicators for varicose veins, and to measure the strength of association between the presence of varicose veins and variables suspected to be causally related to varicose veins. thereby contributing to supporting or refuting such
causal links. To estimate the prevalence of venous reflux in patients with varicose veins. to determine whether the distribution of reflux is compatible with a distal onset of valve incompetence. and to examine the association between venous reflux and variables suspected to be related to the occurrence of varicose veins. thereby contributing to the understanding of the role of venous reflux in the etiology of varicose veins. To
measure the
incidence rate of
oedema,
skin
changes,
ulceration.
thrombophlebitis and varicose vein bleeding in patients with varicose veins. and to examine the influence of more severe venous disorders.
To assess symptoms and quality of life in patients with varicose veins. and to detemine whether varicose veins are assaciated with these outcornes taking into account other hypothesised detenninants. a
10 assess the intensity of health services utilisation in subjects with varicose veins consulting general practitioners and specialists in Beigium. and to identify the effect of varicose veins on specific health care costs taking into account other determinants of resource use.
0
3. Literature review 3.1 Definition and classification of varicose veins
3.1.1 Pathophysiology
A brief overview of pathophysiology of varicose veins will help understanding of several
aspects presented in this thesis. The venous system of the legs comprises a superficial systern in the skin and the subcutaneous fat, and a deep system beneath the fascia.
At its root. the long
saphenous vein of the superficial system joins the femoral vein of the deep system in the inguinal region (Figure 1. right panel). The superficial system is a venous network wtth prominent long and short saphenous veins. and these and other perforating veins pass through the fascia to join the deep veins (Figure 1. upper left panel). The superficial veins. the perforating veins and the deep veins contain valves to prevent backflow. Blood is returned from the periphery by the pumping action of the calf muscles. which compress the deep vein that contain one way valves. When deep veins are emptied, reflux is prevented by the valves and blood is sucked in from the superficiai veins (Hobbs. 1991 ). 'Varicose veins' are dilated. lengthened and tortuous superficial veins. Dilation is often. but not always. associated with incompetence of the venous valves which permits reflux and abnormal pressure patterns. These incompetent valves may be located in perforator veins or in the superficial system (Figure 1. lower left panel). Euans et al. (1994) summarised as follows the theories suggested for the pathogenesis of varicose
veins :
valvular hypothesis : the primary defect is in the valves of the veins which are thought to fail sequentially from above the sapheno-femoral junction downwards; vein wall hypothesis : the primary defect is a weakness of the wall of the vein which
dilates rather than hypertrophies under pressure: arterio-venous anastomosis hypothesis : the primary defect is thought to be multiple small arterio-venous anastomoses. and veins becorne varicose under pressure:
Figure 3.1 Deep and superficial veins and mechanism of varicose veins deveiopment
High p r @ ~ u t e deep vein
Long saphenous NORMAL
Deep high
Supefficiel lowsr
pressure
pressure
Rie 16g veim. saen horn the front
(Drawing reproduced with permission from Laing, 1992. The drawing of the deep and superficial leg veins (nght panel) is reproduced with permission from : 'The Oxford Cornpanion to medicine, edited by John Watson, Paul Beeson and Ronald Bodley Scott for the Oxford University Press, 1986').
O
communicating vein hypothesis : the primary defect is incompetence of the valves of
the perforatinp veins and varicose veins are caused by high pressure flow from the deep system causing dilation.
The 'valvular' and 'vein walll hypotheses would describe the most probable mechanisms of pathogenesis of varicose veins, either singly or in combination (Evans. 1994: Hobbs. 1991: Labropoulos et al., 1997).
Two types of varicose veins are generally identified (Hobbs. 1991):
primary vaficose veins : these are familial and caused by valvular incompetence or weakness of the vein wall;
secondary varicose veins : these are usually caused by deep vein thrombosis. fol lowed by valve destruction during recanalisation of the thrombus. which results in valve incompetence in perforator veins joining the deep and superficial venous systems (Hopkins and Wolfe, 1992;Markel et al.. 1992.Saarinen et al.. 1995); more rarely, they can occur as a result of trauma or surgery (Willén et al., 1982: Lindhagen et al., 1985; Aitken et al.. 1989. Scott et al. 1995). Varicose veins should be distinguished from athletic veins (normal veins that are prominent in healthy muscular legs) and locally dilated venules. also known as hyphenwebs. spider veins or telangiectasia.
3.1.2 Definition
A first methodological issue in the examination of studies on varicose veins is the large
number of tems that have been used to refer to pathological veins visible on the Ieg. such as "varicosities", "varicose veins", "varices". It is not always clear what
is included by these
tems. The second issue is the absence of accepted standard definition of varicose veins (Evans et al.. 1994; Abenhaim et al, 1999). In epidemiological studies, many investigators (e.g. Beaglehole et al, 1976; Hirai et al.. 1990; Laurikka et al., 1993: Maffei et al.. 1986: Malhotra. 1972; Stanhope, 1975) have used Arnoldi's definition of a var~cosity:"any dilated,
@
elongated or tortuous veins, irrespective of size" (Arnoldi. 1 957). Others (e. g. Mekky et al..
1969) used the definition of a varicose vein proposed by Dodd and Crocken (1956) .'A
vancose vein is one which has permanently lost its valvular etficiency and, as a result of continuous dilation under pressure, in the course of time, becomes elongated, tortuous,
pouched and thickened'. This definition implies that the functional status of the venous valves is known. An important difference between definitions has been the specific inclusion (e.g. Hirai et al.
1990;Da Silva et al.,1974; Stvrtinova et al.. 1991: Sadick 1993. Evans et al..1998) or omission (e.g. Cornu-Thénard et al.. 1989; Maffei. 1986: Mekky et al.. 1969: Abramson et al.. 1981 ; Brand et al.. 1988; Canonico et al.. 1998) of telangiectasia and spider veins. Many other studies do not comment about the inclusion or exclusion of these conditions. These differences make direct cornparison of results of prevalence studies difficult since Arnoldis' definition (for example. the tenns 'dilated and 'tortuous?may be subject to different interpretations as to the size and shape of the veins.
3.1.3
Methods used for the dingnosis o f varicose veins
Not only have studies used ditferent definitions, they have also used different rnethods of
measurement. In some studies, the prevalence of varicose veins was estimated by self-administered questionnaire (Bobeck et al. 1966: Franck et al.. 1992a; Laurikka et al.. 1993: Sisto et al.. 1995), supplemented. in one study. with an examination of a sample of subjects in order to validate the method (Laurikka et al., 1993: Laurikka et al.. 1995).
Most studies included a clinical evaluation and various attempts have been made to standardise the conditions under which examinations were perfomed. Examples of such methods are : resting the subject prior to observation (Stanhope, 1975) subrnitting examiners and interviewers to initial and continued training (Evans et al.. 1998:Maffei et al., 1986) using only one observer (Laurikka et al.. 1993: Malhotra, 1972: Mekky et al.. 1969. Weddel, 1966)
a
using joint classification by two or more observers (Evans et al.. 1998; Latto et al.. 1973) having the diagnoses reviewed by a senior physician prior ta final documentatian (Coon et al.. 1973) using diagrams of the limbs to classify veins into grades of severity (Fraser et al.. 1985)
0
using Likert scales to rate severity (e.g. mildlmoderatelsevere or a score from O to 20: Coon et al, 1973; Cornu-Thénard et al.. 1991: Jakobsen. 1979: Large. 1985; Munn et al.. 1981)
a
using a cut-off value of the maximum diameter of the largest varicose vein as a criterion for being included among cases with varicose veins in case-control studies (CornuThénard et al., 1989; Cornu-Thénard et al.. 1994).
O
using photographic slides taken under standardised conditions and analysed by a team of examiners (Evans et al.. 1998: Widmer et al.. 1978; McAdam et al.. 1976).
Some authors have investigated the reliability and validity of clinical criteria used to identify
and diagnose varicose veins. and to rate their severlty. Reliabilitv and validitv of a self-report of varicose veins O
In the Mini-Finland Health Survey of 1978-1981. a self-administered questionnaire included structured questions on whether a physician had ever made a diagnosis of varicose veins in the lower extremities, and on reasons for al1 operations done in the past.
Re-administration of questions in a randorn sample of subjects gave kappa
coefficients of 0.71 in men and 0.83 in women for history of physicians' diagnosis of varicose veins and of 1.O0and 0.90 for operation for varicose veins respectively (Sisto et al.. 1995). r A study in Finland showed that the seif-reported diagnosis of varicose veins had overall
sensitivity and specificity estimates of 0.92 and 0.93 respectively in cornparison to physical examination by a specialist (Laurikka et al.. 1995). O
In a study based on questionnaire followed by a clinical examination. a sensitivity of 0.47 in men and 0.67 in women was found for self-reported diagnosis of variwse veins
ascertained in the questionnaire cornpared to clinical examination: the corresponding figures for specificity were 0.95 and 0.85 (Abramson et al.. 1988). In a preoperative assessrnent in 57 patients. independant ratings of the severity of varicose veins (minor. moderate or severe) were provided by both the patient and the surgeon. The weighted kappa value (calculated from the data presented in the paper) was 0.68 (95% CI: 0.52-0.84): the definition of minor and moderate varicose veins was the only source of discordance (Munn et al., 1981). Inter-observer reliabilitv of a clinical diaanosis O
A study has shown a correlation of 0.97 between two independent clinical specialists'
ratings of the severity of varicose veins on a 0-20 scale: correlations between specialists' ratings and six diameteri length parameters of the same veins ranged from 0.78 for total length to 0.94 for the sum of maximum diameters at each section (Cornu-Thénard et al.. 1991). a
In a clinical trial. inter-observer agreement was poor (43%) for ratings of improvement of cosmetic effects (i.e. presence of 'unsightly veins') (McAdam et al.. 1976). In a prevalence survey of 7.432 Paris policemen examined by 12 specialists. Ducimetière et al. (1981) found that the prevalence of varicose veins varied from 14% to 40% among examiners.
These results suggest a low inter-observer
agreement, but are difficult to interpret without information on the survey methods. Reliabilitv of the photoqra~hicmethod a
In order to test the reliability of the photographic method. the slides of 217 consecutive persons with 'varicosities' were reviewed twice by three physicians at an interval of 10 days. 90% of varicosities were classified similarly in the two examinations but no data of inter- and intraobsewer reliability are provided (Widmer et al.. 1978).
a
Photographic records were also used in a trial to measure treatment success following surgery for varicose veins. Three clinicians provided blind and independent ratings of photographs of patients' affected legs before and after surgery.
Inter-observer
agreement of treatment efflcacy based on photographs was 61.7%. Poor intra-observer
agreement was found between physical examination and visual assessrnent using photographs (25%) (McAdam et al.. 1976). The results on the validity and reliability of a diagnosis of varicose veins presented above are difficult to interpret.
They are derived from studies using different
methodologies and most of these studies are of limited size and provide few details about the way they were conducted. Very good results regarding the reliability of a selfreport of varicose veins were found in two studies in Finland. Laurikka et al. ascribed
these good results to varicose vetns being easily seen in sauna baths (Laurikka et al..
1995). Studies of inter-observer reliability gave inconsistent results and raise concerns about the most scientifically sound method of measuring presence or absence of varicose veins.
3.1.4 Assessment or varicose veins in patients' relatives
The role of heredity as an etiological factor of varicose veins bas drawn considerable
interest and many authors have attempted to assess the presence of varicose veins in relatives of their patients.
In two studies (Cornu-Thénard. 1994: Weddel, l969),the
presence of varicose veins in patients' relatives was assessed by direct examination. In al1 other studies. patients' family history of varicose veins was investigated by a questionnaire administered to the subject (Bobek et al.. 1966; Coon et al., 1973;Cornu-Thénard et al..
1989:Hirai et al.. 1990:Laurikka et al. 1995;Mekky et al., 1969; Sadick, 1992:Scott et al.. 1995: Widmer et al.. 1978). A patient's perception of the presence of varicose veins in
relatives can, however, be quite variable. Weddell et al. (1969)has reported that examination of relatives yielded vastly different results in cornparison with those obtained on questioning. Her conclusion that a history from the proband is an invalid method for detemining prevalence in other family memben has been followed by most authors of epidemiological reviews. A close examination of the Weddel study shows that her results may not be adequate to support her conclusion. Weddel interviewed and examined 88 individuals selected at random from one ward of the City of Cardiff. as well as 201 of their relatives. In each of these 289 subjects. a
detailed family history of varicose veins was taken. and specific enquiries were made into the presence or absence of varicose veins in the subjects' parents. brothers and sisters. Thus. each subject evaluated the presence of varicose veins on his relatives' legs, and his own legs were also evaluated by his relatives. A total of 595 observations were available : 68 on fathers, 86 on rnothers, 195 on brothers and 246 on sisters. The author reports that, in those with varicose veins, 52% of the observations by the rest of the family proved to be correct. while in those without varicose veins 93% were correct. Methodological problems of this study are obviously the lack of independence between the observations and a possible contamination of answers between relatives. Moreover. it is not stated whether physical examinations were performed unblinded as to the rating provided by relatives. No attempt was made to study the inter-observer reliability of the diagnosis of the presence of varicose veins on the same subject by different members of the same family, or to examine whether there were systematic differences in the accuracy of the diagnosis. We re-analysed the results presented in Weddel's publication in terms of sensitivity. specificity and predictive values (Table 3.1.). Table 3.1 Validity of a report of varicose veins in relatives in cornparison to clinical diagnosis (595 obsewations) (Weddel 1969)
Family member's evaluation
W No W Total
Examination W No W 71 76 416 32 487 108
Total t 47 448 595
sensitivity : 0.70 specificity : 0.85 positive predictive value : 0.52 negative predictive value : 0.93 prevalence : 0.18
Results suggest that about 70% of subjects with varicose veins are correctly identified by their relatives and that about 85% of subjects without varicose veins are also correctly identified. Only half of the reports by family members are corroborated by clinical examination, but this positive predictive value is affected by the relatively low prevalence of varicose veins (18.2%) found in relatives. Information on a family history of varicose veins could probably be improved. If each adult subject had to rate the prevalence of varicose veins in hisiher mother, a higher positive predictive value would
9
probably be found because the prevalence increases steeply with age and is iarger in fernales than in males. For example a prevalence of 30% would be associated with a 67% positive predictive value and a 87% negative predictive value. In conclusion, given methodological problems present in her study. Weddel's results are not adequate to refute or confirm the iack of validity of a family history provided by a subject.
3.1.5 Classification o f varieose veins by type and severity
Several classification systems have been proposed in order to evaluate the natural history of venous disorders. to assess the efficacy of treatments and to facilitate universal communication.
Widmer's classification (1978) Based on an epidemiological study (Basle III study) carried out on 4.529 healthy employees of pharmaceutical cornpanies in Basle. Widrner et al. (1978) created a classification with two levels based mainly on objective signs documented by photographs. The first level includes three classes of varicosities; the second level includes three grades of chronic venous insufficiency with increasing severity (Table
Table 3.2 The Basle study venous classification (Widmer et a1.J 978; Madar et al.. 1986: Callarn 1994)
i
No venous disease Varicosities - Hyphenwebs (= telangiectasia, spider veins): intradermal veinectasias - Reticular vancose veiris : dilated, tortuous subcutaneous veins, not belonging to the main trunk or its major branches - Trunk varicosities: dilated tnrnks of the long (greater) or shorter (lesser) saphenous vein and their branches of the first or second orâer Chronic venous insufficiency 1 Dilated subcutaneous veins, 'corona phlebedatica' II Hyperpigmented or depigmented areas III Open or healed leg ulcer
In Widmer's study population. 56% of 3.744 men had trunk varicosities andlor chronic venous insufficiency.
In 44%, the clinical significance of the findings on physical
examination was judged to be minimal. In 1236. changes were considered clinically significant. It was pointed out that the abnormalities coutd be interpreted as a 'disorder' in the first case and as a 'disease' in the second one (Madar et al.. 1986). The latter term. however, introduced a functional definition which relates only to clinical appearance of the limb without adequate diagnostic tests of venous function (Callam. 1994, Abenhaim et al., 1999). WidmerJsclassification has not been forrnally validated in other populations. but it has been used in several epidemiological studies or clinical
trials of treatments of venous disorders (Augustin et al., 1997: Boccalon et al.. 1997: Bradbury et al., 1999; Evans et al., 1998; Neumann et al..1990; Nocker et al.. 1990; Rudofsky, 1988; Widmer et al., 1990).
Porter's classification (1988) The Porter's classification was developed by an international expert cornmittee as part
of a document on general reporting standards in venous disease (Porter et al.. 1988). Four clinical classes of chronic venous insufficiency were defined (Table 3.3). Table 3.3 The Porter's classification of chronic venous insufficiency (CVI) (Porter et al., 1988) class O :
asymptomatic
class 1 :
mild CVI with signs and symptoms including mild to moderate ankle swelling, mild discomfort and local or generalised dilatation of subcutaneous veins; CVI is usually limited to involvement of superficial veins only.
class 2 : moderate CVI including hyperpigmentation of the skin in the gaiter area, moderate brawny oedema and subcutaneous fibrosis, but without ulceration; OUS there is usually prominent local or regional dilatation of the S U ~ C U ~ ~ B ~veins. class 3 :
severe CVI, with chronic distal pain associated with ulcerative or pre-ulcerative skin changes. eczematoid changes andfor severe oedema.
The Porter classification has been extensively used to classify patients in clinical studies on the diagnosis and treatment of venous disorders. The proportions of limbs with superficial or deep venous incompetence in each category of the Porter class may
be assessed from three studies performed in patients referred for a laboratory evaiuation of the venous function (Table 3.4). The results of the three studies show that the proportion of limbs with superficial or deep venous incornpetence increases with increasing disease severity. In the Rosfors
and in the Neglén studies. nearly al1 patients in Porter class 3 had superficial and deep venous insufficiency. Table 3.4 Proportion of limbs with superficial and deep venous reflux reflux. by Porter's class
I
1
Study
Population
Reference test
Rosfors 1990
Consecutive patients
Doppler. plethysmography, foot volumetry
Neglén 1993
Consecutive patients
Doppler, phlebography
Iafrati 1994
Consecutive patients with varicose veins
duplex
Class and sample size (n Iimbs)
Proportion of limbs with reflux (%) superficial deep only mixed only
Assuming that Porter's class was recorded prior to the performance of the diagnostic test (taken as the 'gold standard'), it is possible to calculate the sensitivity, specificity and predictive values of each of the three classes in cornparison to class O (Table 3.5). This analysis pools data reported in the Rosfors and the Neglén studies. Data in Table 3.5. show a decrease in the sensitivity for detecting superficial reflux from class 1 to 3. This result is consistent with the fact that skin changes and ulceration are
mainly indicators of deep rather than superficial reflux. All three classes have a good sensitivity for detecting deep venous reflux. but a high specificity is found only in classes 2 and 3. Globally, this analysis shows that classes 2 and 3 have comparable performances for estimating the presence of superficial and deep venous reflux. Porter's classification is limited in the investigation of patients with varicose veins because they can be faund in the three classes (as show" in Table 3-4). Since none of
@
the classes allows the clear distinction between subjects with and without varicose veins. a specific classification for the identification of varicose vein categories of various severity would therefore be needed. Table 3.5 Validity of categories of Porter's class as indicators of superficial and deep venous reflux (pooled data from Rosfors et al, 1990 and Neglen and Raju, 1993) Porter classes cornpared
Overall prevaience of reflux
Sensitivity
a. superficial reflux 1 versus O 225 2 versus O 146 3 versus O 139
0.53 0.42 0.40
b. deep reflux 1 versus O 2 versus O 3 versus O
0.21 0.34 0.35
Sample sire (1 imbs)
225 146 139
Specificity
Positive predictive value
Negative predictive value
0.80 0.61 0.57
0.70 0.87 0.89
0.76 0.77 0.78
O. 73
0.81 0.82 0.81
0.50 0.93 0.98
0.30 0.85 0.95
0.91 0.91 0.91
O. 76
0.76
The 'NHS' classification (1992) In order to offer the National Health Service (NHS) of the United Kingdorn a means of summarising the wide variation in type and severity of varicose veins, and provide a basis for understanding of the population's health care requirements, Robbins et al. (1992)developed a classification of asymptomatic. mild. moderate and severe varicose
veins (Table 3.6). The classification is based on Porter classes of chrontc venous insufkiency and addresses the limitation of the Porter classification regarding the investigation of varicose veins. Based on literature review, the authors have estimated the distribution of asymptomatic, mild. moderate and severe varicose veins at about 45%. 45%. 8% and 2% respectively.
Although the 'NHS' classification provides a means of classifying varicose veins. it mixes signs and symptoms and includes other signs of venous disorders in the definition of the severity of varicose veins. The four categories are probably not 'allinclusive'.
To our knowledge, the 'NHS' classification has not been used in
epidemiological studies.
Table 3.6 The 'NtlS' classification of varicose veins (Robbins et al., 1992) asyrnptomatic : few, isolated raised palpable veins which are not associated with any pain or discomfoit, or any skin changes. The main problem is cosmetic anxiety. mild :
varicose veins associated with ankle swelling, feelings of heaviness pain or other discomfort. with local or generalised dilatation of subcutaneous veins.
moderate :
varicose veins more likely to be associated with skin changes but not actual ulceration. The symptorns described above are present. with prominent local or generalised dilatation of subcutaneous veins.
severe :
varicose veins are associated with chronic teg pain, ulcerative and pre-ulcerative skin conditions, and there is generally deep venous incornpetence or obstruction.
The CEAP classification (1 995)
The combination of new insights into the pathophysiology of venous disorders. development of new tests, and misunderstandings regarding the use of the term 'chronic venous insufficiency' stimulated an atternpt at a new classification system based on objective clinical signs of chronic venous disease according to 7 classes
(Table 3.7). Table 3.7 Clinical categories defined in the CEAP classification (PoRer et al., 1995) -
-
-
class O :
no visible or palpable signs of venous disease
cfass 1 :
telangiectasia or reticular veins
class 2 :
vancose veins
dass 3 : class 4 :
oedema withlwithout telangiectasia, reticular veins or varicose veins skin changes ascribed to venous disease e.g. pigmentation, venous eczema, lipodenatosclerosis oedema withtout telangiectasia, reticular veins or varicose veins
class 5 :
skin changes as defined above with healed ulceration
class 6 :
skin changes as defined above with active ulceration
-
Clinical categories of the CEAP classification are supplemented with information on the mode of presentation (asymptomatic or symptomatic lirnb). etiology, anatomical
distribution and pathophysiological dysfunction.
The originality of the CEAP
classification has been to clearly dissociate signs. symptoms and results of laboratory tests Therefore. it avoids the difficulties of confusing morphological and functional
terms. However, the feasibility of its use in clinical practice is still to be evaluated and it still does not provide a simple scale for assessing the severity of varicose veins. For example. the presence of venous oedema alone is assumed to represent a more severe condition than the presence of varicose veins alone. irrespective of their size. It is not possible to distinguish among patients with oedema (class 3) . skin changes (class 4) or ulcer (classes 5 and 6) those who have from those who do not have varicose veins.
Ad-hcc classifications
Some authors have developed their own classification system for investigating varicose veins. O
Weddel (1969) used three categories combining the patient's complaint and the observer's examination: 'no vancose veins' (category A): no complaint of varicose veins and none visible to the observer: 'non-clinical varkose veins' (category 6): varicose veins visible to the observer and not fulfilling criteria of category C: 'clinical
varicose veins' (category C): varicose veins noticed by both the subject and the observer which were associated with two or more other symptoms, or with signs of venous insuficiency. This classification has not been used by other authors. a
Jakobsen et al. (1979) and Large (1985) evaluated the effect of treatments of varicose veins by using a classification of varicose veins with four categories : A - no visible varices; B
-
limited varices 25 kg/mmen under average over women under average over
1986 Widmer 1978
'
1.2 *(ideal weight) > 1.2 *(ideal weight)
The sample size corresponds to the number of penons wilh a follow up of 2 years. and the 'numben of subieas with varimse veins' are nurnbers of events having ocairred over a two-year observation time. ' Odds-ratio adjusted for age: relative weigM calculated as Broca index (weight in kg/height in cm, minus 100): only tnink vancosities are considered in this analysis Relative nsk estimate (incidence density ratio) and 95% CI calculated from the data presenled Relative risk estimate (odds-ratio) and 95% Cl (Woolf's rnethod) calwlated from the data presented Relative risk estimate (incidence density ratio) provided by the authors; number of events not stated
Genetic Predisposition Çeveral arguments favour a genetic predisposition to varicose veins. The fint argument cornes from studies that investigated the role of hereditv by collecting data regarding the prevalence of varicose veins in relatives of affected individuals (Table 3.13). Only two of these studies (Cornu-Thenard et al.. 1994: Weddel. 1969) were based
on clinical examination of relatives. The number of cases in Weddel's study was too small to reach any conclusion. In the other study. Cornu-Thenard et al. clinically examined the parents of 67 cases of varicose veins and 67 spouse controls. They found that both the mothers and fathers of cases were more likely than the parents of controls to have varicosities and a history of treatment. Other studies have also suggested an increased prevalence of varicose veins in patients with the condition. even after adjustment for other factors associated with varicose veins, including age, gender and obesity (Scott et al.. 1995). Based on detailed information on relatives of 250 patients with varicose veins.
Gundersen and Hauge (1969) also concluded that the risk of varicose veins was higher in those with affected relatives, especially if the relative was a male. Table 3.13 Case-control analyses about the association between heredity and varicose veins (VV)
1
Study
Measure of exposure
Nurnber of subjeds cases controls
Exposure (%) in cases controls
OR (95% Cl)
!
1
Cornu-Thénard 1989'
W in mother W in father
Cornu-Thenard 1994.
W in mother W in father
Hirai 1990 "
W in relatives
Laurikka 1995
W in relatives
Mekky 1969 a
W in relatives of women
Sadick 1992 a
W in relatives W in relatives
W in relatives
1 Wdmer 1978 ''
W in relatives
* OR adjusted for the effects of age. sex, history of phlebitis and body m a s index
- OR adjusted for the effed of age " Relative risk estimate (odds-ratio) presented in the article
and 95% CI (Woolfs method) calculatecl from the data
The second argument for the genetic predisposition to varicose veins cornes from results of studies that have suggested that ethniciti rnay be important for the development of varicose veins. A suwey in Jerusalem showed that immigrant men bom in north Africa had a significantly lower age-adjusted prevalence rate for varicose veins than immigrants from Europe. Arnerica and Israel (Abramson et al., 1981). Prevalence has also been shown to Vary among wornen from different parts of Europe (Guberan et al.. 1973) and among different ethnicities in New-Zealand (Beaglehole et al., 1975: Beaglehole et al.. 1976).The results of a study on cadavers is interesting. because it showed that black Africans have a higher number of valves than Caucasians when wmparing the same section of leg vein (Banjo. 1987). This result suggests that the number of valves is an inhefited factor Mering between races and thus different parts of the world. A contradictory argument against a primary genetic cause of varicose veins. however. is the case of black Americans. Whereas the prevalence of varicose veins in Africans is generally
accepted as being lower than that in Westerners. the prevalence is equal among black and white Americans (Cleave, 1959). Immigrants from an area of supposedly low prevalence have thus acquired the high prevalence of their adopted country, suggesting an environmental rather than a hereditary cause. If the greater number of valves in black Africans was responsible for their protection from the development of varices, then the prevalence would not be expected ta have increased, despite their change in environment. The results of Cleave's study must, however. be interpreted in view of the conclusion drawn regarding the geographic variation in the prevalence of varicose veins : the high prevalence of varicose veins in black Africans living in the United-States may result from a better
detection: on the other hand, the hypothesis of a multifactorial origin of varicose veins would suggest that the number of valves is not the only etiological factor. Clearly. there is a need for substantial research to c l a m the relative roles of genetic and environmental factors in the occurrence of varicose veins.
Standinci position
The investigation of standing position as a risk factor of varicose veins is affected by the unknown validity of the methods of assessment in relation to adual standing behaviour over a period of time, and by the difFiailty of detemining whether achial standing behaviour is a cause rather than a wnsequence of venous disorder. Thus the results of many studies must be viewed with scepticism. The role of workplace conditions, and more especially of standing at work. has recently been reviewed by Hobson (1997). Several studies have found an association between standing at work and the prevalence of varicose veins in univariate analysis. but other studies came to the opposite conclusion. In the Framingham Study (Brand et al.. 1988), the two year incidence of varicose veins was higher with the length of time women spent sitting or standing. An important study is the study of 2,000 worhers in three French cities where a wide variety of occupational risk factors have been scnitinised. (Sobszek. 1993). Of al1 work factors ascertained, the only factors found to be significantly different between
the venous insficiency and non-venous insufficiency groups were exposure to vibration (less likely in the former group) and work in ambient temperature of more than 19°C (more likely in the former).
Raised intra-abdominal Pressure The role of raised intra-abdominal pressure and of factors that induce intra-abdominal pressure has been put forward by many authors over the last 40 years. In theory. raised intra-abdominal pressure might promote the development of varicose veins but the evidence presented so far is inadequate. Tight undergarments, for example. have been incriminated. In two studies, the prevalence of varicose veins increased with conet-wearing (Abramson et al.. 1981) and the tightness
and stiffness of the undergamient being wom (Mekky et al.. 1969). In Swiss women. the association between corset-wearing and varicose veins was however lost after correcting for age (Guberan et al., 1973). It has also been postulated that iow fibre diet produces small. hard stools that may lead to constipation. repeated increases in intra-abdominal pressure and obstruction of venous
retum from the legs, ultimately leading to the development of varicose veins (Burkitt. 1972: Cleave, 1959). Alternatively, raised intra-abdominal pressure may be lransmitted down the veins of the legs, leading to dilation of the veins and thus rendering the valves incompetent (Burkitt, 1975). A positive risk factor for the development of vaficose veins was constipation in a study in Sicily (Novo et al., 1988) and ltaly but it was negative in other studies (Canonico et al.. 1998; Mekky et al., 1969; Kakande. 1981). Results of such studies are. in any case, of dubious value because of the difficulty in measuring constipation in a valid and reliable way. Squatting to defecate may provide a mechanical protective mechanism for the leg veins helping to prevent the rise in intra-abdominal pressure being transmitted down the veins. Burkitt (1975) argued that this protective mechanism is lost with the use of raised toilet
seats. but the little available evidence does not support this hypothesis (Maffei et al.. 1986: Richardson and Dixon. 1977). The pradice of chair sMng is another Western habit not adopted by primitive wmmunities. Alexander proposed that sitting in a chair increased the hydrostatic pressure exerted on the leg veins, leading to pooling of blood in the legs. venous dilatation and increased tension in the vein wall itself (Alexander. 1972). The effect
of chair-sitting during childhood wuld therefore produce veins with an increased diameter which would bewme more susceptible to factors sudi as pregnancy. There are no data to support this hypothesis.
Other factors Other factors have been implicated in the development of varicose veins (Evans et al.. 1994). These include height (Beaglehole et ai.. 1976: Sisto et al.. 1995; Stewart et al.. 1955) and iack of exercise and cigarette smoking, which were both related to incidence in
the Frarningham Study (Brand et al.. 1988). except for smoking in women. Varicose veins have also been associated with heavy liffingand rnguinal hemiae in males (Abrarnson et al.. 1981), and a history of k g injury (Willén et al., 1982; Lindhagen et al.. 1985; Aitken et al.. 1989; Scott et al.. 1995). Hormone leveis could be relevant. as varicose veins have been
related to late menarche (Amoldi. 1957). hypogalactonhea (Amoldi. 1957) and the use of oral contraceptives (Sadick. 1992). DHerences in AB0 blood groups (higher risk in subjects
with blood group A) have also been implicated (Cornu-Thénard et al.. 1989). The evidence linking these factors to the occurrence of varicose veins is tenuous. and in rnany cases may
be related more to confoundirig by other factors than to a direct causal association.
3.2.1 .-issociated signs and clinical complications Prevalence of sians associated with varicose veins In a population-based survey of variwse veins in Brazil (Maffei et al.. l986), the cornmonest manifestation, occurring in nearly 20% of subjects was oedema. Hyperpigmentation was present in about 6% of subjeds and ulcer (active or healed) in about 4%. The communitybased study in Tecumseh, USA. found that 10% of men and 20% of women with varicose veins had oedema (Coon et al.. 1973). In the Basle Study (Survey II),the prevalence and severity of various grades of clinical chronic venous insufficiency was related to the severity of varicose veins (Da Silva et al.. 1974). The prevalence of chronic venous insufficiency was 81% among those with severe varicose veins (marked trunk varices), and 30% among those with minor varicose veins (reticular veins. telangiedasia or "scarcely visible tmnk varices"). The prevalence of skin changes and ulceration increases when patients with variwse veins who have been referred for clinical assessment are considered (Jakobsen. 1979; Sethia and Darke. 1984;Struckman. 1986a; Struckman, 198613: Struckman et al.. 1986c; Gutman et al.. 1989; Vasdekis et al.. 1989: Almgren and Eriksson. 1990: Zukowski et al.. 1991. Hirai et al.. 1991: Christopoulos et al.. 1991). In these studies. the proportions of affected subjects who have evidence of skin changes and ulcer range from 18% to 74% for skin changes and from 2.5% to 1456 for ulcer. In one study (Gutman et al. 1989), the prevalence of ulcer associated with varicose veins was 39.8% in patients over 70 years old.
Incidence of skin chancles and ulceration Very few data are available about the incidence of skin changes and ulcer in subjects with
varicose veins. The incidence of skin changes and ulceration among subjects with varicose veins in the population has been reported in the Il-year follow-up of chernical workers in the Basle
@
Study (Widmer et al., 1992). Table 3.14 shows that the percentage of subjects developing these complications depended on the sevarity of varicose veins About 50% and 20% of subjects with severe variwse veins developed CVI and ulcers over the 11-year follow-up. Table 3.14 Incidence (%) of chronic venous insufficiency (CVI), oedema and ulcers arnong subjects with varicose veins 0 followed for 11 years (source: Widmer et al..
W at baseline none mild moderate severe
CVI 8 21 34 50
Oedema 10 17 19
Ukers 0.2
30
20.0
0.8 1 .O
Varicose veins : a risk factor for venous thrombosis ? The association between varicose veins and the occurrence of deep vein thrombosis is controversial. Some studies have shown deep vein thromboses in limbs with phlebitis (Jorgensen et al.. 1993; Skillman et al.. 1990). but these thromboses have mostly occurred in circumstances of special risk . such as after an operation or illness (Skillman et al.. 1990: Lutter et al.. 1991) or with a previous history of deep vein thrombosis (Lutter et al.. 1991: Bergqvist and Jarozewski. 1986). In an editorial of the Lancet. Campbell (1996) concluded that there appears to be no evidence that varicose veins is an independent risk factor for thrombosis or phlebitis. Although few data are available. the risk of occurrence of thrombosis or phlebitis would not be higher in people with varicose veins. Varicose veins : a risk factor for cardiovascular disease ? In the Paris study carried out on 7432 male policemen aged 42 to 53 years with a 6.6year follow-up period (Ducimetiere et al.. 1981), coronary heart disease (CHD) occurred one and a half times more frequently in those with varicose veins as compared with the others: this difference was independent of smoking. blood pressure. cholesterol. diabetes. age and BMI. The authors speculate that the mechanism of this association could be a disturbance in the fibrinolytic system that could be at the basis of both varicose veins and arterial disease. In the Framingham study (Brand et al.. 1988), the
incidence of CHD was also higher in subjects with venous disorders; however. after adjustment for BMI and systolic blood pressure, the difference was no longer significant. Thus, although the presence of varicose veins may indicate an increased risk for arterial disease, the association is probably mainly owing to coexistent physical inactivity. obesity or disturbances in the clotting mechanisms and to failure to adequately control for these (De Backer. 1997). 3.2.5
Conclusions
The following conclusions can be drawn from a literature review on the epidemiology of
varicose veins. The prevalence of al1 categories of vafiase veins in western adult populations is in the of
25% to 35% for wornen and 10% to 20% for men. About half of these would consist of those with varicose veins associated with symptoms or more severe signs of venous disorder. The prevalence increases with age. and has a sex ratio of about 23 in favour of women. 6
Despite the fad that many authon have drawn attention to the possibly multifactorial etiology of varicose veins, at the time when the VEINES study was started only the Framingham study had used multivariate analysis to adjust the effect of one determinant for the effect of other suspected determinants. A few other studies have been published since (Scott et al.. 1995; Sisto et al.. 1995). The risk of acquiring varicose veins increases with number of pregnancies and obesity in women. but there seems to be no increased risk of obesity in men. and the association in women is weak.
Genetic
predisposition. lifestyle associated with westemisation and other factors may be biologically plausible as risk factors. but epidemiological evidence is lacking. A large proportion of subjects with varicose veins have oedema and skin changes. In a 11-year follow-up of patients with varicose veins. the incidence of chronic venous
insufficiency and ulcer were 50% and 20% respectively. There is no evidence that the presence of varicose veins is a risk factor for venous thrombosis. There is no evidence that the presence of varicose veins is a risk factor for cardiovascular diseases.
It is now widely racognised that patient-based outcome measures are necessary for a comprehensive evaluation of burden of disease. This shift in attitude is taking place in al1 areas of medicine with good evidenœ that patients can provide reliable and valid judgements of outcomes (Fitzpatrick et al., 1992; Anonymous. 1995: Ware. 1993; Lamping. 1997). The burden of disease for a patient can be physical, and translate into various
symptoms of differing intensity and frequency.
It can also impact on more general
dimensions such as ernotional status. social fundioning, mental health or vitality. An essential requirement for
assessing patient-based outcomes is to ensure that the
information collected is acwrate, credible and useful. This can be achieved by using measuring instruments that are standardised. scientifically sound (e.g.. reliable. valid and responsive) and practical to use. Methods whtch have been used to measure symptoms and quality of life in studies on venous disorders have been recently reviewed by Lamping for the International Task Force on Chronic Venous Disorders of the Leg ( Lamping and Shroter. 1998). Results pertaining to varicose veins are summarised here. 3.3.1 Symptoms
Data on symptoms in patients with varicose veins can be derived from three types of studies : epiderniological studies in which individual symptoms were directly recorded by the investigators. or from the patient using a self-administered questionnaire (e.g. Bradbury et al.. 1999: Laurikka et al., 1993: Widmer. 1978); a
diagnostic test studies where symptoms were investigated in patients with varicose veins referred for pre-surgical evaluation and assessrnent of venous incornpetence (e.g.
Vasdekis et al..1989: lafrati et al.. 1994: Struckman and Mathiesen. 1985: Rutgers et al.. 1993): O
clinical trials of the efficacy of compression (e.g. Fraser et al.. 1985: Struckman. 1986a: or surgery on varicose veins where relief of Snickrnan. 1986b; Struckman et al., 1986~)
symptorns was used as one criterion of efficacy (e.g. Jakobsen. 1979; Sarin et al.. 1992a; Perrin et al.. 1993). Symptoms were rated either by an observer (clinician) rating or by the patient himself. In some studies, patients were asked to give global ratings of improvement in symptoms (Ernst et al.. 1990; Muschietti. 1978; Prerovski et al.. 1972; Burnand et al.. 1980). In other studies, quantitative scales were used to assess symptoms. The types of scales used to evaluate symptoms include Likert scales (a-g. 3-pt scale. O=no symptoms.
3=maxirnum intensity: 7-pt scale, -3=aggravation. +d=amelioration) (Laurent et al..
1988;Languillat et al., 1986; Lambelet, 1973; De Jongste et al., 1989; Camilli and Guarnera, 1994; Roeckaerts and Vanden Bussche, 1987; Widmer et al., 1990; Smith et al.. 1992). and visual analogue scales (VAS) (100 mm lines with endpoints defined as O symptom-free. 10 highest severity) (Zucarelli, 1987; Anderson et al.. 1990: Delecluse et al., 1991; Renton et al., 1994). Studies Vary in how many of the symptoms listed below were.measured: some studies measured only a few key symptoms. whereas others measuied several.
There was
also variation among studies in whether symptom scores were combined to f o n symptom indexes or remained separate.
For example. in some studies symptom
indexes were created by surnming scores on individual symptom scales (Vin et al.. 1992: Fraser et al.. 1985: Widmer et a1..1990).
Other studies derived weighted
symptom scores from a combination of syrnptom intensity and the length of time between walking and manifestation of the symptom (Vin et al.. 1992; Zucarelli et al.. 1988; Vin et al.. 1994). There have been no studies to evaluate the reliability and validity of these rnethods of measuring symptoms.
Due to the variety of symptoms measured. the use of different methods of measurement and the lack of data about the reliability and validity of measuring instruments. only a limited
number of conclusions c m be drawn about the nature and frequency of symptoms in patients with varicose veins and venous diseases in general:
symptoms commonly measured in studies on venous diseases of the leg include: pain. heaviness, fullness, swelling, discomfort, itching, cramps, aching legs. sensation of heaüburning, restless legs, paresthesias, tired legs, skin irritation. discomfort during prolonged standinglsitting, and the urge to elevate the legs; patients may have syrnptoms in the absence of clinical signs of venous disease; the severity of symptoms may not match the severity of clinical signs of venous
disease (Bradbury et al., 1999; Neglén and Raju, 1993: lafrati et al.. 1994: Gooley and Sumner, 1988; Milne et al., 1994; Rosfors et al., 1990). The association of symptoms with varicose veins can be assessed from several studies (Bradbury et al., 1999; Jakobsen. 1979; Sarin et al., l992a: Stnickman. 1986a: Struckman. 1986b; Struckman et al.. 1986~).In most of these studies. very large proportions of symptorns were found. from 40% for itching (Struckman. 1986b) to 88% for swelling sensation (Struckman. 1986a). Since signs of severe venous disorders (oedema. hyperpigmentation. eczema. induration) were also present in proportions ranging from 40% to 80%. it cannot be ascertained whether the large prevalence of symptoms is specifically associated to varicose veins or to more severe venous disease. Results published by Bradbury et al. (1999) are based on a population survey. In the Edinburgh Veins Study. lower limb symptoms such as heaviness. tension, aching, feeling of swelling. restless legs. cramps. itching, and tingling. were extrernely common in the general population. whether or not varicose veins were present. with a prevalence up to 54.7% in women for heaviness and up to 37.6% in men for itching. These symptorns tended to increase with age and were significantly more wmmon in women than in men. The authon wncluded that the poor agreement behveen symptoms and signs in patients with varicose veins raises questions about the venous origin of some symptoms.
3.3.2 Quality o f lire
Comprehensive assessment of quality of life (QOL) in venous diseases of the legs should include both generic and disease-spectfic measures which are known to be reliable. valid
and responsive. Quality of life in venous disease has not been adequately evaluated in
these respects (Lamping, 1997). Studies of the impact of venous disease and its treatment on QOL are scarce, and only a few have specifically addressed varicase veins. Generic Qualitv of Life Measures The Medical Outcornes Study Short Form Health Survey (SF-36) is a 36-item generic measure of QOL which includes 36 items measuring eight health dimensions related to physical functioning, social functioning and mental health. It has undergone extensive psychometric testing in North America, the United Kingdom and Europe and is translated into several languages (Ware et al.. 1995). The SF-36 has been used in three studies in the UK to assess QOL in patients with
varicose veins. Garratt et al. (1993a) evaluated QOL in a postal survey of 281 patients with varicose veins reauited from surgical outpatient departments, and to 542 controls randomly selected from the general population. Both the SF-36 and a clinically-based disease-specific QOL measure were used.
QOL as rneasured by the SF-36 was
significantly lower in varicose veins patients than in the general population sample after adjusting for age and a lower proportion of women. In a second postal survey. Garratt et ai. (1993b) used the SF-36 to evaluate QOL in 1310 patients in four diagnostic groups recruited from outpatient departments and general pradices (278 varicose veins patients, 558 low back pain patients. 271 menorrhagia patients. 203 patients with suspected peptic ulcer). and in 542 respondents from the general population. QOL was lower in variwse veins patients compared to the general population but higher than in patients with low back pain. menorrhagia. and suspected peptic ulcer. Results also showed that patient-reported QOL as measured by the SF-36 was closely related to general praditioners' perceptions of severity, and was lower in patients referred to specialists than in patients not referred.
In order to assess the outcarne of varicose vein surgery, SF36 questionnaires were sent to 150 patients presperatively. and at one and six months post-surgery (Baker et al.. 1995).
Eighty-ni ne (59%) patients answered all three questionnaires.
Presperatively. most
varicose veins scored higher than UK standard population on general health perceptions and lower on pain lncreased pain and reduced rote function were reported at 1 month
post-surgery. By 6 months post-surgery, improvement was s h o w in al1 dimensions except social function and health perception. This study showed good responsiveness to change in the SF36 scale. Drawbacks of the study are the low response rate, the la& of a diseasespecific instrument and the lack of a control group. with the possibility of a placebo effect. The Nottingham Health Profile (NHP) is another standardised QOL measure which includes 38 items covering six domains (physical mobility, pain. sleep. social isolation. emotional
reaction. and energy). It has undergone extensive psychometric testing and is widely used (Hunt et al.. 1980). Tho NHP has been used to assess QOL in one study of patients with chronic venous insufficiency (Augustin et al., 1997) and in three studies of ulcer patients (Lindholm et al.. 1993; Harner et al.. 1994; Franks et al., 1995). Disease-Specific Qualitv of Life Measures
Six disease-specific measures have been developed to assess QOL in venous disease. Three were developed for use with patients with symptoms associated with venous disease (Franks et a1..1992b) or chronic venous insufficiency (Launois et al.. 1994: Augustin et al.. 1997). One was developed for use with patients with varicose veins (Garratt et al.. 1993a).
and two were developed for use with patients with ulcers (Franks et al.. 1994; Hyland and Thomson. 1994).Two other interview-based methods have been used to measure QOL in leg ulcer patients. including stnrctured (Phillips et al., 1977) and qualitative interviews (Charles. 1995). Questionnaires developed specifically for patients with ulcer are not reviewed here.
Questionna;re to measure quality of life in venous disease (Franks et al.. 1992b) This is a 36-item questionnaire designed to measure QOL (symptoms. psychiatric rnorbidity). risk factors. costs and symptoms in patients with venous disease. Symptoms assessed include: swelling of one leg, swelling of both legs. night uamps, itching, restless
legs at night, pain in legs. heaviness of legs. Psychiatrie morbidity is assessed by the Symptom Rating Test. with s a l e scores for depression. anxiety, hostility, subjective cognitive function. somatic symptoms of psychological distress. Costs (use of medical and
social services) and risk factors are also assessed. QOL was assessed in a case-control study where 114 patients, with venous incompetenœ confinned by Doppler scanning and reuuited from three general practices, were matched by age and sex to 114 wmmunity subjects. Cases reported significantly more symptoms than controls. but quality of life measures (e.g. psychiatric morbidity) did not differ between the two groups. The authors suggest that selection bias in the control group (participation rate of only 27% among the community subjeds approached) may have resulted in a more health anxious control group with higher than usual psychiatric morbidity scores.
The reliability, validity and
responsiveness of the questionnaire have not been evaluated. and there are no published data about its use in patients with varicose veins.
Questionnaires to measure quality of life in chronic venous Nisufficiency The Chronic Venous lnsufficiency Questionnaire (CVIQ) is a 20-item questionnaire that
assesses psychological wnsequences. physical consequences. pain. and social consequences of cironic venous insufficiency (CVI) (Launois et al.. 1994: Launois et al..
1996). lt was tested in a survey of 1.001 patients with clinicallydetected CVI and in 934 patients from a randomised clinical trial of a venoadive drug. There is good evidence for the reliability and validity of the CVIQ. It has good intemal consistency and test-retest reliability, has been shown to differentiate CVI patients from non-CVI patient controls. is sensitive to changes in QOL as a function of the severity of pain. and it shows good factorial validity. The CVIQ has also been shown to be responsive to treatment following 2 months of treatment with a venoactive drug, with changes in quality of life correlated with changes in
clinical measures. It is available only in French and has not been specifically tested in patients with varicose veins.
The Freiburger Questionnaire of QOL in venous diseases (FLQA) is a Gemian questionnaire that includes 83 items and differentiates between limitations to QOL in seven areas : physical complaints. everyday life. social life. emotional status. therapy. satisfaction. occupation (Augustin et al.. 1997). It includes general questions taken from the SF-36. the
NHP and the ALLTAG (a Geman questionnaire to detemine stress factors in QOL). To
assess its validity, it was useâ in 246 patients with chronic venous insufficiency in stages 1, II and III in the Widmer classification. Its convergent validity was measured in cornparison
to the Nottingham Health Profile, the ALLTAG and the German version of the Symptom Check-List-90 R, with good results obtained for intemal consistency, convergent validity. discriminant validity, sensitivity to change and acceptance. It is available only in Geman and has not been specifically tested in patients with varicose veins.
Questionnaire to measure qualdy of Me in patients with vancose veins In their study of QOL in variwse veins patients reported above, Garratt et al. (1993a)also used a clinically-based disease-specific measure to assess QOL in addition to the SF-36. It
is a 15-item questionnaire wntaining questions wmmonly used in the clinical assessment of patients with varicose veins, e.g. distribution of veins. duration of pain, tirne of day with worst pain. degree of ankle swelling, use of support stockings. use of diuretics, extent of itching, discoloration. rash or euema, skin ulcer. degree of concern at appearance. influence on choice of clothes, interference with work. interference with leisure, There is some evidence for the reliability and validity of the questionnaire. It shows good intemal consistency as measured by item-total correlation. Factor analysis identified four important health fadon: pain and dysfundion, cosmetic appearance. extent of variwsity. complications.
The validity of the questionnaire has been demonstrated by a high
correlation with the SF-36. This questionnaire is avaiiable only in English. 3.3.3 Conclusions
Svm~toms A large number of symptoms have been measured in studies carried out in patients presenting with varicose veins or venous disorders of the legs. but there has been no standardised and psychometrically-sound way of summarising the burden of symptoms:
scales that have been used have not been psychometrically evaiuated in tems of reliability. validity and responsiveness. O
Symptoms reported by patients with venous disease are also reported by a large proportion of non affected individuals.
Few studies have systemtically examined the prevalence of symptoms in individuals with varicose veins alone, and the conditions in which these symptoms are alleviated or aggravated. a
There are limited data on whether symptoms are associated with the severity of varicose veins.
O
There is limited information on the determinants of symptoms in patients with varicose veins.
Qualitv of life Only few studies have evaluated qualityof-life in patients with varicose veins. Disease-specific quality of life scales have generally not been comprehensively evaluated in ternis of reliability, validity and responsiveness. ln one uncontrolled study, SF-36 has been found responsive in detecting changes in quality of life after surgery for varicose veins. There are no data on the impact of varicose veins on quality of life after adjusting for the effect of other quality of life deteminants.
3.4 Health resources utilisation 3. J. 1 Introduction
Venous disorders are a common reason for interaction with health care systems and use of health care resources. Several aspects must be considered in the evaluation of the economic impact of a disease on both health a r e systems and society: the type of costs considered. the analytical perspective. the time horizon. the sources of data on resource use and the pricing of physical units (Drummond et al.. 1989). Of particular importance in a study on health resource utilisation is the measurement of relevant units for the disease considered. An exhaustive review of available studies and reports on the economic impact of venous
disorders in industrialised countries has been recently perfonned in an International Task Force on the Management of Chtonic Venous Disorders of the Leg (Abenhaim et al.. 1999. Chapter 6).Both individual data. e . g the cost of procedures for a given individual. and
collective data. e.g. individual cost data entered in a population model. were examined. Of the total number of papes reviewed and retained by the Task Force. only 24 provided some information on costs or resource use by patients with venous disorders. In this review, we only consider costs associated with varicose veins. It is noteworthy that most of these studies were undertaken before guidelines for use in phanacoeconornic studies were issued (Drurnmond et al., 1989). Therefore. the costs included and the costing methods varied. Moreover, the studies were al1 cross-sectional and expressed estimated expenditures for venous disorders over a given year.
In this
review, no attempt was made to adjust costs in order to compared diredly berneen countries, even though medical inflation rates Vary frorn one country to another.
3.4.2 Costs associated with the treatment of varicose veins
Seven studies estimated the cost of treating varicose veins. The study done by Lafuma et al. (1994) used a cluster sample of physician prescriptions and hospital discharges. The
study commissioned by the National Health services (NHS) in the United Kingdom used discharge data (Robbins et al.. 1992). Van den Oever (1998) estirnated health resource utilisation in Belgium on the basis of medical care services which are payable by health insurance funds. Other studies used cost data that were wllected in the course of clinical trials or follow-up of case series.
Direct medical costs The total direct cost associated with the management of varicose veins in France was
estimated to be 2.75 billion French francs in 1990 (Lafuma et al.. 1994). The NHS survey (Robbins et al., 1992) estimated the ambulatory and hospital costs to be around £300.000 (1981) for a 250,000 individual district. which could be extrapolated to a total of E72 million for the entire population. Explanations for the lower costs in the UK as suggested in the
NHS report (Robbins et al.. 1992) include reduced availability of services and restricted use of surgical procedures in the UK. In Belgium. the total expenditures for health care for severe variwsis (including haemorrhoids) was estimated to 6.52 billion BEF (US $ 185 million) for the year 1995. which
would represent 1.1% of the 1995 national budget. This amount was divided as follows (in billion BEF): diagnosis (including oflice costs). 1.13; hospitalisations (total costs including surgery), 4.14; venotonic dnigs, 1.O;compression therapy. 0.25 (Van den Oever et al.. 1998). Medical costs of different treatment modalities were also estimated in clinical trials and followvp of cases series. Costs computed in these studies are costs per procedure. They
do not include recurrent costs in case of treatment failure and therefore provide an incornplete estimation of the cost of a complete treatment course. lnjêction sc/erofherapycompressionwas estirnated to cost US$ 100 to 200 per treatment
course. The explanation for the lowest cost figure (Beresford et al.. 1978) is the inflation factor (the cost was estimated in 1977) and the fact that it only included the cost of labour of health care professionals and of supplies in hospitalised patients.
The cost of
ambulatory sclerotherapy in a hospital setting was estimated at US$ 200 to 220 (1986) per treatment course completed (i.e. from 3 to 5 procedures per patient). This cost included the cost of labour and supplies but excluded the cost of tests (Belcaro et al.. 1991). In a trial of compression stockings vs. conventional bandaging technique after sclerotherapy in the UK the direct hospital costs were respectively £ 6 and £ 8 per patient (Scurr et al.. 1985). Inpatient surgery was estimated to cost US$ 450 to 1.500 per procedure. The lowest figure was estimated in 1977 (Beresford et al.. 1978). The highest figure was for private
practice and included hospital costs and surgical fees. In the UK. surgical procedures perforrned in a university hospital on an inpatient basis cost £643 (US $ 1.360) per procedure (Price and Makin, 1991). Outpatient surgery was estimated to cost US $300 per procedure in Sweden (Neglén et al.. 1986) and US $650 in the UK (Price and Makin, 1991).
0
Indirect costs Indirects costs were estimated by the nurnber of working days lost per year In a French study (Lafuma et a1.,1994), the cost wnesponding to the loss of 6 million working days per
year was of the order of magnitude of US$ five hundred million (1990)per year. In the United Kingdom, the cost per patient. considering a working population, varied from US $30 per person per year when the treatment was compression-scterotherapy, to US $600 when the patient undemuent surgery (Beresford et al.. 1978:Piachaud and Weddell, 1972).
4. Patients and Methods :the VEINES study
This work is based on data collected in the VEnous lnsufficiency Epidemiological and Economical study (VEINES), an international study on chronic venous disorders of the leg initiated in 1994 by the Center for Clinical Epidemiology and Comrnunity Studies. McGill University, Montreal (Abenhaim et al.. 1997). The epidemiological investigation of varicose veins was one of the main objectives of the VEINES study. and the study protocol made provision for the collection of data necessary for this investigation. Therefore. this chapter presents the aspects of the VEINES protocol relevant to this thesis. Specific rnethodological considerations are presented separately in the chapters where they apply.
4.2 Scope and objectives The goal of the VEINES study was to assess the burden of venous disorders in those affected. including the consequences on the patients' well-being, and also the economic effect on them. Since the instruments needed to measure these outcomes were either absent or poorly validated, the development and validation of such instruments was also considered an important objective of the study. The specification of the scope of the study emphasised the need to measure outcomes in an environment of current clinical practice. This required the recruitment of an unbiased sample of subjects with venous disorders as they were diagnosed and treated by physicians.
The main objective of the VEINES study was to describe and compare clinical
presentations of venous-related leg disorders with regards to natural history. quality of life. health care utilisation, risk factors and clinical outcome.
The secondary objectives were to develop or identify instruments for the study of venous-related leg disorders, and to test their measurement properties. such as a summary scale of the burden of symptoms, a quality-of-life measure and a tool for economic evaluation. The measure of the prevalence of venous disorders. and more specifically of the prevalence of varicose veins. was not an objective of the VEINES study.
It was
considered that a valid measure of the prevalence would add important constraints to the protocol and that these constraints would affect the fulfilment of the main objective. Moreover, the body of evidence regarding the prevalence of venous disorders in western countries was large enough to estimate the prevalence of venous disorders with sufficient precision. Since other prevalence studies were on-going at that tirne. to collect more data on the prevalence of venous disorders in the general population was not considered a priority.
4.3 Oevelopment and initiation of study 1.3.1 Protocol development
In addition to a comprehensive literature review, an important contribution to the developrnent of the protocol came from the scientific input from members of the international Steering Committee. This Committee was established in September 1993 in order to provide expert opinion on the objectives. design options and methodological
components of the protocol. The Scientific Comrnittee included methodologists and clinicians (angiologists. internists and surgeons) whose field of expertise covered the areas covered by the objectives of the study. Clinicians were requested to provide a clear representation of current knowledge on several aspects related to the design of the VEINES study. such as symptorns and signs presented by patients. severity scales, ciinical outcornes. treatrnent modalities. components of venous-related quality of life. diagnostic tests and existing survey
O
instruments.
A Research Team was established in September 1993 in order to translate
recomrnendations frorn the Steering Cornmittee into a draft study protacol.
The
development of the protocoi followed an iterative process of review, discussion and revision within the research group and in successive meetings with the Steering Cornmittee. Once the decision was made to conduct the study in several countries. the discussions also concerned the choice of countries to be considered. After finalkation of the choice of participating countries, local scientific committees were established in order to provide information on the feasibility of the protocol in local conditions. Since the principle underlying the conduct of the study in different countries was to use the
same protocol in various settings, in some cases views expressed by local committees were translated into proposals that were adopted as elements of the general study protocol to be applied in al1 countries.
4.3.2 Initiation of study
In the first phase. local collaborators were recruited and contacts were made in order to identify local staff and a research centre. Local CO-ordinatorswere in charge of submitting the study protocol to the relevant local ethical committees, to plan the work of the local staff. to recruit investigators, to organise training sessions and to visit investigators who could not attend the training sessions. Initially. the recruitment of general practitioners and specialists was planned to start at the same tirne. in March 1994. Following the Steering Committee's decision to collect a set of specialist-specific data. the recruitment of specialists was delayed in order to adapt the questionnaires and enlarge the recruitment base. These steps were made from September to December 1994.
4.4 Study design 4 . 1 Rationde for the choice of a cohort p o p d i t i o n
Several objectives pursued by VEINES implied repeated measurement of outcomes in the same individuals.
A first reason was related to the need of measuring changes of some outcomes. for example a change of venous symptoms (in order to develop a rating system for assessing an improvement or a worsening of symptoms). a change of quality of life (in order to test the responsiveness of a new quality of life questionnaire), or the occurrence of complications of venous disorders. A second reason pertained to the validity of results.
In a chronic disorder, data
collected at several points in time can provide a more valid assessment of outcomes than a single measure. An example is the assessment of health care utilisation by subjects with varicose vains. This evaluation could have been performed in a single interview with retrospective ascertainment of the amount of health resources used in a defined period of time. This method would be prone to recall bias (patients with more severe venous disorders would better recall events such as investigations. consultations and treatments) and could miss rare events. In order to ensure suffcient quality of data. a retrospective assessment needs to address a short period of observation (e-g. three months). A prospective design allows the collection of more accurate information by ensuring more complete ascertainment of cost units with short delays between each evaluation. A third reason for repeated measurements was related to the amount of person-time of
observation and the number of events of interest that could reiiably be captured in the study. and therefore the precision of the estimates. for example of the mean number of consultations for venous disorders. Repeated rneasurements required the fol low-up of a population experience over time. This population could be closed (a cohoit)or open (a dynamic populatian). A cohort is a closed population whose membership is defined on the basis of some common event. for ever after (Miettinen, 1985. p.321). The common event may be the occurrence of the same disease. exposure to the same drug, consultation with the same physician. or the fulfilment of the same criteria conditioning entry into a study (Miettinen. 1985, p.48). A cohort is qualified as a "closed" population of individuals because. once
a person meets the admission criteria at a particular point in time to. helshe remains a member of it for ever after, even if helshe looses the characteristics defining the membership (Miettinen, 1985, p.49). Time to at which admissibility criteria are met can be hed (for example, a calendar date) or relative (for example, the date of an inclusion
visit). A dynamic population is an open population whose membership is defined on the basis
of sorne state, for the duration of that state (Miettinen, 1985, p.324). An example of a dynamic population in this context would have been the practice of one or several general practitioner(s) or specialist(s). The study of a change in outcome in that practice would imply that patients registered in the practice would be sampled at several points in tirne. A dynamic population was not the best choice for this study. Although it can be used to
measure incidence, a dynamic population is not appropriate to measure changes in the same individuals because there is no guarantee that the same individual will be
evaluated more than once, or that the duration of the state that defines hislher membership in the population is long enough to observe a change.
Moreover. a
dynamic population implies a turn-over of members. lt was not practical in the VEINES study to ask investigators to perform a full baseline assessment each time a new patient consults. as opposed to the use of a cohort population where baseline visits could be grouped in time. The use of a cohort population may also have drawbacks: for some physicians whose rural practice was spread out over a large geographical area. the study period was significantly extended for the time needed to contact and convince subjects to attend a follow-up visit. The same problem was observed by specialists whose patients lived far away and were not willing to travel a long distance to attend a fol low-up consultation. For these patients. the follow-up visit was sometimes delayed until the following regular visit occurred. A cohort population was therefore chosen as the type of population appropriate for the
VEINES study. The VEINES study population is thus a relative-time cohort with
a
admissibility determined at the time of the baseline visit to the investigator.
The VEINES cohort was used as the base for three types of studies : cohofi studies,
cross-sectional studies and case-control studies. A cohort study is a study whose base is the experience of a cohort over time (Miettinen.
1985. p.321). This experience allows a measure of the occurrence of clinical and nonclinical outcornes. and of changes in those outcomes. An example of a cohort study based on the VEINES population is the prospective assessment of health care utilisation. A cross-sectional study is a study whose base is a population cross-section (Miettinen.
1985. p.323). A population cross-section includes members of a dynamic or cohort population as they are exarnined at a particular point in time. A cross-sectional design was applied, for example. to the investigation of quality of life in people affected by varicose veins. A population cross-section may also be used for case-control studies with prevalent
cases (Rothman and Greenland, 1998, p. 113). In the VEINES study, an example of such case-control design is the analysis of risk indicators of varicose veins based on information collected by members of the VEINES cohort at time zero (the baseline examination).
4.42 Definition of source population
The source population (or study base) is the population experience that is used as a source of learning (Miettinen.1985.p.46). The source population is to be distinguished from the study population. which is the group of subjects for which observations have
been collected (Kleinbaum et al., 1982. p. 21). The first choice to be made regarding the source population concerned the type of
population to target as the source of subjects with venous disorders. Two possible options were considered : the general population. and a population of patients consulting physicians. The latter was adopted because it allowed a more accurate assessment of clinical outcomes. could be more easily followed and had advantages in
terms of feasibility. It was also best suited to the general principle of investigating clinical outcomes in a context of current clinical practice. The second choice concerned the type of physicians to be recruited as the source of subjects with venous disorders. In order to collect accurate data on the whole range of venous disorders and on various kinds of treatments, the ideal situation was to include investigators from al1 categories of health professionals wnsulted by patients for advice or treatment of venous problems without being necessarily referred by their general praditioner or specialist. The list of health professionals had to include general practitionen, physiotherapists and specialists such as gynaecologists, surgeons, phlebologists. angiologists, dermatologists. and cardiologists. Such an exhaustive rnethod
of recruitment would have been excessively timeconsuming and practically unrealistic. General practitionen are generally made aware by their patients of consultations made to other health professionals, but important infomation on self-treatment could nevertheless have been missed if they were the only source of infomation. The option taken was
therefore to restrict the recruitment of investigators to general practitioners and specialists frequently consulted by patients for venous disorders (angiologists, phlebologists and vascular surgeons), and to amplement data collected by the physician during consultations by directly interviewing the patient.
Once the choice of a physician-based source population was finalised, the next step consisted in detemining the method of identification of the source population. The use of existing medical records and a prospective registration of subjects consulting physicians were possible options. It was considered that using existing medical records would impede a complete identification of patients with venous disorders. Moreover. this method could lead to a biased source population since medical records are more likely to exist for patients with severe venous disorders. A prospective registration of patients was therefore chosen as the method of choice for the identification of the source population.
A source population has üma and place characteristics Given the choice of a prospective registration of patients. the time parameter was inherently defined by registration dates. In the development phase of the protocol, there was a concem that, according to expert opinion, medical practices and utilisation of health sewices in the field of venous disorders varied from country to country. lmplementing the study in a single country would imply that only a restricted range of consequences of venous disorders could be examined.
An example that illustrates such a variability is the extent of use of
venoactive drugs : none of them are authorised in Quebec, and different reimbursement policies in European countries impact on their prescription by physicians. Other factors. such as clirnatic conditions, could also explain a variability of outcomes. Consequently, the study was carried out in several countries. The following countries were selected according to factors related to clirnate, resource use and operational criteria: Belgium (French-speaking population), France (Ilede-France region and Provence-Côte d'Azur-
Alpes Maritimes region). ltaly and Quebec.
.!
Age limih were also included in the definition of the source population. A lower limit of 18 years was decided given the very low prevalence of venous disorders below that age; a higher limit of 75 years was considered as an age beyond which it became more difficult to obtain reliable information by telephone interviews. The inclusion of patients aged more than 75 years with ulcer was considered. since prevalence of ulcer increases with age and ts an important source of costs. This option was not retained, taking into account that. in the field of venous disorders. epidemiological data are less available for milder forms of venous disorders. Since venous disorders are more prevalent in women. a restriction of the source population as to gender was also discussed by the Steering Committee. However. it was considered that. in clinical practice. about 10 to 20% of consultations are in men.
selection criterion was applied according to gender.
Therefore no
In conclusion, the source population was defined as patients aged 18-75 years old
suffering from a chronic venous-related leg disorder, and consulting a general practitioner or a specialist for that disorder in the French-speaking part of Belgium and in some regions of France, ltaly and Quebec, this consultation occurring during the study period. It is noteworthy that variables such as the socioeconomic status or the presence of severe CO-morbidityare related to the number of consultations to the physician and therefore to the probability of being included in the source population. Other factors Vary across physicians and are either unknown or not measured. These variables are implicitly included in the definition of the source population.
4.4.3 Definition o f study population
The study population comprises representatives of the source population for whom data were wtlected. For descriptive analyses (e.g. amount of resource wnsumption. mean quality of life scores....), it implies that the study population represented the source population in ternis of the distribution of categories of variables of interest. that a known sampling scheme allowed a weighting of the results according to the sampling fraction(s) of individuals in each category. or Mat. at the analysis stage. results were stratified according to the categories of these variables. For instance. physicians could have seledively included patients with varicose veins and ulcer. Results would then reflect the situation of
patients with severe presentation of varicose veins and poor quality of life; these patients would be representatives of a more restricted source population than the one defined
above. For analytical purposes. the sampling procedure also had to ensure absence of selection biases. i.e. that the relation between exposure and outcorne (for example. between varicose veins and quality of Me) was not different for rnembers of the source population who participated and for those who would be theoretically eligible but did not participate (Greenland. 1977).
In the VEINES study, several features have been introduced in the protocol in order to prevent as much as possible a distortion in the remitment of patients, or to be able to measure it : the inclusion phase was preceded by a registration of consecutive patients using standard fonns; this procedure aimed to minimise as rnuch as possible the selection of patients by investigators; the inclusion of registered patients into the study was detemined by the study
CO-
ordinators based on explicit and uniforni criteria; a
sampling of the source population was based on the presence of venous disorder rather than by the type of treatment followed; variables suspeded to be related to the number of consultations and to the probability of being inctuded into the study population were measured.
Consequently. al1 of the following items were necessory to define membership of the study population: 1-
to be a member of the source population (to be a patient aged 18-75 years old suffering from a chronic venous-related leg disorder and consulting a general practitioner or a specialist for that disorder in the French-speaking part of Belgium or in some regions of France, ltaly or Quebec, this consultation occurring during the study period )
2
to be registered by one of the investigators on a special form
3
to be sampled by the study CO-ordinators
4
to be eligible. as evaluated by one of the investigators on the basis of the exclusion criteria below
6-
to attend the baseline visit and give informed consent to one of the investigators
7-
to return the first quality-of-life questionnaire.
Entry into the study was effective once these seven criteria were fulfilled.
in the VEINES study, exclusion criteria were kept to a minimum. They are presented in Table 4.1. They induded a rerhicted list of medical causes of revers oedema other than a venous disease (such as right heart failure and nephrotic syndrome), as well as conditions that might lead to loss of follow-up (such as severe physical disease like cancer) or impair the quality of the data, such as mental illness or an insufficient knowledge of the study
ianguage(s) useci in each country (French in Belgium and in France. ltalian in Italy. French or English in Quebec). Patients who could not be contacted by telephone or who did not consent to participate were also excluded. TaMe 4.1 List of exclusion cnteria Age less than eighteen years or more than seventy-five years Right heart failure Nephrotic syndrome Any severe physical disease or life-threatening conditions likely to affect follow-up Any severe psychiatrie disease or depression likely to affect followup Pamlysis of one or both legs lnsufficient knowledge of study language in country Patient unreachable by telephone Patient declineâ to participate
4.4.4
Size of study population
Due to the multiplicity of outcornes. the involvernent of several countries and the recruitment of physicians from different specialities. it was not possible to apply fixed rules of sample size caiculation. Furthermore. issues of sample size determination raised by the assessrnent of economic and quality-of-life outcornes were unresolved when the study was designed (Drummond and O'Brien. 1993). The estimation of the number of patients to be recruited in each country was based on the following assumptions :
-
sample site should have been sufficient to allow an estimation of the proportions of patients using different types of health services in the general practitionersand specialists-based samples separately in each country:
smaller variability in estimates were expected in the specialist-based sample than in the general practitioners' based sample; a minimum number of 100 patients was required in each group to have sufficient precision in estimates; total sample size should have been sufficient to ascertain rare events or test hypotheses, e.g. association of leg ulcer and past thrornbophlebitis; sample size should have allowed performance of more detailed analyses on treatments in pooled data from France and French-speaking Belgiurn. assuming little variation between the h o countries in patients' behaviours. medical practices and prescribing habits. On the basis of these assumptions, it was estimated that 100 patients had to be recruited in each specialist-based sample in Belgium. ltaly and Quebec. 100 in Francelsle de France region and 100 in France-Provence-Côte d'Azur-Alpes Maritimes (PACA). For the general practitioner-based sample. the required sample size was
estimated at 200 in ltaly and in each of the two French regions. and at 300 in Belgium (the number of patients was increased in Belgium in anticipation of the parallel studies). The total required sample size amounted to 900 patients in the general practitionersbased sample and to 500 in the specialists-based sample.
4.5 Sampling procedures
4.5.1 Recruitment o f physicians
In Belgium. France and Italy. the investigators were general praditioners and specialists of venous diseases. including angiologists. phlebologists and vascular surgeons. In Quebec. a pilot phase suggested that the recruitment of patients by general practitioners would
be extremely ditficul! and would require a lot of effort without guarantee of success. Tharefore, only specialists (phiebologists and vascular surgeons) were recniited. Most of the participating physicians were contacted through medical or research associations or contact groups
A possible issue was the non random selection of
physicians. However. since voluntary participation was nevertheless necessary. it was felt that random selection would not significantly improve the representation of the sample. Contacts were facilitated by CO-ordinators of scientific associations of general practitioners. The number of physicians recniited in each country was initially detenined by having no more than 10 patients followed by each physician. This critefion was set to increase the feasibility of the follow-up, to minimise the time of recniitment and to account for the interphysician variability in the management of dironic venous disorders. The majority of physicians attended one of the training sessions organised in the participating countries. In only exceptional cases, information was provided by telephone. In France and in Belgium. al1 training sessions were given by the coordinator of the Research Team. The method of recniitment varied from country to country as follows.
Belaium The recniitment of general practitioners was co-ordinated by the 'Centre de Recherche et d'Information en Soins de Santé Primaires" (CERISSP). A total of 48 French-speaking general practitioners were reauited from members of the research section of the Departments of General Practice of the Free University of Brussels. the University of Liege and the Catholic University of Louvain. from members of the Fédération des Maisons
Médicales, and from mernbers of the Scientific Society of General Practice (SSMG). All Belgian members of the AssoUation Bénéluxienne de Phlébologie (Benelux Phlebology Association) were also contacted. Of the 15 specialists who accepted to undertake the study. 11 ended the registration phase and included patients. Of these. 6 were vascular or generai surgeons. 4 angiologists or phlebologists and 1 a dermatologist.
France The recniitment of general practitioners was CO-ordinatedby the "Société de Formation a la
Thérapeutique du Généralisten(S.F.T G.) and the ESCOLA association. It was carried-out in two administrative zones: 28 general practitioners were recruited in Paris ( lle-deFrance region) and 31 general practitioners were recruited in the southeast of France (PACA
@
region). Four gynaecologists who participated in the registration phase in the Isleof-France zone were included in the sample of general praditioners. Specialists were contacted through contact groups in Paris, Lyon and Toulouse. Fifty-nine of them agreed to participate in the study. They were all angiologists or phlebologists. except for one surgeon in Lyon. The geographical distribution of specialists who accepted was more spreadout than that of general practitioners. with locations in Paris, the south-
east and the sauth-west of France. ltaly The remitment process was w-ordinated from the Department of Surgery at the University of Padova with the collaboration of the Ambulatono Flebologico association. Thirty-three
general praditioners and 17 specialists (al1 of them general or vascular surgeons) participated in the study. General praditioners were from the Veneto region. whereas specialists were from the Veneto region and from Milan. Quebec The recruitment process was wordinated by the Centre for Clinical Epidemiology and Community Studies at the Jewish General Hospital.
Twenty-seven specialists (15
phlebologists and 12 surgeons) with a practice located in Montreal or Îts surroundings accepted to participate in the registration phase among those contacted either directly or through the Canadian Association of Phlebology. Many then invoked a lack of time and dropped-out of the study at an early stage: 14 specialists (10 phlebologists and 4 surgeons) participated in the inclusion phase.
4.5.2 Registration phase
The objective of the registration phase was two-fold: a
to establish a sampling frame of the source population:
to record the minimum information necessary to relate results obtained in the study population to the source population.
Each participating physician was asked to register 25 consecutive patients of any age wnsulting spontaneously for a cornplaint related to a chronic venous disorder of the leg, of any severity. This criterion allowed for the inclusion of incident as well as of prevalent cases, the only inclusion aiteria being a spontaneous consultation and the presence of a venous-related disorder in the lower limb. However. there was no requirement that the venous-related leg disorder was the prirnary cause prompting the patient to consult. provided that the venous probiem was discussed or mentioned by the patient during the consultation. Standard registration foms were provided to physicians. There were small differences between those used by general praditioners and those used by specialists. The cornmon information was the date of consultation. the patient's initiais. age, sex and occupation. the presence of venous symptoms, and the presence of each of the following signs : varicose veins, skin changes and ulcer. The investigator was also requested to provide an opinion regarding the levei of certainty of the venous origin of the syrnptoms and signs presented by the patient. with three possibilities of answer : certain. probable or possible. Physicians were also asked to state the treatment prescribed or already taken by the patient on the day of consultation. The foms used by specialists included two additional fields : the presence of small varicosities (e.g. telangiectasia) and the presence of oedema. The reasons for this difference are as follows. Small varicosities are highly prevalent in the general population and are not a sign of venous disorder in al1 cases. Since only 25 patients had to be registered by each general practitioner. selecting small variwsities as an independent criterion for eligibility could have led to their over-representation in the sample of registered patients, with a concomitant reduction in the number of patients with more severe foms of venous disease like ulcer or skin changes.
On the other hand. the
presence of vaficosities represent an important reason for consulting specialists. and not including such patients would have ignored a large proportion of patients consulting specialists. Regarding oedema. it was considered that there are various causes of leg oedema. including cardiac, hepatic and renal causes. If venous oedema was included in the list of venous disorders to be registered by general practitioners. many patients with
oedema of non-venous origin would probably have been registered and seleded. The situation was different for specialists. Patients are uswlly referred to them once other causes have been excluded and a venous otigin for the oedema suspected. Telangiectasia and oedema were therefore not included in the registration fom used by general praditioners. For general practitioners, the registration period (defined by the date of registration of the f i ~ and t the last patient) took place from April 1994 to Odober 1994.For specialists. the registration period took place from May 1994 to June 1995. The long duration of the registration phase in the group of specialists is explained by the
fact that sorne of them were recruited as general practitioners in the first stage, but included later on in the group of specialists when the specialists' study started in 1995. In Quebec. two specialists who were contacted later were asked to include consecutive patients without a registration phase.
Registration statistics by country and type of physicians are presented in Tables 4.2 and 4.3. A total of 5688 patients were registered: 3128 patients by 166 general practitioners
(mean : 18.7 per physician) and 2560 by 116 specialists (mean : 22.1 per physician). The mean and median duration of registration were respectively 34 and 29 days for general practitioners and 23 and 15 days for specialists.
-
Table 4.2 Registration statistics general practitioners
Number of physicians al1 physicians with 25 patients registered
Belgium
France
ltaly
58 20
75 31
33
17
Al l countries
166 68
Number of patients al1 patients mean per physician range per physician
Duration of registration (days) ail physicians rnean per physician median range
191 33.6 28.5
2-1 73
167 27.7 25.0 5-1 58
119 47.3 43.0 10-1 10
191 33.8 29.0 2-173
-
Table 4.3 Registration statistics specialists
1
Belgium
France
ltaly
Quebec
All
Number of physicians al1 physicians 25 patients registered Number of patients al1 patients mean per physician range per physician
Durafion of registration (days) al1 physicians mean per physician median range
1.5.3 Sampling procedure
Fifteen patients per physician were sampled by the research team. The only exclusion criterion applied at this stage was an age lower than 18 years or higher than 75 years. In the first round of seledion. lists of up to 15 selected patients were made and sent to the investigators on a Selection F o n . In order to obtain a sufficient number of patients in each clinical category, al1 male patients and all patients with venous ulcer were placed at the top of the list of patients to be included. so that they had to be the first ones to be contacted and included, if they agreed. The other subjects were randomly selected and placed on the list after males and ulcer patients. After evaluation against the exclusion criteria. investigators were asked to contact the first eligible patients on the list until a specified nurnber of patients could be enrolled. This number of patients was determined from the planned sample size by country and type of physicians. and from the number of physicians who accepted to participate. Initially. the numbers of patients to be followed by each general praditioner (GP) were 7 in Belgiurn. 7 in France and 6 in Italy. In the second stage. some Belgian and French GPs were asked to recruit additional patients to compensate for the drop out of other physicians. All specialists were asked to recniit 10 patients.
After informing eligible patients of the study, the investigaton asked them to participate. and, if they agreed, gave an appointment for them to attend an inclusion visit.
The
Selection F o n asked for details of the exclusion criteria, if any, and whether or not the patient declined to participate in the study. For many physicians, the list of eligible patients was not large enough to allow inclusion of the specified number of patients, due to the number of patients meeting at least one of the exclusion criteria or declining to participate. Other physicians were then requested to include up to 12 patients in order to compensate for this. In the s e m d selection round. al1 the remaining patients were listed in randorn order. The majority of general practitioners
then received a second list of eligible patients to contact.
4.5.4 Assessrnent of exclusion criteria
Exclusion criteria invoked by general practitioners and specialists in each country are presented in Tables 4.4 and 4.5. "Patient declines to participate" and "Patient unreachable by telephone" were the most frequently invoked reasons for exclusion. It ts possible that these criteria were used as pretexts for not including patients or that only one phone cal1 was given to some patients with no further attempts to contact them. Other eligible patients were not enrolled by physicians although no exclusion criteria were mentioned. The large nurnber of patients with exclusion criteria rnay be related to the fad that some physicians did not contact their patients according to the order specified on the Seledion Fom. but solicited them as they
came back to their consultation office. The la& of compliance with the inclusion procedures may have two possible consequences : 1- some patients who should have been included in the first order of priority (e-g. male or ulcer patients) were not solicited at all: 2- the randorn order of the other patients was not respeded. resulting in the inclusion of patients who consult more frequently. This phenomenon was found more frequently with specialists than with general practitioners. It could be explained by practical reasons. The catchment area of specialists is generally rnuch larger than that of general practitioners. with many patients having to cover long distances to consult. Some specialists were therefore reluctant to ask patients to
travel for their inclusion visit only. In addition. when a patient was referred to them by a general praditioner, some specialists judged it to be necesrary to infom the general practitioner of their intent to enrol the patient into the study, or to ask himlher the authorisation to do so, a task considered too time-consuming. Table 4.4 Nurnbers of patients in each category of exdusion criteria invoked by general
r--
Exdusion criterion
Belgiurn
France
ltaly
Age outside 18-75 years range Right heart failure Nephrotic syndrome ~ e v e r ephysical disease or life-threatening condition likely to affed follow-up Severe psychiatricdisease or depression Iikely to affed followup Paralysis of one or both legs Insuftlcient knowiedge of study language Patient unreachable by telephone Patient deciined to participate Urispecified
Table 4.5 Numôers of patients in each categoiy of exclusion mena invoked by specialists
1
Exclusion criterion -Age outside 1û-75 years range Right heart failure Nephrotic syndrome Severe physical disease or life-threatening condition likety to affect follow-up Severe psychiatric disease or depression likely to affect follow-up Paralysis of one or both legs lnsufficient knowiedge of study language Patient unreachable by telephone Patient declined to participate
Belgium
France
ltaly
Quebec
4.5.5 Inclusion
Once the patient agreed to participate. an inclusion visit was scheduled. The physician sent the patient the first quality of life questionnaire. to be returned at the inclusion visit.
Data collected during this visit included a fomal assessment of exclusion criteria and informed consent, demographic variables and clinical and economic data. lnvestigators infoned the study cosrdinators that an inclusion visit took place by filling and returning
an inclusion card. An eligible patient was recruited when three conditions were fulfilled : 1) heishe gave hislher informed consent, 2) the baseline visit was performed, 3) the first quality of life questionnaire was returned. For general practitioners. the inclusion phase took place from July 1994 to May 1995. For specialists, the inclusion phase took place from February 1995 to October 1995. The long duration of the inclusion period for general practitioners is explained by a second
round of selection of patients included by some physicians. The selection procedure explained above led to the inclusion of 1.475 patients. In addition. 13 patients were recniited by three specialists who did not participate in the registration
phase. and a total of 51 patients who had not been registered were nevertheless recruited by investigators.
The total number of patients initially recruited in the VEINES study is thus 1.539. Eight of these were excluded at the data analysis phase : 6 were aged more than 75 years. and two lacked either a baseline quality of life questionnaire or a clinical examination f o n .
Table 4.6 below shows the minimum number of patients to be recruited in the general practitioner- and specialist-based samples in each country (as presented in section 4.4.4) and the actual sample size in each patient group. The pre-specified number of
subjects to be recruited was achieved. Table 4.6 Required and actual samgle sites per country and type of physician
General practitioners required Belgium France ltaly Quebec Total
actual
Specialists required
actual
4.6 Data collection 4.6.1 Schedule of data collection
Prospective versus retrospective data collection The course of a population is inherently prospective. Events occurring during this course rnay be collected retrospectively (only the events that occurred before the patients' inclusion are ascertained), prospectively (only the events that occurred after the patients' inclusion are ascertained) or both. A retrospective data collection (e.g. by inspection of medical charts) has several
practical advantages: it uses already existing data that can be collected over a short period of time. it allows the coverage of a long period of time and the ascertainment of rare events, such as venous ulcer, and it avoids the effect of observation biases that
can occur in prospective surveys when participation in a study affects patients' and investigatorsJbehaviours. A retrospective data collection had disadvantages in the context of the VEINES study:
most of the required information was not available (for example : quality of life and economic outcomes). retrospective interviews covering a long period of time were not valid enough regarding a number of outcomes. such as symptoms. drug utilisation or number of consultations, and a retrospective design was obviously inconsistent with the testing of the reliability of the quality-of-life questionnaire or with the validation of clinical signs of venous disorders. The VEINES study design used both retrospective and prospective methods of data collection.
The retrospective survey included clinical and economically-related
information on disease history. treatment history and health services utilisation. The period covered by the retrospective evaluation was lifetime when disease history and surgical procedures were concerned. and 12 months for the assessrnent of treatments prescribed and other consurnption of health resources such as number of consultations to specialists. The prospective survey allowed repeated rneasures of clinical. quality-
0
of-life and economic outcomes over time.
Ouration of observation time For operational rearons, the duration of the observation of patients enrolled in the mhort was restfided to one year after admission into the study population. In the course of the study, some investigators, mostly general practitioners, were asked to compensate for the low recruitment of patients by other physicians. Since this situation extended the time span of the inclusion period for these investigators, calendar dates were therefore chosen to define the end of the follow-up period : 30 November 1995 for general practitioners and 31 March 1996 for specialists. Therefore. some patients did not have a follow-up of 12 months. Timina of data collection Historical. baseline and follow-up information were collected during the one-year period. Three main methods of data collection were used : assessment of past and present medical history by the investigators, self-administered questionnaires and phone interviews. Retrospective medical and ewnomic information covered the 12 rnonths before the recruitment visit. Two follow-up clinical examinations visits were planned. The first follow-up visit was allowed to take place from 2 to 6 months after the recniitment visit. As instructed. this flexible schedule allowed the investigators to perfonn at least one clinical examination during the warmest season of the year, or when symptorns were most
severe. The second follow-up visit was scheduled to take place 12 months after the recruitment visit. Two quality-of-life self-administered questionnaires were included during the observation period. They were provided to the patient by the investigator at the time of each follow-up examination and were mailed back to the study madinators in a pre-stamped envelope. Three phone calls were planned. each phone cal1 covering a three months period. The first phone cal1 was planned for three months after the baseline visit, and the subsequent phone calls were planned for every three months thereafter.
4.6.2 Algorithm For data collerfion
The baseline visit corresponded to the inclusion visit. Data collection was made in 8 steps: 1. The list of patients sampled by the research team from the list of al1 registered
patients was sent to the investigator. 2. The investigator assessed the exclusion ctiteria for al1 patients sampled. 3. The investigator called the eligible patients to explain the study and ask for a
preliminary consent to participate in the study. 4. If the patient accepted, the investigator gave him an appointment for the baseline
visit.
'
5. Before the baseline visit, the investigator sent the first quality-of-life questionnaire ("Pl" questionnaire) to the patient; this questionnaire was to be returned at the baseline visit. 6. The investigator performed the baseline visit (or inclusion visit) ("Ml" questionnaire); this visit also included provision of informed consent. administrative work. and
evaluation of exclusion criteria. After completion of the visit. the investigator also had to send an inclusion card to the local CO-ordinatingcentre. 7. The investigator was visited by the clinical research assistant who reviewed and collected the "Ml" questionnaire. 8. Starting three months after the inclusion visit. the clinical research assistant began
three-monthly telephone interviews to the patient. A sirnilar procedure was followed for each of the two follow-up visits: 1. For the 1" follow-up visit ("M2").the physician was allowed to perform the visit
anytirne between 2 and 6 months after the baseline visit: the 2ndfoltow-up visit ("M3") had to be planned for about 12 months after the inclusion visit.
2. After each of the follow-up visits. the investigator had to send a card to the local coordinating centre to inform it that the visit had taken place: at the tirne of the visit. the investigator provided the patient with the second ("P2")or third ("P3")quality-of-life questionnaire to be filled in at home.
3. The patient was asked to send the questionnaire back to the co-ordinating centre in a
pre-stamped envelope. 4. The Clinical Research Assistant visited the investigator.
5. Telephone interviews were performed every three months independently from the visit to the patient.
4.6.3 Socio-demographic information
Socio-âemographic data were collected from the first phone cal1 to patients, or from the investigator's questionnaire. They fell into one of the following categories : 1) descriptive information useful to characterise the study population : age. sex. origin
(European, African, American, other), educational level, current and past occupation, number of children and age of youngest child. family size:
2) data useful for the interpretation of econornic data, such as family status, educational level, employment status, number of persons living at home. family net monthly incorne, health insurance status;
3) possible consequences of venous-related leg problems on work (work stoppage. change of work. number of workdays lost during the previous 3 months) and physical activities.
4.6.1 History o f venous disease and associated factors
Information on disease history and associated factors was collected from the investigator's questionnaire, the patient's questionnaire and the first phone d l .
- history of venous disease :
length of presentation of venous disease as stated by the
investigator and the patient. time since first consultation and reason for first consultation. number of consultations to general practitioners and specialists (angiologist/phlebologist, surgeon. dermatologist. gynaecologist. other) in the 12
months preceding the first medical visit and since the last VEINES visit in each of the follow-up visits. history of superficial thrombophlebitis. deep thrombophlebitis. pulmonary embolism. bleeding of varicose veins and ulcer. laboratory tests undergone
in the past and during follow-up (e.g. U.S. Doppler, echo-Doppler, phlebography. plethysmography, etc...), hospitalisation for venous problems during the past 12 months and during follow-up;
- heredity :presence of varicose veins or ulcer in father, rnother, brother. sister or other farnily member;
- history of trauma or surgical intementions on the lower legs : this information covered past history and events occurring during follow-up;
- concomitant medical conditions : weight and height. blood type.
smoking status.
CO-
morbidity (diabetes, high blood pressure, chronic pulmonary condition. cardiac disease. chronic hepatitis or cirrhosis. hypothyroidism. rheumatoid arthritis. arterial disease of lower limb. lymphoederna, joint arthritis, other disease to be specified). delivery in previous year and during follow-up, pregnancy in previous year and during follow-up. menopause. hormonal therapy:
- history of treatments used in the past and during follow-up : sclerotherapy, stripping surgery or other surgery. drugs prescribed for venous disorders during the last 12 months and during follow-up (NSAIDS. analgesics. diuretics. venoactive drugs. other drugs). history of undesirable effects of treatments in the past 12 months and during fol low-up.
4.6.5 Symptoms
Symptorns of venous disorders were assessed using two methods : by investigators during each of the three medical consultations. and by the patient in each of the three selfadministered quality of life questionnaires. The rneasurement of symptoms is detailed in Chapter 11 (Symptoms and quality of life). section 11.3.4. 1.6.6 Signs
The clinical examination made by generai practitonen included investigation for the following signs on each leg:
-
telangiectasia or reticular veins varicose veins (if present. also for presence of pain)
:
oedema hyperpigmentation
-
eczema or dennatitis along the venous pathway
-
atrophie blanche
-
ulcer (if present, number of ulcers and inflammatory aspect)
lipodematosclerosis
The presenœ of each sign was assessed as a yes or no response. A patient was considered to present a specific sign if it was recorded for either the right or the left leg. The following objective measures were also assessed on each leg : measure of the leg circumference. measure of the maximum diameter of the largest varicose vein (if present) using a calliper, and measure of the size of the largest ulcer (if present). Additional information was requested from specialists. including :
-
presence of ankle flares and additional information on ulcers and oedema
-
additional information on varicose veins (W): territory : short saphenous. long saphenous. non-saphenous pain on palpation of at least one W on each leg: present or absent location: abdomen. thigh. lower leg, foot severity rated on a 4-point scale : mild. moderate. severe. very severe.
1.6.7 Laboratory measures
Laboratory measures were not requested in the study. lnvestigators were nevertheless requested to state any laboratory examination undergone in the past by their patient. The following list of investigations was plethysmography.
rheography.
elicited : Doppler. echo-Doppler (duplex).
phlebography,
thermography.
venous
pressure
measurement. scanner and magnetic resonance imaging. Any. other test performed could also be cited. lnvestigators were also asked to state whether any of superficial or deep venous incompetence were demonstrated any time previously.
The location of the
incompetence was also requested : long saphenous vein reflux. short saphenous vein reflux. deep vein reflux reflux in perforators
Specialists were also requested to report results of any Doppler. echo-Doppler or RimaTrendelenbourg tests if performed during the course of the medical visit. If mntinuous-wave Doppler was performed, results had ta be reported for each leg as presence or absence of reflux in the superficial femoral vein. the long saphenous vein Doppler or the poplitea fossa. If performed, duplex had to be reported as presence or absence of reflux in the same locations as Doppler. plus the short saphenous vein in the pcplitea fossa. gernellar veins. the popliteal vein and perforators at the upper or lower half of the leg.
5.6.8 Economic evaluation
Two sources of information were used to collect economic data : the investigator (on each of the patient's visits to the investigators) and the patient hidherself (by phone interviews administered three times at three months intervals). The questionnaires were designed conservatively in such a way that only use of health services related to the venous disorder were collected.
The schedule of data collection and the items covered by the
questionnaires are presented in Chapter 12 (Health service utilisation in Belgium) section
12.3.3. (Methods of data collection).
4.7 The Duplex study
4.7.1 Objectives
The main objective of the Duplex study was to examine the association between various clinical presentations of chronic venous disorden of the leg and the presence of superficial and deep venous incornpetence as ascertained by specialists of venous diseases. using standard measurement criteria. A secondary objective was to measure the validity of the clinical dîagnosis of venous
disorder by general practitioners in cornparison to the diagnosis made by specialists.
1.7.2 Rationale
The investigation of the association between venous disorders (including varicose veins) and superficial andlor deep venous inwmpetence aimed to assess the predictive values of signs of venous disorders for evaluating the presence of venous reflux. This information
was considered important to general practitioners in the management (diagnosis and treatment) of patients with venous-related disorders.
It also intended to examine
hypotheses on the etiology of varicose veins. Since the inclusion criteria for patients did not include objective information. it wuld not be guaranteed that only patients with venous disorders were recniited into the VEINES study.
The Duplex study gave also the opportunity to assess the validity of the diagnosis of venous disorders by general practitionen as well as their correct interpretation of signs presented by the patients.
The rationale for undertaking a parallel study in a sub-sample of the VEINES cohort is further explained in introductions to chapters 7 (Validity of the diagnosis of variwse veins by general practitioners) and 9 (Varicose veins and reflux).
1.7.3 Patients and methods
Desian Infomed consent was cbtained from 149 patients followed in the general practitioner-based
VEINES study to undergo a clinical examination and a test of venous fundion by a specialist of venous disorden. Patients who already had a duplex examination perfomed in the last 12 months were excluded.
In order to ensure unifomity of the diagnosis,
investigators used the same diagnostic aiteria. The test of venous function included an echo-Doppler (duplex) and a short assessrnent of the clinical signs.
lnvestigating centres Seven investigating centres participated in the study. This number was kept as small as possible in order to reduce the likelihood of variability in echo-Doppler testing and clinical examination. Due to the geographical spread of patients in Belgium. four centres were
seleded in that country in order to ensure adequate reauitment. The selection of investigating centres was guided by their recognised expertise in duplex testing.
lnvestiaations
Laboratory test. The protocol of the duplex testing was approved by al1 investigators. These were asked to state whether reflux was present or not in the veins investigated for each leg. Laboratory tests are detailed in Chapter 9 (Varicose veins and venous reflux).
Section 9.3.3.
Clinical assessment. The clinical assessment consisted of an abbreviated version of the clinical examination perfomed by specialists in the VEINES study. It is detailed in Chapter 9, Section 9.3.4.
Sam~lingprocedure a. Eligbility A subject was eligible if he/she fulfilled the following criteria :
1)
helshe was included in the VEINES study;
ii)
he/she was not investigated by echo-Doppler in the 12 months before the inclusion visit or during the follow-up period; this criterion was included to avoid. if possible. patients included in the Duplex study being those with the most severe disorders:
iii)
hisher general practitioner agreed to participate in the Duplex study:
iv)
heishe gave an inforrned consent.
6.Recruitment The desirable sample size was estimated at 150 subjects (50 in each country). Their recruitment was perfomed in five steps : 1)
General practitioners participating in the VEINES study. with a pradice located within a reasonable distance from one of the specialised centres. were asked to
collaborate.
ii)
Subjects enrolled by the general praditioners recruited in i) were identified by the
VEINES study team in each country; a list of subjects to be wntacted by was sent to the general practitioner, as well as the desiraMe number of subjects to be included. Each study team was asked to remit 50 patients and to try to r e m i t an equal number of patients in each CEAP category. iii)
The general practitioner proposed to the selected patients their participation in the
Duplex study; when a patient refused, the physician then selected the following patient on the list. v)
Clinical research assistants were informed of the names and phone numbers of patients who agreed to participate. as given on their infomed consent foms.
vi)
Clinical research assistants organised visit schedules to the specialists.
4.7.1 Study population
Results are available for 147 of the 149 patients who agreed to participate in the study in Belgium (67) France (30) and ltaly (50). One French subject enrolled in the duplex study was 77 years old and excluded at a later stage from the VEINES study population.
This patient was also excluded from the Duplex study. Two other subjects who did not attend any of the follow-up visits of the VEINES study were nevertheless recruited into the Duplex study. These two subjects were excluded from some analyses.
4.8 Quality control
Quality control was carried out at different stages of the VEINES study : 4.8.1 Single dossier per patient
All questionnaires (except those used in the telephone interviews) and study documents were attached to a unique study document for each patient. Each questionnaire had two carbon copies. Both copies could be tom out and taken by the clinical research
assistant (CRA) at the time of the visit. The patient's dossier also contained a summary of the procedures. the pre-stamped and pre-addressed cards to be sent to the local CO-
ordinating centre after each visit, the consent form, the form for ascertaining exclusion criteria and the form for administrative information. Moreover. a unique sequential patient identification number (four digits) was pre-printed in rad ink on each page. This single dossier for each patient has several advantages : O
given the large number of questionnaires used in the study, it limited the loss and mixing of documents; it gave the investigator al1 the information needed to include and follow a patient;
a
it allowed the
C M to have a clear and rapid view of the study progress for each
patient.
4.8.2 Patient identification
Patients were identified by two methods : a four-digit sequential number corresponding to the number pre-printed on al1 pages
of the patient's dossier used by the investigator at the time of the inclusion; an identification consisting of the physician's number (3 digits), a patient number allocated at the time of the registration (from 1 to 25) and two initiais. These two identification numbers were used simultaneously in the verification of questionnaires and in the data validation process.
1.8.3 Check visits and data control
Clinical research assistants were in charge of ensuring data quality.
They work
included the following activities : communicating with investigators : transmission of study documents. recall of visits to be done. check visits. requests for additional information keeping statistics regarding recmitment and study progress checking questionnaires for inconsistencies and obviouç errors follow-up of quality-of-life questionnaires returned by physicians ( P l ) and patients
(P2 and P3) coding of drugs and pathologies
a
transmission of documents for data entry. All investigators were visited at least once by a clintcal research assistant. and al1 general practitioners were visited at least twice. The first visit was organised as soon as possible after the first clinical examination (inclusion visit) in order to introduce corrections in the physician's working procedure as needed. This visit consisted in a review of the questionnaires, followed by a request for additional information in case of missing or inconsistent data, a control check of the patient's dossier, and a reminder about the procedures to be followed in subsequent visits. Possible difficulties with the patient encountered in the phone interviews or in the self-administered questionnaires were also discussed.
1.8.1 Newsletters
A VEINES Newsletter was created at the international CO-ordinatingcentre (Liege) in order to keep physicians aware of the overall progress of the study. The Newsletter also gave important reminders about study procedures. information regarding the
study
progress.
opinions
It also provided other of
ethical
committees.
implementation of parallel studies. etc... Except for the first Newsletter. different Newsletters were written for specialists and general practitioners.
4.8.5 Data audit
After data entry, al1 data files and questionnaires were collected centrally at the McGill Centre for Clinical epidemiology and Community Studies at the Jewish General Hospital.
As there were up to 12 questionnaires per patient (about 15.000
questionnaires in total). an audit of data was performed in order to get confidence in the database and start the analyses.
In the first step, the following procedures were followed : to file al1 paper questionnaires in one "safe" place to verify if ail patients in database did exist on paper
to veriw if al1 paper questionnaires were entered in the database to do a random check comparing database and paper data in 10 % of the data to verify if the newly created variables such as disease severity scales, symptom scores, etc. .were good and valid (Le. well programmed in SAS). The random check comparing database and paper data did not identify any problems worthwhile of further investigations. After the first check of available questionnaires, a certain nurnber of paper documents were identified as missing. Most of them were retrieved in various locations. For some patients. a copy of some of the questionnaires was requested from the physician. In the second step, discrepancies between patients and physicians were analysed. In the VEINES study. a set of questions wncerning signs and symptoms of venous disorder was asked to both physicians and patients. The consistency of physician-patient responses to essentially the same questions was investigated. There was no differenœ between the mean discrepancy scores for patients seen by specialists and those seen by general praditioners. An analysis of variance by physician did not show a statistically difference across physicians.
4.9 Ethical considerations
The study protocol was submitted to relevant ethical cornmittees in Belgium. France and Quebec. This step was not necessary in ltaly because VEINES is an observational study. Favourable opinions were received the Ethics Committee of the Medical Faculty of the University of Liege in Belgium. the "Comité National Informatique et Liberté" (C.N.I.L.) in France and the Research and Ethics Committee of the Jewish Generai Hospital in Quebec. As an observational study, the VEINES study was not covered by the Hurié law on the protection of private life in France. and was therefore not fomally submitted to a 'CCPRB' cornmittee.
5. Source and study populations 5.1 Source population
5.1.1 Introduction The registration of consecutive patients consulting physicians in five countries for a venous-related leg disorder was an unique feature of this study. The objective of this registration phase was two-fold : to establish a sampling frame for the source population, and to record the minimum information necessary to relate results obtained in the study population to the source population. The number of items to be recorded was purposely limited in order to prevent the investigators from not registering consecutive patients due to workload. The characteristics of the source population are presented in this chapter.
5.1.2 Results The characteristics of the 5688 patients registered (3128 by general practitioners and 2560 by specialists) are s h o w in Tables 5.1 and 5.2. When interpreting specialist-based data (Table 5.2).it should be kept in mind that different types of specialists participated in each country : almost exclusively angiologists in France and surgeons in Italy. and a mixed group
of phlebologists and surgeons in Belgium and Quebec. In both groups. about 84% of registered patients were women. Little variation between countries was obsewed. with proportions of women ranging from 80.6% (Quebec) to 87.4% (France) in the specialist-based sample.
The mean (s.e.) ages in the specialist- based
sample and the generalist-based sample were respectively 58.9 (0.3) yean and 50.1 (0.4) years. In the generalist-based sample. the age distribution was comparable in the three countries. with the mean age being slightly higher in Belgium (61.4 years) and lower in ltaly (55.9 years). In the specialist-based sample. registered patients were on average younger in Belgium (mean age: 48.9 years) and in Quebec (mean age : 48.3) than in France (mean
a
age: 50.4) and ltaly (52.3).
Symptoms were reported by 70% of patients seen by general praditioners and 66Oh of patients seen by specialists.
Ulcer was rewrded for respectively 5.3% and 4.1% of
patients. In both samples, patients seemed at a more severe stage of disease in Belgium than in the other countries, with proportions of individuals with ulcer and skin changes being 10.7% and 17.2% respectively in the specialist-based sample and 6.6 % and 36.9% respectively the GP-based sample.
Investigators were also asked to state the level of certainty of the venous origin of the clinical pidure patients presented with. The objective was to assess whethec such information was related to outcomes of venow disorders. such as quality of life or health service utilisation. In both samples. more than 70% of physicians declared certainty that the signs and the symptoms were of venous origin. In analyses presented in this thesis. there
was no effed whatsoever of the level of certainty on the dependant variables. and it will not be wnsidered again.
Another question investigated the type of treatment followed by the patient. The question was aimed at assessing the relative distribution of venous-related treatments in real clinical pradice. This information on treatments was however asked as an open question and can only give a global picture of the relative importance of various treatrnents. There was a
large number of missing values : 24.5% in the GP-based sample and 11.3% in the specialist-based sample. Moreover. no distinction was made between current and past treatments. Venoadive medication (53% of patients) and compression (14%) were the types of treatment most frequently mentioned by general practitioners. In the specialistbased sample. venoactive medication was rnentioned for 27.0% of patients, but this proportion was mainly accounted for by French patients (46.1%). whereas lower percentages were mentioned in Belgium (15.6%). ltaly (10.4%) and Quebec (1 patient). Sclerotherapy was mentioned for 40.5% of patients in the specialist-based sample. In Belgiurn. ltaly and Quebec, surgery had been perfomed or was planned in equal proportions (2042%).
5.1.3 Discussion The data from the registration phase show similar characteristics of patients consulting general practitioners on one side and specialists on the other side in Belgium. France. ltaly and Quebec. The differences in the types of treatment between countries suggest however that the detailed analysis of health service utilisation in Belgium presented in this thesis may not reflect the situation in other countries. Frorn the differences in clinical presentations observed between countries and types of physicians. it can be concluded that the recruitment of patients from several countries was fully justified. The data' from the registration phase in the general practitioner-based sarnple in France can be compared to those reported by Boccalon et al. (1997) who examined the characteristics of 895 outpatients presenting with chronic venous insufficiency (CVI) to 142 general practitioners and 3 angiologists. Women represented 91% of subjects with a mean age (standard deviation) of 47 (16) years. Men had a mean age of 53 (13
years). Varicose veins or history of varicose veins were present in 67% of the sample. which is higher than the 53.1% reported in the registration phase of the French VEINES study. Oedema. skin pigmentation. atrophy and venous ulcer were found in 68W.23%.
5% and 1% respectively of patients. Moderate or severe heaviness was reported by 92% of patients. to be compared to a prevalence of about 75% in the VEINES study.
On the basis of the French data. patients included in the registration phase by general practitioners seem therefore to present with a less severe venous disease than in the ,
results reported by Boccalon et al. The data from the specialist-based registration data can be compared to those reported by Zuccarelli and Vincent (1990) who surveyed 2.737 women attending 193 consultants for angiology or phlebology in France. They reported a mean age of 42 years. which is lower than the mean of 49 years found in this study in France. Medical treatment. compression. sclerotherapy. stripping and other treatment were used by respectively 51%. 14%. 26%. 13% and 4 %of subjects. The main difference with this study is the
lower proportion of subjects having had sclerotherapy. About 88% of women presented
with 'varices'. Since it is not known what this term exactly meant, this figure cannot be compared with the proportions of wornen with varicosities or varicose veins reported by specialists participating in the VEINES registration phase.
5.2 Study population
5.2.1 Introduction The procedure used for sampling the study population from the source population has
been explained in Chapter 4, Sections 4.5.3. 4.5.4 and 4.5.5. One of its objectives was to increase the sampling fraction for men and for ulcer patients. This Chapter presents clinical and socio-demographic characteristics of the study population and the follow-up rate during the observation period. Patients are stratified using the CEAP classification (Porter et al.. 1995). which divides a study population into classes of presumably increasing severity of disease. The
CEAP classification has been presented in Chapter 3. Section 3.1.5. In brief. it includes 7 classes :
r,
class O :
no visible or palpable signs of venous disease
0
class 1 :
telangiectasia or reticular veins
class 2 :
varicose veins with or without telangiectasia
+ class 3 : oedema with or without telangiectasia or varicose veins class 4 :
skin changes ascribed to venous disease - e.g. pigmentation. venous eczema. lipoderrnato-sclerosis. with or without any of the above
class 5 :
healed ulceration with or without any of the above
class 6 :
active ulceration with or without any of the above
Socio-dernographic data include distributions of patients by type of investigator and country. age. gender. educational level and net household income.
5.2.2 Clinical and socio-demographic characteristics
The numberr (and proportions) of subjects in each CEAP class were the following : class 0: 58 (3.8%),class 1: 201 (13.1%), class 2: 367 (24.0%), class 3: 197 (12.9%), class 4: 556 (36.3%), class 5: 117 (7.6%). and class 6: 35 (2.3%). Distribution bv CEAP class, countw and investiaator The distributions of patients by CEAP class showed differences behrveen the GP based- and speciaîist-based samples (Tables 5.3 and 5.4). Presence of skin changes with or without oederna. varicose veins or telangiectasia (CEAP class 4) was the most frequent clinical presentation in bath samples [>30°h).
About 20-25% of patients
presented with varicose veins alone (CEAP class 2). An important difference between the two types of physicians was the much larger proportion of patients with telangiectasia alone in the specialist-based sample (21.6Oh) than in the GP-based sample (7.9%). This difference was accompanied by lower percentages in almost al1
other clinical classes. except class G.
In the GP-based sample, the proportions of patients with healed or active ulcer in Belgium, France and ltaly were about 15%. 10% and 5% respectively.
These
percentages are approximately twice as much as in the registration phase (6.6%. 2.8% and 3% respectively), which could indicate that the sampling procedure was successful in increasing the sampling fraction for patients with ulcer. This was also true in the
specialist-based sample when al1 patients are considered. The main difference between countries in the GP-based sample was the prevalence of 40% of skin changes in patients recruited in ltaly and in Belgiurn. as compared to
28.8% in France. This difference was compensated by a higher proportion of patients
in classes O (4.3%) and 1 (10.9%)in France. Distribution bv CEAP classLaae and clender About 23% (2201947) of patients in the GP-based sample were men. as compared to
a
14% at the registration phase (Table 5.5). Corresponding figures for the specialist-
based samples are 20.5% (1201584) and 13.9% (Table 5.5). Thus, the sampling procedure was also successful in increasing the sampling fraction for men. although the actual number of eligible men who were not recruited is not known.
The
proportions of patients aged 18-39, 40-59 and 60-75 in the total sample were 17.4%. 40.9% and 41.7% respectively. The distribution of CEAP classes varied by age and gender with more severe presentations of venous disorders being associated with higher age and male gender, particularly in the specialist-based sample. Distribution by educational status and revenues
On average, patients consulting specialists have completed a higher level of education than patients consulting general practitioners, with 34.6W and 6.6% of them having completed the higher secondary school level, the university level or the equivalent (Table 5.7). This difference was reflected in the difference of estimated net household incomes in patients who provided or knew this information (Table 5.8).
5.2.3 Follow-up rate and observation time Phvsician-based follow-up In the study design. an observation time of 12 months was planned for al1 patients. during which two follow-up examinations had tu be performed in addition to the baseline examination.
However. this duration was not possible to obtain for al1
patients. among others for some of those seen by specialists due to the delay in their recruitment. Of the 1531 patients. 1435 patients (93.7%) had at least one clinical examination during follow-up : 1235 (80.6%) had two follow-up clinical examinations and 200 patients (13.156) had one follow-up clinical examination. Follow-up information
is lacking for 96 patients (6.3%). The observation time (in months) was calculated as the difference between the date of the last visit to the study investigator and the date of the baseline visit. divided by 30.4.
Table 5.9 presents the number of patients for whom follow-up information is available. as well as the number of months of observation. by physician type and country. A total observation tirne of 14778 rnonths was captured in the study. The rnean observation-
time was 11 months in the general practitioner-based sample and 9 months in the specialist-based sarnple. Patient-based economic evaluation Three phone calls to patients were scheduled at an interval of three months starting 3 months after the baseline visit to the investigator. At least one telephone questionnaire could be administered to 1457 (95.2Oh)of the 1531 patients: 1 questionnaire to 101 (6.6%) patients, 2 questionnaires to 226 (14.8%) patients. and 3 questionnaires to
1130 (73.8Oh) patients. By protocol, each questionnaire addressed events having occurred during the last three months. This time period was chosen as the longest period that could be explored by phone interview with acceptable validity.
An
observation time of 3 months was taken when the two phone calls were separated by 3 months or more. Thus, the range of the observation time is 3 to 9 months.
Table 5.10 presents the statistics of the observation time in the patient-based survey. by physician type and country. A total observation time of 11175 patient-months were
captured in the study. The mean observation-time was 7.88 months in the general practitioners-based sample and 7.33 months in the specialists-based sample.
5.2.4 Discussion The recruitment statistics show different profiles of patients between countries and types of physicians. These differences reflect both the type of physicians (particularly the type of specialists) who participated in the VEINES study, as well as differences in medical practices. Patients consulting specialists were on average younger. be!ter educated and had milder forms of venous disorders (namely telangiectasia) than patients consulting general practitioners. Milder forms were also observed in French patients consulting general practitioners. One of the objectives of the sampling procedure was to increase the proportion of male and ulcer patients. Cornparison of the registration and recruitment data shows that the proportion of patients with ulcer was doubled in the recruitment phase and the proportion of male patients was increased by approximately 50%.
The follow-up rate was high in this study. In the patient-based evaluation, 80.6% of patients attended the two follow-up visits, and 13.1% of patients attended one of them. In the patient-based evaluation, 73.8% answered to 3 telephone interviews, 14.8% to
226 and 6.6%to 1. Losses to follow-up were therefore limited to 6.3% in the physicianbased survey and to 4.8% in the patient-based survey.
93
-
Table 5.1 Characteristics of registered patients general practitioner-based patient population Belgiurn (n=985)
Gender men women unknown
Aae (years) c 18
18-29 30-39 40-49 50-59 60-69 70-75 >75 unknown
Clinical ~resentation symptoms varicose veins skin changes ulcer unknown Certaintv of venous oriqin cerîain probable possible unknown T v ~ e sof treatment nme mentioned none counselling physiotherapy venoactive dnig nonvenoactive or unknown drug compression sclerotherapy surgery (past or planned) local therapy unknown treatment
France (n=1449)
ltaly (n=694)
AH countries (n=3128)
-
Table 5.2 Characteristics of registered patients specialist-based patient population
Belgium (n=340)
Gender men Jvornen unknown
Acle (years) 18 18-29
c
30-39 40-49 50-59 60-69
70-75 >75 unknown
Clinical ~resentation syrnptorns varicosities varicose veins oedema skin changes ulcer unknown
199 58.5 204 60.0 98 28.8 62 18.2 61 17.9 38 11.2
26
7.6
CeRainty of venous o r i ~ i n 259 76.2 certain probable 21 6.2 possible 20 5.9 unknown 40 11.8 T Y D ~ofStreatrnent none mentioned none counselling physiotherapy venoactive drug otherlunknown drug compression sclerotherapy surge'Y local therapy unknown treatment
France (n= 1289)
ltaly (n=420)
Quebec
(n=511)
All (n=2560)
Table 5.3 Distribution of subjects inciuded in VEINES by CEAP class, countiy and type of investigator general practitioner-based sample.
-
-
-
--
96
1 n O h
28 3.0 5 1.4 16 4.3 7 3.5
75 7.9 20 5.4 41 t 0.9 14 6.9
O n
All Belgium
France ltaly
2 n
CEAP class 3 n
4 n
5 n
Oh
%
O!
Oh
136 14.4 37 10.0 71 18.9 28 13.8
370 39.1 171 46.3 108 28.8 91 44.8
75 7.9 42 11.4 27
239 25.2 81 22. O 1O5 28.0 53 26.1
7.2 6 3.0
6
n Y0 24 2.5 13 3.5 7 1.9 4 2.0
al1 n Oo /
947 100.0 369 100.0 375 100.0 203 100.0
fable 5.4 Distribution of subjects included in VEINES by CEAP class, country and type of investigator specialist-based sample.
-
O
1
n
n Oh
%O
Al l
Belgium France Italy Quebec
30 5.1 7 6.1 8 3.7 10 6.4 5 5.3
126 21.6 24 20.9 49 22.5
25 16.0 28 29.5
2 n
CEAP dass 3 n
O h
128 21.9 26 22.6 52 23.9 32 20.5 18 19.0
4 n O/o
61 10.5 8 7.0 18 8.3 25 16.7 9 9.5
186 31-9 35 30.4 73 33.5 54 34.6 24 25.3
5 n ?'!a
42 7.2 12 10.4 14 6.4 7 4.5 9 9.5
6
n %O
Il 1.9 3 2.6 4
1.8 2 1.3
2 2.1
all n O h
584 100.0 115 100.0 218 100.0 156 100.0 95 100.0
Table 5.5 Distribution of subjects included in VEINES by CEAP class, age (years) and gender general-practitioner-based sample.
Males Ail
Females All
-
O
1
2
CEAP class 3
5 2.3
6 2.7
68 30.9
22 10.0
90 41 .O
16 7.3
13 5.9
220 100.0
23 3.2
69 9.5
171 23.5
114
15.7
280 38.5
59 8.1
11 1.5
727 100.0
4
5
6
al1
Table 5.6 Distribution of subjects included in VEINES by CEAP class, age (years) and gender specialist-based sam ple.
-
CEAP class
Males All
1 8-39
al1
O
1
2
3
4
n
n
n
n
n
O10
O h
n Oh
5 n
6
n
O!
O ?
Oh
Oh
55
19 15.8
22 18.3 9 47.4
O
5
7 5.8 1 5.3 3 5.4 3
23
9
2.2
0.0
11.1
6.7
51 .l
20.0
22 4.7 9 6.9 11 4.9 2 1.9
1 22 26.3 58 44.3 54 24.0 10 9.3
1 06 22.8 35 26.7 55 24.4 16 14.8
54
131 28.2 18 13.7 67 29.8 46 42.6
8 6.7 2 25.0
40-59
5
60-75
8.9 1
4 3.3 O 0.0 4 7.1
O */
8
14.3
45.8 4 21 .O 28 50.0
3 15.8 7 12.5
5
4.2
O 0.0 1
1.8 4 8.9
120 100.0 19 100.0 56 100.0 45 100.0
Fernales
Al l 18-39 40-59 60-75
11.6 11 8.4
25 11.1
18 16.7
23 5.0
6 1.3
O
O
0.0
0.0 3 1.3 3 2.8
10 4.4 13 12.0
464 100.0 131 100.0 225
100.0 1 08 100.0
Table 5.7 Distribution of subjects included in VEINES by type of investigator (generalistor specialist-based sample) and educational level Generalists Specialists al1 (n=1531) (n=947) (n=584) n (Oh) n (%) n (%) - --
Hiqhest level of education achieved None or primary school Lower secondary or technical school Higher secondary school or university Unknown
221 25.6 413 47.8 230 26.6 83
66 13.5 254 51.9 169 34.6 95
287 21.2 667 49.3 399 29.5 178
Table 5.8 Distribution of subjects included in VEINES by type of investigator (generalistor specialist-based sample) and income Generalists Specialists al1 (n=1531) (n=947) (n=584) n (%) n (%) n (Oh) -
-
--
-
Net household income (in FQ < 5,000
5,000 - 9,999 2 10.000 unknown
109 19.3 261 46.2 t95 34.5 382
42 15.1 89 32.0 147 52.9 306
151 17.9 350 41.5 342 40.6 688
Table 5.9 Physician-based economic evaluation: number of patients followed and number of person-months of observation between the first and the last dinical examination, by physician type and country Belgium General ~ractitioners al1 patients n (96)patients followed patient-months mean (std. error) range Soecialists al1 patients n (%) patients followed patient-months mean (std. error) range All phvsicians al1 patients n (96) patients followed patient-rnonths mean (std. error) range
France
ltaly
Quebec
All countries
1O0 Table 5.10 Patient-based follow-up: number of patients followed and number of person-months of observation between the first and the last interview, by physician type and country Belgium General ~ractitioners al1 patients n (Oh) patients followed N patient-months mean (std. error) range
369 364 (98.6) 2964 8.14 (0.05) 3-9
Specialists ail patients n (%) patients followed N patient-rnonths mean (std. error) range
115 113 (90.4) 888 7.86 (0.12) 3-9
Al! ~hvsicians al1 patients n (%) patients followed N patient-rnonths rnean (std. error) range
484
477 (98.6) 3852 8.08 (0.05) 3-9
France
ltaly
218 192 (88.5) 1113 5.80 (0.14) 3-9
156 155 (91.O) 1290 8.32 (O. 10) 3-9
Quebec
All countries
101
6. Classification and clinical description of vaiicose veins
Several classification systems of venous disorders have been developed for the epidemiological investigation of varicose veins. They were presented in Chapter 3. Section 3.1-5. Widmer et al. (1978) proposed a classification with two levels based matnly on objective signs. The first level included three classes of varicosities (hyphenwebs. reticular varicose veins and trunk varicosities) that could be combined with three classes of a second level related to 'chronic venous insufficiency' (Le. dilated subcutaneous veins, hyperpigmented or depigmented areas and open or healed leg ulcer). Widmer's classification attempted to differentiate between 'uncomplicated' varicose veins and varicose veins with 'chronic venous insufficiencyJ(CVI) based on objective signs. It has
been used in a large number of epidemiological studies. In practice, hyphenwebs and reticular varicose veins were often grouped together and CVI was generally used only in combination with trunk varicosities. However, there was no explicit classification that combined varicose veins and other signs of venous disorders. Moreover. Widmer's classification is not exhaustive. For example. it does not consider presentations such
as varicose veins with oedema but without skin changes. Another system is the 'NHS' classification, proposed by Robbins et al. (1992) following Porter's classification of venous disorders (Porter et al.. 1988).
lt is a complex
classification system mixing signs and symptoms. The definitions of the categories are however not straightforward and may lead to different interpretations. Other 'ad-hoc' classification systems have been used in other studies. These systems were either insufficiently documented or not exhaustive enough to be used in this thesis. As none of the existing classifications was satisfactory. there was a need for developing an original one.
This chapter presents the classification system developed for this thesis as a means for stratifying and cornparing patients with varicose veins
The distributions of socio-
dernographic, anthropometric and clinical variables in the categories of the classification are presented. A clinical description of varicose veins diagnosed in subjects examined by specialists is also reported.
6.2 The 'VEINES' classification of varicose veins.
The 'VEINES' classification system has been developed following the 'new' CEAP classification of venous disorders published during the implernentation of the VEINES study (Porter et al., 1995). The principle underlying the CEAP classification was to dissociate different dimensions of venous disorders. such as objective signs. symptoms and presence of reflux. It is all-inclusive and rnutually exclusive. Since the original classification is not specific to varicose veins. it could not be used as such for the purpose of this thesis and needed to be adapted. The 'VEINES' classification contains 6 categories defined as follows :
no signs or telangiectasia alone O
varicose veins alone (with or without telangiectasia)
0
varicose veins and oedema (with or without telangiectasia) varicose veins and skin changes (with or without oedema or telangiectasia) varicose veins and healed ulcer (with or without any of the above signs) varicose veins and open ulcer (with or without any of the above signs).
'Skin changes' is a composite variable created following the definition used in the CEAP classification. It was equal to 1 in presence of at least one of the following signs: hyperpigmentation. eczema or denatitis along the venous pathway. lipodermatosclerosis. and atrophie blanche. It took the value O if al1 the signs were absent. The rationale of this grouping was to provide a scale related to the presence of varicose
veins and the severity of the associated venous disorders. Contrary to the CEAP classification, it allows to operate the following contrasts :
patients without signs or with telangiectasia alone vs. patients with varicose veins alone patients with varicose veins alone vs. ail patients with varicose veins and other signs of more severe venous disorders patients with varicose veins alone vs. patients with varicose veins and each of the other signs of venous disorder. It must be emphasised that the above classification was created only for operational purposes. It was not the intent of this thesis to validate it against other classification systems, although the richness of the data available is probably unprecedented for a classification of venous disorders. The categories of the 'VEINES' classification are referred to hereafter as 'varicose vein categories'.
6.3 Description of varicose vein categories
6.3.1 Socio-demographic characteristics
Of the 1531 patients participating in the VEINES study. 218 had oedema, skin changes or healedlactive ulcer without varicose veins and were therefore not concerned by the classification. On the 1313 rernaining patients. 259 (19.7%) had no other sign than telangiectasia, 367 (28.0%) had varicose veins alone. 125 (9.5%) had varicose veins and oedema. 431 (32.8%) had varicose veins and skin changes. 100 (7.6%) had varicose veins and healed ulcer. and 31 (2.4%) had varicose veins and active ulcer. The distribution by type of specialists and country (Table 6.1) shows comparable
percentages of patients with varicose veins alone. varicose veins with skin changes and varicose veins with ulcer.
Marked differences are seen between generalists and
specialists in the proportions of patients without varicose veins (12.4% and 32.3% respectively) and with varicose veins and skin changes ( 3 7 . 7 2 and 24.4% respective1y ).
Important inter-country differences were observed in the GP-based
sarnple. with larger proportions of Belgian patients having varicose veins with skin
changes or ulcer (healed and active). Among GP patients, French patients had milder signs of vanous disease associated with varicose veins. Of the 1313 patients concerned by the classification. 1013 were women (77.2%) and 300 were men (22.8%).
The mean age of women was 53.1 years.
Their age
distribution (Table 6.2) shows a marked increase of the mean age according to the severity of signs associated with varicose veins, except for varicose veins associated with active ulcer. Men were on average older that women (mean=57.5 years), with smaller age differences between clinical categories. Table 6.1 Distribution of subjects incîuded in the VEINES study, by type of investigator. country and varicose vein category (percentages are row percentages)
Country (sample size)
No signs or tetangiectasia (n=259) n %
Generalist-based sample All (830) 103 12.4 Belgiurn (335) 25 7.5
France (317)
57
18.0
ltaly (178)
21
11.8
Specialist-based sarnple All (483) 156 32.3 Belgium (98) 31 31.6 France (168) 57 33.9 ltaly (129) 35 27.1 Quebec (88) 33 37.5
Varicose veins alone (n=367) n %
W and oedema (n= 125) n %
W and skin changes (n=431) n Oh
W and healed ulcer (n=100) n O h
W and active ulcer (n=31) n Oh
Table 6.2 Age of subjects included in the VEINES study, by gender and varicose vein category No signs or telangiectasia (n=259)
Varicose veins alone (n=367)
W and oedema (n=125)
W and skin changes (n=43t)
W and heaied ulcer (n=l00)
W and active ulcer (n=31)
Men -
n mean (se.) range
?(se.) mean range
An interesting feature of the distribution of educational level in each clinical category (Table 6.3) is the increase of the proportions of patients with a low level of education with the severity of the signs associated with varicose veins (from 17.6% to 28.6%). An inverse trend was observed for the proportions of patients with a high level of education.
The differences observed between patients without signs or with
telangiectasia alone and patients with varicose veins alone is also striking. These differences could be due to several factors and rnay be associated with referral patterns: patients with a high level of education are more likely to worry for the presence of telangiectasia and consult a physician. Table 6.3 Distribution of subjects included in the VEINES study, by educational level and clinical category
Educational level
No signs or telangiectasia (n=259) n
%
16
7.1
Varicose veins alone (n=367) n %
w and oedema (n=125) n Oh
W and skin changes (~431) n %
Low Intermediate
1O0 44.6
58 17.6 177 53.8
High
IO8 48.2
94 28.6
30 28.0
91
38
18
36
Unknown
34
20
W and healed ulcer (n=100) % n
W and acîive ulcer (n=3l) n %
30.6
121 30.6
24 28.6
57 46.3
183 46.3
46
54.8
14 48.3
23.0
14
16.7
6 20.7
16
9 31.0
2
6.3.2 Anthropometric measures There is a rnarked increase of body rnass index with categoties of varicose veins (Table 6.4). The average body mass index increases from 24.7 in patients wi?h varicose veins
alone to 31.3 in patients with varicose veins and active ulcer. The median in this category is also 31. which means that half of the subjects are beyond the threshold for obesity (BMI >30). Table 6.4 Height, weight and body m a s index (BMI) of subjects included in the VEINES study, by varicose vein category
No signs or telangiectasia (n=259)
Varicose veins alone (n=367)
W and oederna (n= 125)
W and skin changes (n=431)
W and healed ulcer (n= 100)
active and ulcer (n=31)
1 ,
Height n mean (s.e.) range Weight n mean (s.e.) range BMI -
n mean (s.e.) range
6.3.3 Associated signs
Telangiectasia were found in 80%-87% of the patients. irrespective of the clinical category they belonged to. Oedema was found in 40% to 60% of patients with varicose
veins and ulcer. and skin changes were found in about half of patients with healed or active ulcer. Skin changes are usually considered as the clinical stage that precedes the occurrence of ulcer.
Although it is possible that skin changes were not
appropriately recorded. this is unlikely because this sign was assessed using different questions concerning the aspect of the skin. The results suggest that. in patients with
@
varicose veins. ulcer could occur without skin changes. or that some of the uicers had a non venous cause.
Table 6.5 Distribution of signs of venous dirorders in subjects included in the VEINES study. by varicore vein category
Sign
Telangiecîasia
No signs or telangi-
Varicose
ectasia (n-259)
alone (n=367)
201
77.6
veins
293
79.8
W and oederna (n=l25)
W and skin changes (n=431)
W and healed ulcer (n=100)
W and active ulcer (n=3 1)
112 89.6
376 87.2
87 87.0
26 83.9
125 100.0
163 37.8
46
46.0
19 61.3
55 55.0
14 45.2
100 100.0
27 87.1
Oedema
O
O
Skin changes
O
O
O
431
Healed ulcer
O
O
O
O
Open ulcer
O
O
O
O
100.0
O
31
100.0
6.3.4 Clinical description of varicose veins in specialist-based sample
Distribution of varicose veins
In the specialist-based population (Table 6.6), the non saphenous domain was the most frequently affected by varicose veins (154i261 or 47.1%), either alone (43.3Oh)or in association with varicose veins located in the short saphenous domain (0.4%) or the long-saphenous domain (13.0%). In the saphenous domain, long saphenous veins were more frequently affected by varicose veins than short saphenous veins (51.3% vs. 7.6%). There were no differences between the two legs for the most frequently affected
domains. About 66% (1641327) of varicose veins affected the whole leg. In a limited number of patients. varicose veins were observed above the knee only (101347 or 4.0%). In about a quarter of patients. varicose veins (27.9%) affected the leg below the knee. Three patients were reported to have varicose veins on the foot only. There was no difference between the two legs for the most frequent location of varicose veins.
1O8 Table 6.6 Location of varicose veins' on left leg (n=257)
Varicose veins on right leg (n=275)
on any leg (n=327)
Domain short saphenous only long saphenous only non saphenous only short and long saphenous only shortsaphenousaridnon saphenousonly long saphenous and non saphenous only short, long and non saphenous non specified Location leg above knee only leg below knee only whole leg (excepted foot only) foot only non specified
-
ascertained in 584 specialists' patients
Pain on palpation of varicose veins
Pain on palpation of varicose veins was reported in 21% of patients. and this percentage was similar whether varicose veins involved the left leg or the right leg. Severity of varicose veins Specialists were asked to give a global rating of the severity of varicose veins on the lef? and the rights legs in four categories ('mild'. 'moderate'. 'severe' and 'very severe'). The percentages of mild. moderate and severe (including 'very severe') varicose veins were
27.2%. 51.4% and 21.4% respectively. Similar percentages were observed for varicose veins present on the left or the right side. Lateralitv of varicose veins Table 6.7 shows that bilateral varicose veins are found in 35.1% of the population of patients seen by specialists. and in 56.0% of patients with varicose veins (3971584). In 122 patients with unilateral varicose veins. these are seen more frequently on the right
leg (n=70) than on the left leg (n=52).
Table 6.7 Laterality of varicose veins
Varicose veins on right leg
no
Yes
total
Yes left leg
total
6.4 Discussion
To our knowledge, the 'VEINES' classification of varicose veins presented in this chapter is the first comprehensive classification that allows a distinction between patients with varicose veins alone and patients with various signs of more severe venous disorders. Following the CEAP classification. signs have been dissociated from symptoms and from the presence (and location) of reflux. The proportions of patients with reflux in various locations of the leg system are presented in Chapter 9 and symptoms associated with each category are presented below in Chapter 11.
This chapter offered descriptive data regarding the distribution of the varicose vein categories according to type of physician consulted, country. age, educational status and anthropometric measures.
These data showed important differences between
categories. It suggests that the pooling of patients with varicose veins and other signs of venous disorders into a single class of patients with .varicose veins'. as it is done in many epidemiological studies. can be misleading and mask important differences
between patients. These data and other data to be presented in subsequent chapters also contribute to a better understanding of factors that need to be considered in etiological studies on varicose veins.
7. Validity of the diagnosis of varicose veins by general practitioners
7. f Introduction In a limited number of studies, inconsistent results were found regarding the interobserver reliability of a clinical diagnosis of varicose veins.
These results raise
concerns about the vaiidity of measuring the presence or absence of varicose veins. Since the scope of the VEINES study was to assess outcornes presented by subjects with venous disorders in a natural environment, clinical examination was the only source of information used to determine the presence of varicose veins. In ternis of economic evaluation. the la& of objective information does not represent an issue if one is interested in venous disorders as ascertained by general practitioners. However. it would
affect the results of etiological investigations. If the diagnosis were inaccurate for some patients, results would relate to a mixed population of subjects with and without varicose veins. The concern seemed greater for general practitioners than for specialists. Many patients going to specialists are presumably referred by other physicians because a search for other causes of the disorder was negative or because the diagnosis was ascertained and required specific treatment.
Moreover. it was postulated that there was smaller
variability between specialists in deciding whether or not the disorden are of venous origin. even if disagreement may exist in detemining the severity and the best management
strategy. A study was therefore carried out in order to assess the validity of the clinical diagnosis of varicose veins made by general practitioners.
7.2 Objectives 7.2.1 Main objective To assess the validity of the clinical diagnosis of varicose veins made by general
practitioners in comparison to the diagnosis made by specialists. and to identify
e
variables associated with misclassification.
7.2.2. Specific objectives In order to examine the sources of disagreement found between general practitioners and specialists, additional questions were raised : Is disagreement explained by failure to correctly diagnose telangiectasia ? a
Is misclassification accounted for by the diagnosis made by a small number of general practitioners or is it a widespread problem ? Is misclassification accounted for by variability of the diagnosis made by special ists ?
a
Can a restricted definition of varicose veins improve the validity of the clinical diagnosis ? Are there any variables associated with differential misclassification ?
7.3 Patients and Methods 7.3.1 Patients and methods This study compares the result of the last clinical examination performed by general practitioners in the course of the VEINES study and the result of the clinical examination performed by 7 specialists in the Duplex study. A description of these studies was presented in Chapter 4. Varicose veins were assessed separately in each leg and a subject was classified with varicose veins if this diagnosis was recorded in one or both leg. The clinical examination made by general practitioners included the masure of the maximum diarneter of the largest varicose vein using a slide calliper graduated to 0.1 mm according to Cornu-Thénard's technique (Cornu-Thénard et al..
1991 ). The Duplex study included a questionnaire to be filled in by each of the seven specialists to whom participants were referred. This questionnaire was comparable to that filled in by investigators at each of the three follow-up visits. Specialists were instructed to perform and record the clinical evaluation before the duplex scan. Six of the 7 specialist were working in a hospital and in these centres the duplex scan was perforrned by another person than the specialist who made the clinical examination.
Data of the clinicai examination performed by a specialist were available for 144 of the 149 patients included in the VEINES study and selected to be referred to one of the
seven specialists participating in the Duplex study in Belgium. France or Italy. The protocol specified that patients in whom a duplex scan had been performed in the last 12 months should not be referred. At the selection phase. information regarding the presence of venous incompetence was available for only 41 (28.4%) of the 144 patients referred, but this information had been collected before the previous 12 months for the majority of cases.
Only 6 patients were referred to the specialist
although fhey had had a Doppler ultrasound (4) or a duplex scan (3) during follow-up (one patient had had both). These six patients were nevertheless kept for the analysis. It should be reminded that the selection of subjects was driven by an attempt to
increase the proportion of subjects in categories of the CEAP classification (Porter at al.. 1995) with a lower sample size.
This objective was only partly obtained. the
distribution of patients in CEAP classes O to 5 being 8.3%. 21.5%. 18.8Oh. 13.2%. 30.6% and 7.6% respectively (patients with active ulcer were grouped in class 5).
Secondly. according to the investigator Doppler ultrasound or duplex scanning had not been performed during the 12 months prior to being seen by the specialist.
The mean and median delays between the assessrnent made by GPs and the
assessment made in the Duplex study were 146 and 131 days respectively.
7.3.2 Restricted definition of varicose veins In an attempt to increase the validity of the diagnosis of varicose veins by general
practitioners. a stricter definition of the presence or absence of varicose veins was created by using clinical information collected during the follow-up time. It is generally accepted that varicose veins do not spontaneously resolve. However. among those included in the Duplex study. 7 of the 111 subjects with varicose veins at the baseline visit did not have varicose veins at the last follow-up visit. although they
were neither operated on nor had their veins sclerosed between the two visits. For 6 of these 7 patients, the presence of varicose veins was not confirmed by the specialists consulted in the context of the Duplex study. This suggeçts that the diagnosis of varicose veins made in these individuals by the general practitioner at the baseline visit could indicate a non pathological and transient increase in the volume of a vein at the time of the baseline visit, or an error in reporting. It could also be assumed that an average follow-up time of 11 months was too short for the occurrence of new varicose veins. Results of the Duplex study show that 5 of the 7 cases of 'new' varicose veins detected by the general practitioner at the last visit were not confirmed by the specialist consulted. An apparently new occurrence of varicose veins at the last visit could also indicate a misclassification at one of the two visits. or an error in reporting.
Consequently. a restricted definition of varicose veins was used : 'absence of varicose veins' rneant absence of varicose veins at the baseline and at the last follow-up visit. and 'presence of varicose veins' meant a reporting of varicose veins both at baseline and follow-up examinations.
7.3.3 Data analysis In the analysis. the diagnosis of varicose veins made by the general practitioner (GP) at the last follow-up visit (12 months after the baseline visit) was compared with the result of the specialist's evaluation.
The following parameters of validity were
computed : sensitivity. specificity. the positive and negative predictive values. as wel l as their 95% confidence interval. The number and proportion of disagreements between GPs and specialists were calculated for each general practitioner and specialist.
When error of diagnosis occurs. it is important to determine whether a subject has a presumed higher (or lower) probability of a diagnosis of varicose veins by a general practitioner because he presents (or lacks) characteristics suspected to be associated with the presence of varicose veins. In the Duplex study. specialists were kept blinded
of the characteristics of the subjects referred to thern. Therefore, it can be assumed that the diagnosis was unaffected by these variables.
On the other hand, if the
likelihood of reporting varicose veins in patients exarnined by general practitioners is influenced by the presence or absence of a suspected deteninant, the sensitivity and specificity of this diagnosis would Vary across categories of the suspected determinant. In order to examine whether the presence of suspected determinants influenced the diagnosis, separate calculations of sensitivity and specificity were made in categories of three types of variables : Variables suspected to influence the number of consultations to a general practitioner and. consequently. the likelihood of detecting a venous problem : age, gender, level of education and CO-morbidity. Variables suspected to influence the occurrence of varicose veins and, consequently. the probability of a positive diagnosis: number of pregnancies. obesity. materna1 history of varicose veins. history of leg trauma. and past thrombophlebitis. Variables indicating a severe venous problem and, consequently. influencing the likelihood that enlarged veins are ciassified as pathological varicose veins : venous oedema and skin changes.
The number of subjects in this validation study was too small to allow estimation of the validity of the diagnosis of varicose veins in the narrow categories of variables listed above. and levels of al1 variables were collapsed into two categories.
7.4 Results 7.4.t Main result Table 7.1 presents the validity of general practitioners' diagnosis of varicose veins in comparison to the diagnosis made by specialists. Table 7.1 Cornparison of the diagnosis of varicose veins (VV) made by general practitioners and by specialists
General practitioners
W No W Total
Specialists W No W 74 35 6 29 80 64
Total 109 35 144
sensitivity specificity positive predictive value negative predictive value
: 0.925 : 0.453
: 0.679 : 0.829
The prevalence of varicose veins measured by specialists was 0.555. The diagnosis of
varicose veins by general practitioners has a sensitivity of 0.925 (95% confidence interval : 0.867-0.983) and a specificity of 0.453 (95% CI: 0.331-0.575). The low specificity means that about half of those classified by specialists as varicose vein-free were wrongly classified by general practitioners. Due to the relatively high prevalence
of varicose veins in this particularistic sample (56%). the positive predictive value is still 0.68.
7.4.2 Effect of errors in the diagnosis of telangiectasia A possible source of error in the diagnosis of varicose veins by general practitioners is
the diagnosis of telangtectasia as varicose veins.
Table 7.2 below presents the
distribution of patients according to the diagnosis of telangiectasia and varicose veins made by general practitioners and specialists.
Table 7.2 Distribution of patients according to the presence of telangiectasia (TEL) and varicose veins ON) on either leg, by general practitioners and specialists' diagnosis
r-
Specialists' diagnosis
No TEL, NoW
TEL, NoW
No TEL. W
TEL and W
All subjects
1 TEL. No W 1 No TEL. W
2
6
3
1
12
6
15
1
1
1
3
4
4
and W
7
24
10
56
16
48
18
62
Generalists' diagnosis
No TEL, No W
1 TEL
1 All subjeds
23
l2 97 la
l 1
1
1
A first interesting observation is the fact that specialists reported telangiectasia (7
subjects) and varicose veins (4 subjects) in 12 patients for whom neither telangiectasia nor varicose veins were reported by general practitioners. Secondly. 3 subjects were classified with varicose veins alone by general practitioners and with telangiectasia alone by specialists. Thirdly, a total of 8 subjects diagnosed with telangiectasia alone by general practitioners were classified as presenting varicose veins by specialists. Therefore. an important source of errors appears to be in the recognition andlor reporting of telangiectasia which were either not recognised andlor not reported. or diagnosed as varicose veins. This explains why the validity of general practitioners' diagnoses is not improved when telangiectasia are included in the varicose vain definition (sensitivity: 0.89, specificity: 0.17 ) .
7.4.3 Analysis by general practitioner
The 144 patients included in this analysis were referred by a total of 48 general practitioners. with a mean of 3.0 patients per physician (range : 1-7). The 41 patients with a discordant diagnosis were referred by 26 GPs. who had enrolled a total of 87 patients into the Duplex study (mean per GP: 3.3). Misclassification occurred by 54% of GPs (about equally distributed in Belgium. France and M y ) and in 28% of patients. For one GP. there was a discordance for 4 of 5 subjects. but the exclusion of this GP
a
did not substantially improve the values of the validity parameters. It can be concluded
that the poor specificity of the diagnosis of varicose veins is distributed across a large nurnber of general praditioners (about 50%).
7.4.4 Analysis by specialist Statistics by specialist are presented in Table 7.3. Table 7.3 Numbers and proportions of disagreements between general practitioners and specialists, by specialist
Specialist 1 2 3 4 5 6 7 Al l
N subjects examined
37 8 9 12 15 13 50 144
N subjects with disagreement
Proportion of disagreement (Oh)
15 5 1 2 3 1 14
40.5 62.5 11.1 16.7 20.0 7.7 28.0 28.5
41
Specialists 1 to 4 are from Belgium, specialists 5 and 6 from France. and specialist 7 from Italy. The proportion of disagreements is much higher for Specialists 1 and 2 (40.5% and 62.5%) than for the others. Exclusion of these two specialists increases the
specificity from 0.45 to 0.57. Importantly. it should be underlined that specialists had at their disposal a piece of information unavailable to general practitioners : the result of the duplex test. Although specialists were instructed to record the results of their clinical examination before doing the duplex test (and the Duplex test was performed by a different person in some centres). it was not possible to verify that they cornplied with this instruction.
7.4.5 Effect of using a iestiicted definition of varicose veins This restriction lad to the exclusion of 16 subjects with varicose veins detected at only one of the clinical examinations. Table 7.5 presents the validity of the restricted definition of varicose veins presented above. Table 7.5 Cornparison of the diagnosis of varicose veins made by general practitioners and by specialists Specialists W No W General practitioners -
-
W No W Total
72
30
4 76
22 52
Total 102 26 128
sensitivity : 0.947 specificity : 0.423 positive predictive value : 0.706 negative predictive value : 0.846 prevalence : 0.594
-
These results show that restricting the definition does not affect the sensitivity and specificity estimates. The specificity is not increased since 12 of the 16 subjects with varicose veins at one of the two examinations did not have varicose veins according to the specialists' examination. and these 12 subjects were almost equally classified as with (5) or without (7) varicose veins by general practitioners.
7.4.6 Assessment of variables infiuencing the diagnosis of varicose veins The last clinical examination made by general practitioners was perforrned on average 11 months after the baseline visit of the VEINES study. where information on variables
suspected ta be associated with errors of diagnosis took place.
Therefore. the
restricted definition of varicose veins was used : a patient was considered as having varicose veins if the diagnosis was made both at the baseline and at the follow-up visit.
and without varicose veins if varicose veins were absent at both visits. Subjects with varicose veins reported at only one of the two examinations were excluded from the analysis. Three sets of variables were examined :
variables suspected to influence the number of consultations to a general practitioner and. consequently, the opportunity to detect venous problems (Table 7.6)
variables suspected to influence the occurrence of varicose veins and. consequently, the classification of enlarged veins on the legs as varicose vems (Table 7.7) variables indicating the severity of the venous disorder and, consequently. the misclassification of enlarged veins on the legs (Table 7.8). Among the variables investigated in Table 7.6. a trend towards an effect on the diagnosis (increased sensitivity and decreased specificity) is observed with high educational level. However. even with collapsed categories, large confidence intervals are obsewed and the number of observations do not allow statistical confirmation of the existence of this effect. A decrease of only the specificity of the diagnosis is also seen with increasing age and co-morbidity. Tablc 7.6 Sen~itivihand spccificity of the diagnosis of \.aricose veins (VV) by gcncml pracritioners in cornparison to ihc diagnosis madc by specialists. by categop of variables suspectcd to inilucncc the nuniber of consultations ro general pnctitioners
Variable All subjects
Specialists' diagnosis W No W (n) (n) 52 76
Sensitivity
estirnate
95% CI
0.95
0.90-1 .O0
Specificity
estirnate
95% CI
Gender
- male
- female Age (years) - 18-59
- 260
45
5
26 49
0.88 0.98
0.76-1 .O0 0.94-1 .O0
34 18
27 49
0.96 0.94
0.89-1.00 0.87-1 .O0
8 48
23 52
0.87 0.98
0.73-1 .O0 0.94-1 .O0
30
27 49
0.93 0.96
0.83-1 .O0 0.90-1 .O0
Educational level
- none or primary school - secondarylhigh school Co-rnorûidity
- no
- yes -
---
22 -
lncludes at least one of the following conditions : diabetes, hypertension, cardiac disease. renal insufficiency. chronic respiratory disease, osteoarthritis, hepatitis, hypothyroidism, peripheral arterial insufficiency of the lower legs. lymphoedema.
Table 7.7 shows the results regarding the effect of variables suspected to influence the occurrence of varicose veins and. consequently, the probability of a positive diagnosis. A trends towards differential misclassification is observed for reproductive history.
history of past thrombophlebitis and history of leg trauma. No misclassification seems to exist according to the presence of varicose veins in the mother. or with the level of physical activity. A decrease of the specificity of the diagnosis of varicose veins is seen with high body mass index. Table 7.7 Sensitivity and specificity of the diagnosis of varicose veins (VV) by general practitioners in cornparison to the diagnosis made by specialists, by category of variables suspected to influence the occurrence of varicose veins
Variable All subjects
Specialists' diagnosis No W W (n) (n) 52 76
Sensitivity estimate 0.95
95% CI
0.90-1.O0
Specificity estimate 0.42
95% CI
0.29-0.56
Body mass index
- c 25
27 25
48
6 36
20 40
-
45 7
55 21
-
33 19
46 30
13 37
27
47
69 7
- 225
28
Number of pregnancies
-O
-11
Past thrornbophlebitis no yes Mother with VV no yes Physical activity IOW - moderate to high History of leg trauma no yes
-
-
5
47
Table 7.8 presents sensitivity and specificity estimates in relation to the presence of
oedema and skin changes (including healed and active ulcer). as ascertained by the general practitioner at the last clinical examination.
The data suggest that the presence of oedema is not associated with a different probability of correctly diagnosing or of reporting varicose veins. On the contrary, specificity is decreased in the presence of skin changes. Table 7.8 Sensitivity and specificity of the diagnosis and recording of varicose veins (VV) by general practitioners in cornparison to the diagnosis made by specialists, by category of variable indicating a severe venous disorder
Variable All subjects
Specialists' diagnosis No W W (n) (n) 52 76
Sensitivity
Specificity
estimate 0.95
0.90-1.O0
estirnate 0.42
0.94 0.96
0.88-1.O0 0.88-1.O0
0.38
95% Cl
95% CI
0.29-0.56
Venous oedema
- yes Skin changes ' - no - yes
39 13
52 24
0.44
0.28-0.59 0.12-0.65
lncludes healed and active ulcers
7.5 Discussion 7.5.1 The problem of diagnosis of varicase veins The only universaily accepted opinion regarding the definition of varicose veins is precisely that there is no generally accepted definition. and that one is critically needed.
The definition proposed by Arnoldi in 1957. at a time when objective
investigations were uncornmon. has been used in many studies on venous disorders using either clinical examination or selfieporting of varicose veins. Arnoldi's definition
of varicosity as "any dilated, elongated or tortuous veins, irrespective of size" leaves room for a subjective appreciation of the "dilated. elongated and tortuous" aspect of a vein. As stressed by several authors (Callam. 1996; Evans et al., 1996:De Backer.
1998). interpretation of Arnoldi' definition. e.g. the inclusion or exclusion of srnall
varicosities, varies between studies.
Few objective diagnostic tools are available to help physicians adopt a standard and validated clinical definition of varicose veins.
Cornu-Thénard et al. (1991) have
proposed a system of classification based on wooden pegs of unequal diameters to measure the maximum diameter of the largest vein. In one study, this system was validated by the inventors of the method against specialists' clinical assessment. obtaining favourable results. To our knowledge, this study has not been replicated. The principle of measuring the diameter of the vein (currently a slide calliper measured to the nearest 0.1 millimetre is used) has not gained much popularity among specialists of venous diseases, perhaps because as a measure of severity of varicose veins it cannot match the amount of information provided by Doppler ultrasound and duplex scanning. Laboratory tests, especially duplex scanning, are now frequently employed and promoted (Evans et al.. 1993). Applied to the examination of varicose veins, they represent a measure of severity and prognosis (Evans et al.. 1995) rather than of diagnosis because varicose veins can occur without valve incompetence. Various other methods have been used in epidemiological studies to reduce variability
in measurernent. such as taking photographs (McAdam et al.. 1976: Widmer et al.. 1978: Evans et al.. 1998) or joint classification by various observers (Latto et ai.. 1973. Evans et al.. 1998), but none of these methods are useful for clinical practice. In the absence of an objective methsd available for clinical practice. clinical examination remains the main source of information. Assessing its validity is another difficulty. In fact. there is no universal gold standard in phlebology because the purpose of diagnostic techniques differs, from anatomical description to functional investigation (Abenhaim et al.. 1999). There have been very few attempts in the past to assess the validity of clinical diagnosis of varicose veins. One study assessed the validity of physical examination to detect the presence of venous incompetence in cornparison to quantitative plethysrnography and colour Doppler (lafrati et al.. 1994). The study was carried out in 45 consecutive patients (59 legs with varicose veins) presenting to a university
vascular laboratory. The great saphenous vein was classified as abnormal if it was felt
to be dilated and varicose on palpation by an experienced vascular surgeon. The cough impulse was also checked over the inguinal area when possible. This physical examination had a sensitivity of 43%. a specificity and a positive predictive value of 100%. and a negative predictive value of 58%. These results suggested that physical
examination is not sensitive enough to detect venous reflux. In the absence of an accepted definition and diagnostic criteria, diagnostic test studies have generally focused on the reliability of measurement systems rather than their
validity (McAdam et al; l976,Widmer. 1978;Cornu-Thénard et al.. 1991).
7.5.2 Strengths and weaknesses of this study In this study, we measured the validity of the diagnosis of varicose veins made by general practitioners in 144 subjects, in cornparison to the diagnosis made by specialists after a median delay of 131 days. It should be stressed that the primary atm of the study was to measure the prevalence of venous incompetence in subjects with
venous disorders at various stages of severity. The design would certainly have been different had a diagnostic test study been the main objective. The long delay between the h o examinations represents an important weakness of this study. Two features of
the design were nevertheless introduced in order to approach what would have been be done in the situation of a diagnostic test study :
1) The two evaluations were ~ndependeniand blinded to each other. The clinical
examination made by the general practitioner (GP) took place several weeks before the visit to the spectalist. and specialists were aware of neither the results of GPs' assessrnent nor the patients' medical history.
2) Subjects' eligibility for inclusion into the study was determined by the study
CO-
ordinators. Two main criteria were used in order to avoid interference by the patient's medical history on their referral to the specialist (and therefore on the probability of an abnormal test of venous incompetence). Firstly. the selection of
S U ~ J ~ C was ~ S driven
by an attempt to increase the proportions of subjects in each category of the CEAP classification (Porter at al.. 1995). Secondly. according to the invesfigator Doppler
ultrasound or duplex scanning had not been performed during the 12 months prior to being seen by the specialist. According to the specialists' diagnosis, the prevalence of varicose veins was 55.5% in the subjects referred to them. The diagnosis made previously by general practitioners has a sensitivity of 0.925, a specificity of 0.453, a positive predictive value of 0.679. and a negative predictive value of 0.829. These results mean that general practitioners are able to detect the majority of subjects with varicose veins (92.5%),but that their diagnosis is specific in only 45.3% of the cases, meaning that less than half of the individuals without varicose veins are correctly classified. An important finding in relation to this level of disagreement is the fact that it was obsewed in the three countries where the study was carried out (Belgium. France and Italy) and concerned 26 of the 48 general practitioners participating in the study.
Thus. the
problem
appears to be widespread.
7.5.3 Possible causes of disagreement Before discussing the implications of such disagreement. it is necessary to consider its causes in order to dissociate those attributable to weaknesses in the study design from those attributable to problems of measurement.
Clinical disagreement
due to
measurement problems is in principle amenable to improvement (Sackett et al.. 1985. p.32). The causes of disagreement have been classified into five categories using the
terminology proposed by Sackett et al. (1985. p. 33) : biological variation in the systern being examined effects of illness and medication ensnarement by diagnostic classification schemes entrapment by prior expectation a
simple ignorance.
Bioloiaical variation in the svstem beina examined There was a median interval of 131 days between the two examinations. Since the sequence of measurements was identical for al1 cases (measurements made by general practitioners always preceded those made by specialists), it is theoretically possible that false positive cases concern subjects who actually had varicose veins when examined by a general practitioner (GP), but in whom varicose veins were not detected by the specialist. This sequence could be explained by two hypotheses . varicose veins resolved spontaneously, or they did not appear at the second examination. Firstly, spontaneous resolution of varicose veins is exceptional. De Backer (1998) asserts that it is well known that varicosities c m appear during pregnancy and that they generally disappear after delivery. but this statement is not docurnented. Even so. none of the women were pregnant at the time of the clinical examination made by the general practitioner. It is not impossible that in some women who had previously been pregnant, varicose veins disappeared between the two examinations. Among the 35 false positive cases. only 3 were women aged less than 40 years old and could possibly be concerned by this hypothesis. A time-dependant biological variability would also be associated with an increase of
validity as the delay between the h o examinations shortens. This hypothesis is not supported by the sensitivity and specificity values calculated according to the delay between the two examinations : delay of 4 months or less: 94% and 60%: delay of 5 months : 88% and 35%: delay of 6 rnonths: 92% and 40°h: delay higher than 6 months: 100% and 40%.
Secondly, the hypothesis of biological variability would be supported if varicose veins that resolved were small ones when exarnined by the GP. Although misclassification of small varicose veins may be independent from the delay between the two examinations. it would be possible that external conditions such as climatic conditions or the conditions for the clinical examination (e.g. patient's position. lighting conditions. etc.) influence the appearance and prominence of varicose veins and therefore the likelihood of being detected. There is some evidence to support the influence of the
site of the veins on the diagnostic validity : the maximum diameter of the largest vein has a median value of 61 mm in falre positive cases and of 91 mm in true positive cases. Effects of illness and medication Under this sub-heading is considered the possible effect of sclerotherapy and surgical treatment (although Sackett refereed to the effect of illness and medication on the conditions of the clinical examination). Since the visit to the specialist was performed after the end of the follow-up period. there is no information about possible treatments received between the two examinations. From data collected during follow-up, 6.7% of patients recruited by GPs experienced at least one of those procedures during an average of 11 months of follow-up. On the basis of this figure. about 3 (3.2%) of the 109 individuals diagnosed with varicose veins could have received sclerotherapy or
surgery during the average delay of 146 days between the two examinations, a figure that is much lower than the 35 false positive cases presented in Table 7.1.
Ensnarement bv diamostic classification schemes This category refers to the use of different definitions and classifications by general practitioners (GPs) and specialists. There is no doubt that such a problem exists. given the imprecision of current varicose veins definition.
It is also most likely that the
definition varies between GPs and specialists. This variation is certainly a cause of disagreement between physicians. Entrapment bv ri or expectation One tends to find what one expects or hopes to find. In this study. the maxim could apply to both general practitioners and specialists.
For operational reasons. the
reference clinical examination and the duplex scanning were performed in the same subject on the same day. in the same centre and. in one centre, by the same investigator. Although investigators were instructed to perform and record the result of their clinical examination prior to performing the duplex scan there is no guarantee that
they followed this instruction. It is therefore possible that the diagnosis of varicose veins was influenced by the result of the test.
The specialist investigators were
carefully selected for their professional expertise in the field of varicose veins. but whether their diagnosis is valid and recorded independently from the result of the duplex scan has not been ascertained. The influence of several variables on the diagnosis made by general practitioners is strongly suggested by the results of this study. They show that differential misclassification of leg veins exists in the presence of characteristics generally suspected to be associated with the occurrence of varicose veins. such as body mass index. number of pregnancies, past thrombophlebitis and history of leg trauma (Table
7.7). Misclassification exists also in the presence of skin changes (Table 7.8) and with greater age (Table 7.6.). Therefore, there seems to be a tendency to classify enlarged veins or small varicosities as "varicose veins" in the presence of a factor suggesting that the subject is at higher risk of presenting varicose veins. This phenomenon is also related to the tendency 'to record inference rather than evidence' (Sackett. 1985. p. 32) : the physician would interpret and report varicosities or non-pathological enlarged
veins as varicose veins because this inference offers a more comprehensive and quicker-to-report description of a whole clinical picture.
However. the drawing of
inferences. rather than the recording of clinical signs. is precisely the main source of disagreement between physicians and represents the diagnostic challenge. Simple ignorance Whether misclassification of varicose veins can also originate from a lack of knowledge about differences between small varicosities (or telangiectasia) and varicose veins cannot be fully ascertained without investigating at which step of the clinical examination a false positive diagnosis of varicose veins was established. It is quite possible that the error occurred at the final stage of the recording of the diagnosis due to a confusion in the terminology. This hypothesis is consistent with the data reported in Table 7.2. showing that. in 53% of subjects (771144). generai practitioners failed to report the same clinical status in terms of presence of telangiectasia and varicose veins
as reported by specialists.
'Ignorance' could also be invoked to explain that
telangiectasia were detected by specialists in 14 of the 144 subjects examined (97%) in whom no such diagnosis was made by generalists.
The problems of definition and lack of knowledge have been mentioned as possible causes of misclassification of varicose veins. One possible way to distinguish between these two causes would be to examine the validity of the diagnosis of varicose veins made by specialists.
We are not aware of such an investigation having been
performed. However, the poor reliability found in the scoring of photographs and in assessing the outcorne of surgery has been alluded to above. In the Paris policemen study. the prevalence of varicose veins, observed by 12 physicians who examined at least 200 men varied from 14% to 40%. It is therefore quite possible that lack of agreement also existed among the 7 specialists participating in our validity study. but this hypothesis cannot be examined with our data. In summary, several factors could explain the misclassification of varicose veins : differences in definition. classification and terminology used by the investigators, the influence of variables judged by the examiner to be associated with the risk of varicose veins, the effect of treatments administered between the two examinations and the effect of biological variation in the pathology.
7.5.4 Implications Several authors have recently underiined the implications of an invalid diagnosis of varicose veins for epidemiological investigations (Abenhaim et al.. 1999. Callam. 1994: De Backer. 1998: Evans et al., 1994) and assessrnent of the cost of treatment for the health systems (Bradbury et al.. 1999). Despite its limitations. this study contributes to the discussion by providing evidence that the diagnosis is influenced by postulated risk factors for varicose veins. The major implication of this finding appears to be the fact that GP-based data should be used with caution for etiological research in the field of varicose veins. It is inherently applied to the investigation of risk indicators presented in the follawing chapter.
This study is the first one to document the problem of the validity of the diagnosis of varicose veins in general practitioners' clinical practice. The diagnostic problem in the field of venous disorders is, however, not restricted to varicose veins. Blankfield et al. (1998) investigated the validity of a clinical diagnosis of bilateral leg oedema made by family physicians in 55 ambulatory adult patients.
Historical information, physical
findings and clinical impressions were cornpared to iaboratory evaluations including echocardiograms, duplex scan, serum albumin levels and 24-hour urine analyses. Venous insufficiency was the initial clinical impression in 71% of patients. but only 22% of subjects actually had this condition. Pulmonary hypertension was the cornmonest medical condition involved. In other fields of medicine, a lack of diagnostic specificity was found in the diagnosis of depression in the primary care setting, with differential misclassification dependent on knowledge of farnily history of depression (Klinkman et al.. 1998). and in the diagnosis and management of dyspepsia in primary care on the basis of symptoms and clinical history alone (Hansen et al.. 1998). In contrast. severity-of-asthma score was a valid measure in generalist-treated subjects on the basis of good interna1 consistency and concurrent validity in comparison to the SF-36 General Health and Physical Functioning scales (Eisner et al.. 1998). In another study. the agreement of clinical diagnosis with diagnostic criteria was studied for 5 chronic diseases (hypertension. chronic ischemic heart disease. diabetes rnellitus. chronic non-specific lung disease
and osteoarthritis) in 7 general practices in the Netherlands (Schellevis et al.. 1993). The agreement was high. ranging from 96% true positive cases in diabetes mellitus to 58% in chronic non-specific lung disease. The highest rate of false positive was 4%.
8. Analysis of risk indicators 8.1 Introduction In the literature review, it was concluded that the evidence regarding the relationship
between suspected determinants and the presence of varicose veins is adequate only for pregnancy and obesity, although the importance of the risk associated with these
two variables had not been adequately quantified. A relationship between other variables and varicose veins has been suggested. but conflicting results have been reported, and the association found with several variables disappeared after adjustment for age. Most authors of epidemiological reviews on venous disorders have concluded that varicose veins would have a multifactorial origin. but only one study using a multivariate analysis of several risk indicators was available at the time of initiation of the VEINES study (Brand et al.. 1988). Two other studies were published in 1995 (Scott et al.. 1995; Sisto et al.. 1995). Although these studies controlled for the
effect of a number of variables. neither made a distinction between subjects with varicose veins alone and subjects with varicose veins associated with more severe signs of venous disorders. In these studies, the associations identified could have therefore been confounded by determinants of more severe stages of venous disease.
Varicose veins are a chronic condition with slow development and III-defined onset times. In such a condition. a design based on incident cases would require the followup of a very large number of subjects free of varicose veins at the start of the
observation period. We know of only one study on risk factors of varicose veins based on incident cases: the population-based cohort of the Framingham Study, with a total of 44.382 person-years at risk (Brand et al.. 1988). Applied to the VEINES population. a
longitudinal design would have required that each of the 477 subjects free of varicose veins at time zero would have been followed for an extended observation period in order to obtain only a fraction of the amount of person-years availabie in the
0
Frarningham Study. This practical difficulty explains why a case-control design based
on prevalent cases is most often used for investigating factors associated with varicose veins. It was also adopted in this thesis. Given the limitations of this approach. it would not be possible to ascertain causal factors for varicose veins. The study was nevertheless designed to identify 'risk indicators', using Last's definition of a risk indicator: "An attribute that is associated with an increased probability of occurrence of a disease and that c m be used as an indicator of the increased risk. Not necessarily a
causal factor." (Last. 1995, p.149).
8.2 Objectives
8.2.1 General objective
To identify risk indicators for varicose veins.
8.2.2 Specific objectives
To rneasure the strength of association between the presence of varicose veins and variables suspected to be causally related to varicose veins. and thereby to contribute to supporting or refuting such causal links. The following variables were specifically examined :
number of pregnancies obesity defined as a body mass index (BMI) 2 27 blood group A history of varicose veins in mother history of superficial or deep thrombophlebitis past history of leg surgery
smoking status height
Oestrogen use and level of physical activity were initially included in the variables to be investigated. As explained in lhis chapter, constraints irnposed by the study design precluded a valid assessrnent of their association with varicose veins. - ,.
8.-3 Patients and methods '
a
Design : casetontrol study based on prevalent cases; the assumptions needed for the validity of such design are examined. Study population : members of the VEINES cohort included in the cross-section of the cohort at time zero, the 'cross-section' being represented by the subjects' visit to the investigators. The analysis has been restricted to women. Cases : subjects with varicose veins alone as reported by specialists: given the lack of specificity of the diagnosis of varicose veins made by general practitioners. patients with varicose veins ascertained by generalists were excluded. Controls : subjects without varicose veins or signs of more severe venous disease as ascertained by general practitioners and specialists; subjects with a history of sclerotherapy or vein surgery of the lower lirnbs were excluded.
a
Statistical analysis . multivariate logistic regression.
a
Sensitivity analyses : alternative definitions of the categories of some variables were used in specific models: a model included in the case group subjects with varicose
veins and either oedema. skin changes or ulcer. The use of a case-control design based on prevalent cases identified at one moment in time of the VEINES cohort had several implications (Rothman and Greenland. 1998. p. 113). They are examined below in relation to the following aspects :
use of prevalent cases for estimating an Incidence Density Ratio (IDR) (section
8.3.2) characteristics of the source population. restricted to subjects consulting a physician for a venous-related complaint (section 8.3.3)
r
definition of cases and controls and completeness of case/control ascertainment (sections 8.3.4 and 8.3.5).
8.3.2 Prevalent vs. incident cases
In the situation of a chronic disorder with slow development, the follow-up of individuals exposed to an etiological factor needs to be of long duration in order to include the risk period. The disorder is then described as having an extended risk period (Kleinbaum et al.. 1982, p. 112). In a cohort. the parameter of interest for estimating the risk of such a disorder is the incidence density and the parameter for etiological research is the incidence density ratio (/DR). The same parameter needs to be used when the estimation of the risk ratio is based on prevalent cases identified in a cross-section of the cohort. In the absence of migration out of the prevalence pool (steady-state assumption), the IDR has the following relation with prevalence in exposed (Pi) and non-exposed (Po) individuals:
where IDoand ID, are the incidence densities in non-exposed and exposed individuals respectively, and where To and Tt are the corresponding rnean durations of disease (Kleinbaum et al.. 1982. p. 147). It follows that :
where ?OR is the Prevalence Odds Ratio. Therefore. the POR is equal to the IDR only if Ti z TO: the determinant does not affect the duration of disease and the disease does
not affect the exposure status (Rothman and Greenland. 1998. p.113). Varicose veins having a very ill-defined time of onset. an additional requirement for the validity of the
prevalent odds ratio is the assumption that the tirne relationship between disease occurrence and suspected determinants is in the right direction, for example that a high
BMI was present before varicose veins occurred. and not the inverse. For several reasons, the following assumptions necessary for the validity of a POR are generally considered strong enough to make cross-sectional studies unattractive: except for stable deterrninants such as blood type, it cannot generally be assurned a
prion that the determinants of interest are a cause rather than a consequence of the outcome. unless accurate data on disease history and personal characteristics are available; in many chronic diseases, the exposure is also a prognostic factor and independence between exposure and duration of disease cannot be assumed: the steady state assumption cannot be assumed in narrow age categories (Kleinbaum et al.. 1982: p.148). These assumptions are discussed below in relation to venous disorders. Correct time sequence between exposure and disease. In the absence of information. confidence in the assumption of a correct time-relationship is often a matter of prior knowiedge and judgement. Suspected determinants for which there is a strong doubt about the sequence between exposure and disease should not normally be included in
the analysis. For example. finding a strong association between pregnancy and the presence of varicose veins may not necessarily mean that varicose veins are caused by pregnancy. However, it would be unlikely that the presence of varicose veins increase the probability of becoming pregnant. A strong association between pregnancy and varicose veins would therefore favour the hypothesis that pregnancy is a determinant of varicose veins. provided that potential confounding factors are
below. adjusted for. The discussion for the other determinants is held in section 8.3.9.
lndependence between suspected determinants and duration of disease. Venous disorders represent chronic disorders with a very iow lethality. and it is accepted that
they do not resolve with medication . The only way for individuals with varicose veins to exit the prevalence pool, excluding death. is by surgical extirpation or sclerotherapy. Misclassification in the case-control status would exist if subjects in the control group had their varicose veins previously sclerosed or removed by stripping, this history thereby conferring on them a possible case status. A nondifferential misclassification would result in a bias towards the nuIl value if the decision to operate was independent frorn one of the risk indicators. However, if the presence of a determinant of interest was a contra-indication for surgery or sclerotherapy (e.g. severe obesity ). resulting in more non exposed individuals than exposed ones being operated on or sclerosed, the prevalence of varicose veins in non exposed individuals would be deflated and the assumption of independence between determinants and duration of disease needed for the comparability of POR and IDR would not apply. In the example of BMI, the oddsratio would be biased away from the nuIl value. A differential misclassification may. however. go either way (Rothman and Greenland. 1998. p. 127). The actual situation is more complex. Although past history of venous surgery and sclerotherapy was recorded in the VEINES study, there is no information in the VEINES database regarding the indications for past venous surgery and sclerotherapy. Venous surgery may have been perfoned for a venous disorder other than varicose veins. and sclerotherapy is also frequently performed for telangiectasia in patients without varicose veins. It is therefore quite possible that the exclusion of controls with a past history of sclerotherapy or venous surgery would on its own introduce a bias if these procedures were related to one of the risk indicators of interest. such as age. obesity, socio-economic status or presence of a CO-morbidcondition. This possible problem is not as important for venous surgery, because the majority of procedures consisted of vein stripping. which is specifically related to varicose veins. Other venous surgery may be performed in patients suffering from more severe venous disease and these
patients would in any case be excluded from the control group. The following verifications were made in order to assess the impact of excluding
patients with past sclerotherapy and surgery :
estimation of the proportion of subjects with and without varicose veins who have a history of sclerotherapy. vein stripping or other venous surgery (Table 8.1); association between suspected determinants and history of sclerotherapy in patients without varicose veins (Table 8.2).
Past historv of sclerothera~vor surcierv Table 8.1 shows that about half of women already had a history of sclerotherapy or venous surgery at the time of the inclusion visit. whether they had varicose veins or not. Table 8.1 Numbers and proportions of women with and without varicose veins at the baseline visit who had experienced vein stripping, other venous surgery or sclerotherapy any time before the inclusion visit Women with varicose veins
n
O */
Women without varicose vein
n
%
AI/ women
777
Vein stripping
196
25.2
101
24.4
38
4.9
25
6.0
Scterotherapy
272
35.0
204
49.3
Any of the above
342
44.0
230
55.6
Other venous surgery
414
Past history of surgery was found in similar proportions in the two groups of women. but sclerotherapy was found more frequently in women without varicose veins.
This
difference was accounted for by patients recruited by specialists : in the specialistbased sample,
past sclerotherapy was reported in respectively 61% and 72% of
women with and without varicose veins; in the generalist-based sample. these percentages were about 23% in both groups. The high percentages found in the sample of women consulting specialists is perhaps accounted for by sclerotherapy performed for telangiectasia. but there are no data in this study to confirm this hypothesis. The exclusion of women with past venous surgery from the control group was not a
matter of choice.
In contrast. the exclusion of a large group of women with past
sclerotherapy may exclude valid controls It was therefore important to examine the
implications of such exclusions in relation to the selection of subjects with specific characteristics. Since these characteristics may represent levels of the determinants of interest, the exclusion of subjects with past sclerotherapy would result in a selection bias. The association between these deteminants and the use of past sclerotherapy in subjects without varicose veins at the baseline examination is shown in Table 8.2. Only determinants for which there is a strong suspicion of an association with varicose veins are presented in the table. Table 8.2 Association between suspected determinants and numbers and proportions of subjects without varicose veins at the inclusion visit (women who had a positive history of sclerotherapy any time before the inclusion visit only) No sclerotherapy
n
OO /
Sclerotherapy
n
O h
Al/ subjects Age (yeas) - 18-39 40-59 260 Body mass index < 25 25926.9 1 27 Number of pregnancies O 1 -22 missing
-
-
Mother hashad varicose veins
- no - yes
Previous thrornbophlebitis
- no
- yes Of al1 variables listed in Table 8.2. only one (number of pregnancies) was not
statisticaily associated with a history of sclerotherapy (p=0.12 at the chi-square test). The trend goes towards a high probability of previous sclerotherapy with age iess than 60 years. low body mass index. materna1 history of varicose veins. and history of
previous thrornbophlebitis. A high level of education and the absence of a CO-morbid condition were other variables associated with a higher likelihood of past sclerotherapy.
These data suggest that the exclusion of patients with previous sclerotherapy from the control group could induce a selection bias if subjects with sclerotherapy were actually valid members of the control group. It could. however, be argued in the opposite direction : the association found between suspected determinants of varicose veins and sclerotherapy suggests that the patients actually had varicose veins. Looking at these
aspects of the argument, the decision was taken to exclude subjects with past sclerotherapy as it was considered the safest attitude in terms of bias reduction. An important lesson from this discussion is nevertheless the fact that disease duration may be influenced by some determinants of interest and, therefore. that one of the main assumptions for interchangeability of the prevalence odds ratio and the incidence odds ratio would not hold if, for example. previous surgery or sclerotherapy were not reported.
Steady state assumption in narrow age categones. Table 8.2 shows that the steady state assumption cannot be assumed even in broad age categories on the basis of differences in the proportions of subjects having used sclerotherapy. Age differences also exist in proportions of patients operated on : the proportions of subjects aged 1839 years old. 40-59 years old and
60 years old who were operated on in the past are
4.9%. 5.4% and 2.5% in 1054 subjects with varicose veins and 2.1%. 5.6% and 0.0% in
477 subjects without varicose veins. Although the differences are not large and based on small numbers. they indicate that age is a variable that has to be controlled for in the analysis of risk indicators for varicose veins.
8.3.3 Choice of source population The source population is 'a source of learning about an abstract occurrence relation' (Miettinen. 1985: p.46). When interpreting the results of the analysis. it would be uninteresting if the conclusion only applied to the source population. defined in Chapter 4 as patients 18-75 years of age suffenng from a chronic venous-related leg disorder,
@
and consulting a general practifioner or a speciaiist for that disorder in the Frenchspeaking part of Belgiurn, in France, in italy or in Quebec.
Scientifically, it would be more useful to conclude about an association independently of the characteristics of the source population in which it was studied. Since the choice
of the source population rnay still influence the parameter estimated in the study. the extrapolation of results may require a specification of sorne characteristics of the individuals surveyed or of some study conditions. Before undertaking the analysis. it was therefore important to examine to what extent the use of the VEINES cohort could influence the conclusions to be drawn frorn the study.
Two aspects must be
considered:
In this study, given the choice of the source population, individuals without varicose veins are not individuals free of venous disorders. Since the source population is a rnix of affected and non-affected individuals, the association found in this study
between a risk indicator and varicose veins would be biased towards unity if the risk indicator was also associated with the conditions present in the control group. i.e. telangiectasia and symptoms. When extrapolating the conclusions of the study. the importance of this bias shouid not be overestimated. Recently, the group of the population-based Edinburgh Veins Study reported a prevalence of more than 80% of individuals having telangiectasia in a Scottish population (Evans et al..1998), and a proportion of more than 30 % of subjects having typical symptoms of venous disorders (Bradbury et al.. 1999).
Membership of the source population was influenced by factors related to the probability that subjects with varicose veins corne to the attention of investigators
and are subsequently included into the VEINES cohort. This situation can lead to selection bias; they must be considered as potential confounding factors and treated as such (Rothman and Greenland. 1998, p.?38). For example, it is possible that individuals with varicose veins consult more frequently if they are obese than if they have a normai weight for height. This selection precludes a dissociation of two effects of obesity : its effect on the occurrence of varicose veins and its effect on the probability of consulting a physician and being included in the VEINES cohort.
Consultations are also influenced by many other factors such as age. gender, socioeconomic factors, employment status, CO-rnorbidity. and country-specific medical culture. Following Rothman and Greenland (1998). adjusting for these factors in the analysis would (at least partly) account for the differing probabilities of coming to the attention of the investigator. However. since sorne of these variables are unknown. the inferential statement should not overlook that the source population is a population of patients consulting a family or a specialist practice.
8.3.4 Cases
In a case-control study, a methodological problem is an incomplete case identification with the degree of incornpleteness dependent on the deteninant (Miettinen. 1985. p.78). This is an example of differential misclassification that would affect the validity of
the results and for which there is no remedy. A loose or inaccurate case definition may also lead to non differential misclassification if the degree of misclassification is independent from the exposure status. It generally reduces the strength of an exposure-disease association.
In this study. cases are subjects with varicose veins. A common problem in the ascertainment of varicose veins is the lack of objective measurement. In some studies (Widmer et al.. 1978: Evans et al.. 1998: Laurikka. 1994: McAdam. 1976) photographs were used. but this method requires standardised conditions of work including ideal lighting conditions, specific lenses and accurate view angles. These conditions are very difficult to apply in practice and it did not seem possible to obtain from a large number of general practitioners photographs in standard technical conditions. The diagnosis of varicose veins was therefore defined on the basis of a clinical examination. The results of the validity study on the diagnosis of varicose veins by general
practitioners have been presented in Chapter 7 (Tables 7.1. 7.6. 7.7 and 7.8). The low specificity (45%) was associated with a large number of false positive cases. Moreover. specificity decreased with the number of pregnancies. history of past
@
thrornbophlebitis (as recorded by the general praditioner). histary of k g trauma. high
educational level and presence of skin changes. According to this. the misclassification of cases would be differential. with a bias away from unity because false positive cases are subjects presenting risk indicators of varicose veins.
In order to avoid these
problems of misclassification. the choice was made to restrict the case series to the set of patients with varicose veins diagnosed by specialists. Arnong the 242 cases diagnosed by specialists, 106 (43.8%) had varicose veins alone. 32 (13.2%) also had oedema without skin changes or ulcer. 80 (33.1%) also had skin
changes without ulcer. 19 (7.9%) also had an healed ulcer and 5 (2.1%) also had an open ulcer. Therefore, 136 of the 242 cases (56.2%) had varicose veins with signs that have a highest degree of severity on the s a l e of venous disorders according to the
CEAP classification (Porter et al., 1995). In these 136 subjects, variables associated with the probability of being a member of the case group may include two sets of variables : variables
specifically associated with the probability of varicose veins
occurrence. and variables associated with the probability of developing more severe venous disorders. In order to assess the possible effect of the first set of determinants independently from the second. the case group was restricted to subjects with varicose veins alone. More severe signs will be included in a sensitivity analysis.
In summary. the definition of the case group included the two following conditions : 1. The presence of varicose veins on either leg was recorded by a specialist investigator at the baseline visit. 2. Oedema. skin changes or healedlopen ulcer were absent on both legs.
8.3.5 Controls
In a case-control analysis. controls are representatives of the source population at a particular time if "were helshe to become a case at that time, he/she would (...) become a member of the case series of the study" (Miettinen. 1985.p. 54: Wacholder et al.. 1992a). thereby providing an estimate of the distribution of the categories of the
determinants in the source population (Miettinen. 1985.p. 338).
This representativity of the control group is a major issue in every case-control study. The problem is particularly important in situations where controls are defined secondarily to cases, because a conceptual definition of the source population from which cases arise is required.
In this study, cases and referents are members of the
VEINES cohort and are identified secondarily to the cohort definition.
Therefore.
controls are valid representatives of the source population unless some restrictions or conditions imposed to the definition or ascertainment of cases imply de facto that the actual source poprllation has been re-defined. following question must be raised :
Does the
In the context of this study.
restriction
the
of the case series to cases
diagnosed by specialists have implications in terms of re-definition of the source population ? In other words. should the control group be limited to subjects recruited by
specialists ? This question is relevant since the validity issues invoked to reject cases diagnosed by general practitioners apply also to controls. Subjects consulting specialists originate from two sub-populations : a population of
.
individuals consulting specialists directly and a population of subjects referred by general practitioners. As a matter of principle. controls should therefore also arise from these Wo populations. A restriction of the control group to controls detected by specialists would have been justified if GPs were also referring patients with syrnptoms or telangiectasia alone to specialists. This is not the case. as shown below in Chapter 12 (Utilisation of health services). Subjects consulting specialists for venous disorders
constitute a homogeneous group in terms of age. number of pregnancies and other important variables: the restriction of controls to such subjects would also entail a bias towards the nuIl in the risk ratio estimate since the same variables are associated to consultations for telangiectasia and to consultations for varicose veins. A remaining issue is the question of whether subjects with signs of more severe venous
disorders than varicose veins (oedema. skin changes. ulcer) should be included in the control group. The answer is negative. because the occurrence of these signs is probably associated with the same variables as the occurrence of varicose veins. The
prevalence of risk indicators in such subjects would therefore be higher than the prevalence in the source population.
In summary, the definition of the control group included the three following conditions
1. The absence of varicose veins on either leg as recorded by one of the investigators
at the baseline visit. 2. The investigators' reports that sclerotherapy, vein stripping or other venous surgery had not been performed before or at the time of the baseline visit. 3. Oedema, skin changes or healedlopen ulcer were absent on both legs.
8.3.6 Sample size
Of the 1531 subjects of the study population. 367 (24.0%) had varicose veins alone. Of these. 277 (75.5%) were women; 106 (38.3%) of these were diagnosed by specialists and represented the cases. A total of 477 subjects (31.2%) did not have varicose veins on either leg, and 414
(86.8%) of these were women. Among these 414 women. 184 (44.4%) did not have a history of sclerotherapy, vein stripping or other venous surgery: 94 (51.1%) of these did not have oederna. skin changes or ulcer on either leg and represented the
controls. Among the 94 controis. 67 were recruited by general practitioners and 27 were recruited by specialists.
8.3.7 Statistical power
It was estimated that the minimum relative risk estimate to be detected with statistical significance and a type I error of 5% in this study was 2.0. Table 8.3 therefore presents the power of the study to detect a odds ratio of 2 for a type I error of 5%. a sample stze of 100 in the case and control groups. and a prevalence of exposure in the control
e
group ranging from 10% to 70% (Schlesselrnan. 1982. p. 149).
Table 8.3 Power of the study to detect an odds ratio of 2 (a = 0.05) with a sample size of 100 in the case and control groups r
Prevalence of exposure in controls
Power
Table 8.3 shows that. irrespective of missing values, the statistical power is not adequate to detect an odds ratio (OR) of 2.0. The power in the range of exposure of 20% to 60% is equal to or higher than 80% only if the OR to be detected is 2.5 or higher. Alternatively, about 150 subjects would have been needed to detect an OR of The anticipated consequence of the sub-optimal statistical power is an imprecise estimation of the relative risk estimates. The sample size reduction that resulted from
exclusion criteria applied at different stages of the identification of study groups was motivated by a concern for unbiasedness. but this objective was achieved at the cost of an important decrease in the precision of the estimates.
8.3.8 Suspected deteminants
A number of variables were assessed as exposure variables. In view of the use of prevalent cases as a basis for a case-control study. the relevance of these variables must be discussed in terms of their time relationship with the disease as well as of the likelihood of recall bias. This aspect is discussed below.
.
a. Specifications
age: 3 categories: 18-39.40-59, 60-75 years
a
height : 3 categories: 4 60 cm, 160-164 cm. 2165 cm
O
body mass index : categories of body mass index were defined according to
Garrow's classification as modified by Seidell (1986) as follows : e25.0. 25.0-26.9,
27.0-29.9, 230.0 kglm2. In this analysis, the threshold for obesity was set at 27 kg/m2.and 3 categories were used : : ~25.0. 25.0-26.9, S 7 . 0 kg/m2 blood group : 2 categories : blood groups O. B or AB. and blood group A. a
smoking status: 2 categories: never. ever; number of pregnancies : 4 categories :O,1.2, >2:
a
history of leg trauma or non-venous related leg intervention before onset date of venous disease : 2 categories: yes/no
a
past history of superficial or deep thrombophlebitis : 2 categories : yeslno
a
family history of varicose veins, tested with several variables (2 categories in each) :
- history of varicose veins in mother (as reported by subject) : yeslno - history of varicose veins in father (as reported by subject) : yeslno
- history of varicose veins in brother or sister (as reported by subject) : yesho - history of varicose veins in any family relative (as reported by subject) : yeslno These variables were assessed in the self-administered questionnaire both as closed and open-ended questions.
In the statistical analysis. missing values were
interpreted as an absence of varicose veins. 'History of varicose veins in mother' was used as an indicator of family history in the main model. because it is assumed
to be the most valid variable for risk estimation. The other variables were introduced in specific models.
In specific models. alternative definitions were also tested for some variables: parity :
number of pregnancies : 0-1. 1-2. 22 (as in the Framingham Study) number of pregnancies : 0.21
obesity : alternative categories of body mass index were l6:OO hours;
cunent employrnent status : 2 categories: no paid employment, paid employment: education : educational level was measured using 6 categories corresponding to completed school levels; the subject was assigned the value corresponding to the highest school level completed; the question was designed in such a way that any same value described comparable educational levels in the different countries: for this analysis, three categories were defined : low (no education or primary school level). intermediate (professional studies or lower secondary school level), and high (higher secondary school level or university level). At the analysis phase. the 'low' and 'intermediate' categories were rnerged;
incorne : total net household incomes were assessed by a variable with 6 categories: each category was defined by a range of income values determined in each country in order to represent a comparable economic status from very low to very high: for example. the ranges used in France were : < 3.000 FF: 3.0004.999
FF: 5.000-9.999 FF: 10,000-14,999 FF: 15.000-24.999 FF: 225,000 FF; for this analysis. 3 broad income categories of incomes were defined : low (in France :
c5,000 FF); intermediate (5,000-9.999 FF) and high (210.000 FF). In the analysis.
O
the 'low' and 'intermediate' categories were merged.
country : 4 categories : Belgium. France, ltaly and Quebec. Due to the small number of subjects without varicose veins in Quebec. 'ltaly' and 'Quebec' were merged in the analysis.
8.3.1 0 Statistical analysis
Logistic regression was used to estirnate the strength of association between a suspected determinant and the presence of varicose veins. adjusting for the effect of other determinants and potential confounding variables. The rnultivariate model was fitted as follows. In the firrt step. al1 exposure variables and variables suspected to be confounders were examined in univariate analysis with the purpose of identifying appropriate categories for the use of indicator variables; for confounders. identification was guided by frequencies of representation in each category, such as for educational level or categories of revenue. In the second step. al1 variables were tested in a bivariate analysis as binary variables
with levels taking the values O (absent) or 1 (present). lndicator variables were created where a variable contained more than two categories. In the third step. a multivariate model ('Main model') was fitted.
The variables
introduced into the multivariate model were selected on the basis of one of the two following criteria:
i-
a variable is suspected of being a determinant of the presence of varicose veins.
and a measurement of its effect is identified as one of the specific objectives of the analysis: al1 variables that met this criterion were forced into the rnultivariate model
irrespective of the strength of their association with varicose veins in bivariate analysis; ii-
a variable is suspected of being a confounder of the association between the presence of varicose veins and one of the determinant of interest: it is considered as a potential coafounder and retained in the multivariate model if the p-value. corresponding to the chi-square calculated as the difference of the -2 log likekihoods computed in the models with and without that variable. is 0.10 or less.
In the fourth step. specific models corresponding to each of the variables identified in the objectives were fitted. For these specific rnodels. al1 other variables included in the main model were introduced as potential confounding factors Specific models were also used when the number of missing values for one of the variables affected significantly the power of the model. This principle applied for the effect of blood group. for which there was a large number of missing values. Several specific models were used with family history of varicose veins as determinant of interest : presence of varicose veins in mother. presence of varicose veins in mother
measured only in women who agreed with the investigator on the presence of varicose veins on their own legs. presence of varicose veins in father. presence of varicose veins in brother or sister. presence of varicose veins in any relative.
In the fifth step. a sensitivity analysis was performed by including in the case group subjects with varicose veins and oedema, skin changes or ulcer. For this sensitivity analysis. al1 suspected confounders were re-tested in bivariate analysis to identify a potential confounding effect.
All statistical analyses were performed with SAS 6.10.
8.4 Results 8.4.1 Unadjusted analyses (
Results of the bivariate logistic regression models for determinants are presented in Table 8.4, along with numbers and proportions of cases and controls in each level of a variable. Among the variables listed in Table 8.4, age. height, number of pregnancies. history of thrombophlebitis and family history of varicose veins were associated with an increased risk of presenting with varicose veins at the baseline examination.
Unexpectedly,
history of leg trauma and smoking were associated with a decreased risk of varicose veins. The 95% confidence intewals around the odds-ratio exclude the value of 1.O0in only a limited number of variables : age. number of pregnancies, history of leg trauma and variables related to family history of varicose veins. For only 12 persons was there a report of leg intervention or trauma before the year of varicose vein occurrence as stated by the patient. Based on this low number, the multivariate mode1 was fitted with and without that variable. Since its removal did not affect the odds ratio estimate of the other variables in the model. history of leg intemention or trauma before varicose vein occurrence was not retained in order to gain in precision. Body mass index was also redefined as c25 and 225.
Among variables tested for a potential confounding effect of the association between suspected determinants and the presence of varicose veins. education. rnonth of consultation. time of day at consultation and country fulfilled the criterion for inclusion into the rnodel.
A consultation held at the end of the day (>16:00 hours) and
educational level were inversely associated with varicose veins.
154 Table 8.4 Unadjusted logistic regression models with suspected deteminants Distribution of determinant in:
A_a_e (yeam 18-39 40-59 2 60 H e i ~ h(cm) t 460 160-164 r 165 unknown Body m a s index ~25.0 25.0-26.9 227 unknown Smokinci status never ever unknown Blood qroup O.0 or AB A unknown Number of oremancies O 1 2 >2 unknown Historv of leu trauma or non-venous surqery no Yes Supenicial or deep thrombo~hlebitis
no Yes
controls (n=94) n Oh
cases (n=106) % n
47 50.0 34 36.2 13 13.8
35 55 16
30 32.3 34 36.5 29 31.2 1
25 25.0 34 34.0 41 41 .O 6
65 69.9 11 11.8 17 18.3 f
73 73.7 11 11.1 15 15.2
46 65.2 36 34.8 12
58 65.2 31 34.8 17
54.8
35 48.0 38 52.0 33
40
33 45.2 21 25 13 21 21 14
31.2 16.2 26.3 26.3
33.0 51.9 15.1
7
11 22 31 20 22
13.1 26.2 36.9 23.8
84 89.4 10 10.6
104 98.1 2 1.9
87 92.5 7 7.5
89 84.0 17 16.0
unadjusted odds-ratio
Table 8.4 (continued) Unadjusted logistic regression models with suspecteci determinants Distribution of determinant in: cases
@=IO6) n % Varicose veins in mother no Yes Varicose veins in mother ' no Yes excluded Varicose veins in father
46
unadjusted odds-ratio
43.4
60 56.6
42 42.0 58 58.0 4
80 75.5 26 24.5 Vancose veins in brother 68 38
Varicose veins in any relative no Yes
19 87
64.1 35.9
17.9 82.1
Exclusion of women who disagreed with the investigator regarding the presence of
varicose veins on their own legs.
Table 8.5 Unadjusted logistic regression models with individual potential confounders Distribution of variable in:
1
Current emolovment non-ernployed employed unknown Educational level lowlintermediate high unknown lncome level Iow/intermediate high unknown
Month of consultation January-March April-June July-September October-December Time of day at consultation c 1200 hours 1200-1600 hours >1600h0ut-s unknown Country Belgium France ltaly Quebec
controls (n=94) n O h
cases (n=106 ) n %
unadjusted odds-rat io
pvalue (-2 log likelihood chi-square)
46 54.8 38 45.2 10
59 67.8 28 32.2 19
1.O0 0.57 (0.31 1.07)
< 0.10
26 54.2 22 45.8 46
21 42.0 29 58.0 56
1.O0 1.63 (0.73 - 3.63)
0.1 1
24 16 38 16
25.6 17.0 40.4 17.0
13 12.3 48 45.3 38 35.8 7 6.6
1.O0 5.54 (2.30 13.36) 1.85 (0.82 4.15) 0.81 (0.27 2.46)
28 34 28
31.1 37.8 31.1
42 40.4 46 44.2 16 15.4 2
1.O0 O.gO(0.47-1.73) 0.38 (0.18 0.83)
4
-
-
-
< 0.10
< 0.10
8.4.2 Muhivariate analyses
A. Main model Results of the goodness of fit of the 'main' model are presented in Table 8.6. Due to missing values accounting for a lack of information on number of pregnancies. educational level and smoking status, the initial sample size of 106 cases and 94 controls was reduced to 76 cases and 69 controls.
It results in imprecise OR
estimates. with a relative risk of 2.0 detected with a power ranging from only 43% to 53% for an exposure prevalence of 20% to 40%.
This lack of precision is obvious when estimating the association between age and varicose veins. In comparison to the 18-39 age group. the likeiihood of presenting with varicose veins was increased by a factor of 2.1 from 40 to 59 years and of 3.7 from 60 to 75 years after adjusting for the effect of other suspected determinants. All 95% confidence intervals (CI) include the value of 1.00. Other variables which showed a strong, although imprecise, association with varicose veins are the number of pregnancies and a history of varicose veins in the mother. The relative risk of presenting varicose veins was two to three times higher when the woman has been pregnant at least once in comparison to nulligravid wornen. and the risk does not increase with the number of pregnancies. Therefore. the increased risk of occurrence of varicose veins seems to be related to the first pregnancy. History of varicose veins in the mother seemed to be associated with a risk that is more than doubled.
-
Table 8.6 Multivariate analysis of factors associated with varieose veins Main model in women aged 18-75 years (sample size in model : 76 cases, 69 controls) Distribution (Oh) of deteminants in Deteminant
cases
controls
Adjusted odds-ratio '
Aae (yean) 18-39 40-59 60-75 Heiqht (cm) 460 160-164 2165 Bodv mass index c25 225 Number of preanancies
Past thrombo~hlebitis
Varicose veins in rnother no Y=
Smokina status never ever
Model adjusted for the effects of educational level, month of year at consultation. time of day at consultation and co~ntry.
B Specific models
Parity The main model indicated that women who have experienced at least one pregnancy
have a 2 to 3 times higher risk of presenting varicose veins than nulliparous women (Table 8.7). Two observations are noteworthy : 1) the risk of varicose veins associated with parity was not affected by the adjustment made for the effect of other variables: 2) although confidence intervals cross each other. the risk was lower when the woman has had more than 2 pregnancies than when she has had only 1 or 2. There is no clear-cut explanation for the decreasing risk. except for chance or a possible reverse 'selection' effect : women with varicose veins may avoid pregnancy due to leg pain or cosmetic problems already experienced during the first pregnancies. Table 8.7 Association between nurnber of pregnancies and varicose veins (models denved from main model in women aged 18-75 years) Distribution (Oh) of variables in controls
(n=94)
cases (n= 106)
unadjusted odds-ratio
adjusted odds-ratio
-t
Number of preqnancies '
O 1 2 >2 unknown
Number of premancies
O z 1
Number of oremancies
'
O-1 2-3
>3 unknown
O
-
3
OR adjusted for the effects of age. height. body m a s index. past thrornbophlebitis. history of varicose veins in mother, smoking status, educational level, month at consultation, time of day at consultation, and country; sample size reduced to 76 cases and 69 controls From main model Categories defined as in published results of the Framingham Study (Brand et al.. 1988)
When a category of women with at least one pregnancy was created. the odds-ratio was 2.96, with a 95% CI of 1.02 to 8.64. When pregnancies were categorised as 0-1. 2-3 and >3, as reported in results derived from the Framingham cohort, the odds-ratios
(OR) were 1.59 (95% CI: 0.66- 3.85) in women having experienced 2 to 3 pregnancies as compared to O or 1, and 1.14 (0.24-5.41) in women with >3 pregnancies (Table 8.7). These results are of the same order of magnitude as those published by Brand et al. (1988),with incidence density ratios of respectively 1.17 (95% CI: 0.98-1.41) and 1.30 (1.06-1.60). It suggests that the small increase in the risk of occurrence of varicose
veins with pregnancy reported in the Framingham study is perhaps related to an unsatisfactory definition of pregnancy categories.
Body mass index
The main model shows that there was no increased risk of presenting varicose veins with an increasing body mass index (BMI) in bivariate or in multivariate analyses. In contrast with expectation, the risk was decreased in individuals with a BMI 225 (OR : 0.66) (Table 8.6). The adjusted OR of presenting varicose veins in obese women with
a BMI 27 compared to women with a BMI < 27 was 0.70 (95% CI: 0.23-2.15). Blood group A The addition to the main model of variables indicating whether patients had blood group A or blood group O. B or AB showed a similar risk in the two groups (Table 8.8).
.
Table 8.8 Association between blood group A and varicose veins (model derived from main model in women aged 18-75 years old) Distribution of variables in controls (n=94)
cases (n= 106)
unadjusted odds-ratio
adjusted odds-ratio
Blood aroup 0.8 or A 8 A unknown
OR adjusted for the effects of age. height. body mass index. past thrombophlebitis. history of varicose veins in mother, smoking status, educational tevel, month at consultation, time of day at consultation. and country; sample size reduced to 76 cases and 69 controls
Family history of varicose veins As explained in the Methods, history of maternal varicose veins was taken as the first choice variable for indicating famiiy history of varicose veins. Several other variables were nevertheiess tested (Table 8.9). After the exclusion of 21 women who disagreed with the investigator for the presence of varicose veins on their own legs. an adjusted
OR of 3.63 (95% CI: 1.35-9.78) was found. which is higher than in the adjusted model fitted with these women included. There may be two explanations for the increase : a more valid ascertainment of maternal history of varicose veins. or a bias away from unity given the fact that women who disagreed and were excluded were more likely to be Young, highly educated patients seen by specialists in Quebec and with a lower
prevalence of varicose veins (about 18% of the initial control group and 4% of the initial case group were excluded). Our data do not aliow discrimination between these two explanations. The results obtained for other variables tested confirmed the higher risk of a family
history of varicose veins reported by affected subjects : 2.39 for a history of varicose veins in father. 4.15 for a history of varicose veins in brother or sister and 3.97 for a history of varicose veins in any relative.
162 Table 8.9 Association beîween family history of varicose veins and varicose veins (models derived from main model in women aged 18-75 years old) Distribution of variables in controls (n=94)
Varicose veins in mother no Yes Varicose veins in mother no Yexcluded Varicose veins in father no Yes Varicose veins in brother or sister no Y'= Varicose veins in any relative no Yes
'
-
'
cases (n=106)
n
%
56 38
59.6 40.4
46 43.4 60 56.6
51 26 17
66.2 33.8
42 42.0 58 58.0 4
81 13
86.2 13.8
80 75.5 26 24.5
80 14
85.1 14.9
68 38
39
41.5
55
58.5
19 87
n
unadjusted odds-ratio
adjusted odds-ratio
'
%
64.2 35.9
17.9 82.1
OR adjusted for the effeds of age, height. body mass index. past thrombophlebitis. pregnancy history. smoking status, educational levei. month at consultation, time of day at consultation, and country; sample size reduced to 76 cases and 69 controls From main model Exclusion of 21 women whose evaluation of W on their own legs disagreed with the investigator's evaluation
History of thrombophlebitis After adjustment for the number of pregnancies and other variables. history of past thrombophlebitis was poorly associated with the risk of varicose veins (OR: 0.77. 95%
CI: 0.21-2.88). The removal from the model of the variables used to designate the different numbers of pregnancy led to a moderate increase of the OR associated with thrombophlebitis (OR: 1.08. 95% CI: 0.32-3.65).
Past history of leg trauma or surgical intervention
Due to small numbers. past history of leg trauma and of non-venous surgical intervention were not used in the analysis. When this variable was added onto the other variables. an odds-ratio of 0.33(95% CI: 0.04-2.88) was found.
Smoking status The main madel (Table 8.6) showed that this variable was weakly associated with varicose veins in the bivariate and multivariate models (OR: 0.67: 95% 0.29-15 3 ) .
Height There was a small trend towards an increasing risk of varicose veins with increasing height (Table 8.6). A height greater than or equal to 165 cm was associated with a higher risk of presenting varicose veins in comparison with women smaller than 160 cm. but the 95% confidence interval includes unity. (OR: 1.77; 95% CI: 0.61-5.15).
C. Sensitivitv analvsis The 'enlarged' model is presented in Table 8.10. In comparison to the main model. the 'enlarged' model included 104 additional cases who had varicose veins and oedema. skin changes or ulcer. In univariate analysis. COmorbidity and employment status were identified as potential confounders and introduced into the model. The effect of this model extension was observed on several variables. mainiy age. especially the 60-75 years age group for which an odds ratio of 5.30 (95% CI: 1.5518.2) was seen, and history of thrornbophlebitis (OR: 2.08. 95% CI: 0.73-5.87).The
association between body mass index and varicose veins was not strengthened by the inclusion of severe venous disorders among cases. even after removing height from the model. The main result of this model is certainly the suggestion that past thrornbophlebitis is
an indicator of a higher risk of occurrence of severe signs of venous disorders rather than of occurrence of varicose veins.
Table 8.10 Multivariate analysis of factors associated with varicose veins, where individuals with oedema, skin changes or healedlopen ulcer combined with varicose veins are induded among cases. 'Enlarged' model in women aged 18-75 years (sample sire in model : 180 cases, 69 controls) Distribution (%) of determinants in Determinant
& (years)
18-39 40-59 60-75 Heiciht (cm) cl 60 160-164 2165 Body m a s index
cases
controls
21-1 51.1 27.8
52.2 36.2
Adjusted odds-ratio
'
11.6
21.7 35.0 43.3
37.7 29.0
83.9 16.1
87.0 13.0
12.2 18.3 34.4 35.0
33.3 14.5 26.1 26. t
72.8
27.2
91.3 8.7
49.4 50.6
59.4 40.6
63.9 36.1
47.8
33.3
Number of preanancies
Past thrombo~hlebitis
Varicose veins in mother
Smokinq status never ever
52.2
Model adjusted for the effects of educational levet, CO-morbidity,employrnent status. month at consultation. time of day at consultation and country
8.5 Discussion
The investigation of risk indicators for a chronic condition with an ill-defined onset time sets numerous problems in epidemiology, from collection of exposure data to valid ascertainment of events. An unbiased design would require a large population free of varicose veins to be followed for a long period of time. The Framingham study (Brand et al.. 1988) is exceptional in this respect, although information was not available on a large number of suspected determinants. This study took the forrn of a case-control design based on prevalent cases identified in a cross section of the source population. The analysis examines the relationship between varicose veins and variables of interest at one point in tirne, narnely the baseline inclusion visit to the study investigator. Limitations of this approach have been delineated in the Methods section. They include difficulty in ascertaining the temporality of events (i.e. distinguishing
cause and effect) and the potential for information bias due to lack of precise documentation and/or difftculty with recall for variables of interest that occurred in one group of patients more than in another group. These limitations are inherent to the design and their understanding helps to guide interpretation of the data. To ouf knowledge. this study on risk indicators is the first one that has investigated variables specifically associated with the occurrence of varicose veins alone. a design which allowed removal of the effects of determinants of more severe venous disorders. Before discussing the implications of these results. it is necessary to consider several methodological issues. classically divided into precision of estimates. selection bias. information bias. confounding effects and generalisation.
8.5.1 Methodological issues
Precision of estimates Cases and controls of this study were selected from members of a closed cohort derived from a source population of patients consulting general practitioners and specialists for venous disorders. The use of a 'prirnary base' as a source of cases and
controls has advantages in t e n s of comparability of the two groups (Wacholder et al..
l992b). It also has limitations. The use of a cross-section from a closed cohort limited the size of the case series and the number of available controls. As a consequence of this constraint, the analysis began with a fixed sample size. The main objective of the VEINES study was to examine the multiple consequences of venous disorders for the patients. For this purpose. the sarnpling procedure focused on ensuring an adequate representation of male patients and of patients with leg ulcer of venous origin. No atternpt was made to increase the representativity of patients with
varicose veins alone.
Moreover. two methodological issues were not anticipated
beforehand. The first is the differential misclassification of varicose veins by general practitioners according to suspected determinants, which led to the exclusion of al1 cases diagnosed by general practitioners. The second is the large number of subjects without varicose veins who had had sclerotherapy or venous surgery in the past. casting doubt on their 'control' status. Although sclerotherapy was probably not related to the presence of varicose veins in al1 cases. the absence of information on its indication led to the exclusion of al1 patients in whom sclerotherapy had been recorded. These exclusions were of great consequence as they reduced the sample size ta about 100 individuals in each group. In the final model. about one third of subjects were also excluded due to missing values for one or several variables.
Therefore. al1 95%
confidence intervals around parameter estimates are imprecise. The final sample size could not ensure an adequate statistical power for detecting a relative risk of 2 with a power of 80%: a sample size about twice as large would have been required for this purpose. The minimum relative risk that could be detected with type I error of 5%. a power of 80%. a prevalence of exposure of 30 to 50% and a sample size of about 70 cases and 70 controls ( final model) was 2.7.
Selection biases Selection biases are distortions that result from procedures used to select subjects and from factors that influence study participation. The common element of such biases is that the relationship between exposure and disease is different in those who participate
@
from those who should be theoretically eligible for study, including those who do not participate (Rothrnan and Greenland, 19%. p. 119). In rnost cases. selection biases cannot be resolved at the analysis stage. because, by definition. there are no data available on those subjects who were eligible but did not participate. An advantage of using a 'primary base' as a source of cases and controls is to match study subjects by variables that determine membership to the source population. In this study, however, a sampling took place in order to establish the VEINES cohort. The sampling was at random with the exception of male subjects and subjects with ulcer who were forced into the sample. In principle. this procedure should not affect the participation of subjects without varicose veins or with varicose veins alone. However. actual inclusion into the cohort depended on both the physician and the patient. Variables that influenced patients' inclusion are factors that might lead to selection bias and be considered as confounders if they acted differently for cases and controls (Rothman and Greenland. 1998. p. 138; Wacholder et al.. 1992b).
The only
information available in the registration database to assess these factors was age. gender and clinical status. It was assumed that other factors that could play a role were socio-economic status, CO-morbidity,employment status. month of consultation and time of day at consultation. and these variables were assessed as potential confounding factors. Selection bias may have been introduced by the exclusion of controls with previous sclerotherapy or venous surgery. In fact. sclerotherapy was associated with some of the variables of interest (body mass index. history of varicose veins in mothers and past thrombophlebitis), and this exclusion could lead to a selection bias if people with sclerotherapy were val id controls.
Unfortunately. this hypothesis is unverifiable
because the indication for sclerotherapy was not known. If it were present. a selection bias would act away from the nul1 for history of varicose veins in mothers and for past thrombophlebitis (since such a history was elicited more frequently in excluded subjects) and towards the nuIl value for body mass index (subjects excluded had a lower BMI than nonsxcluded subjects).
The restriction of the cases series to cases diagnosed by a specialist does not constitute a selection bias, since al1 cases fulfilling this criterion were included. Information biases Information bias c m occur whenever there are errors in the measurement of variables, but the consequences of the errors are different, depending on whether the distribution of errors for one variable (e.g., exposure or disease) is dependent on the actual value of other variables (Rothman and Greenland. 1998, p. 126). In order to appreciate the validity of the prevalence rate ratio, it was important to assess three types of measurement errors in this study : errors occurring on exposure variables. errors occurring on disease status. and differences in proportions of missing values.
Errors N, measurement of exposure. Two sources of information bias were presented in the Methods section : incorrect time sequence between exposure and disease, and recall bias. An incorrect time sequence can occur whenever the disease influences the exposure. As explained in the Methods section. such bias could theoretically occur for the association between varicose veins and body mass index. thrornbophlebitis. smoking status, physical activity and parity. Although it cannot be excluded for any of these variables, the likelihood of a reverse association has been considered important
enough only for level of physical activity. Body mass index (BMI): it is postulated that cuvent BMI is an indicator of BMI before varicose vein occurrence. namely that it was not significantly affected by the occurrence of varicose veins. Thrombophlebitis : it is postulated that 'past thrombophlebitis' refers to a period preceding varicose vein occurrence and. following Campbell (1996). that thrombophlebitis is an unlikely complication of varicose veins: however. a reverse association could exist and lead to either a direct (if varicose veins entail surgery and immobilisation) or indirect association (if varicose veins induce less risky activities and less trauma).
Smoking status : it is postulated that current smoking status is an indicator of smoking behaviour before varicose vein occurrence, namely that it was not significantly affected by varicose veins; this postdate is consistent with a rather high prevalence of about 34% of smoking in cases and in controls : if the figure of 34% in the case group was the result of a decrease of smoking after the occurrence of varicose veins. the prevalence of smoking before varicose vein occurrence would have been high. Physical activity : it was considered that current level of physical exercise cannot be considered as an indicator of physical exercise before varicose vein occurrence because it may be included in the changes to lifestyle promoted after the diagnosis of varicose veins; this variable was therefore removed from the list of variables of interest. Parity : it was considered unlikely that the presence of varicose veins could influence parity; it is however possible that the presence of varicose veins could influence the number of pregnancies by fear of aggravation of varicose veins during pregnancy or
by experience of painful symptorns during a previous pregnancy. Information on leg trauma and non-venous surgical intervention was also collected and much care was exercised at the analysis stage to select events that occurred before the onset of venous disease reported by the patient.
This process led to the
identification of a very srnall number of events (12), and this variable was not introduced into the multivariate rnodel. Since more than 80% of subjects tncluded in the VEINES study told that their disease had lasted for more than 5 years. it is possible
that a certain number of events were not recalled.
Errors on disease status. Differential misclassification in case-control status would exist if the probability of being (erroneously) classified as a case or a control depended on some of the variables of interest. Two opportunities for information bias were discovered while preparing the analysis of this study. Firstly, the validity study carried out with general practitioners showed that the specificity of the diagnosis made by
general practitioners was influenced by variables suspected to be determinants of the disease, such as body mass index, pregnancy history, history of thrombophlebitis and history of leg trauma. This bias justified the exclusion of cases diagnosed by general practitioners. Secondly, a large number of control subjects had had sclerotherapy or venous leg surgsry in the past, suggesting they could have had varicose veins and thus could not be included in the control series. Moreover. the likelihood of undergoing one of these procedures was influenced by variables of interest such as varicose veins in rnother or history of thrombophlebitis. Had past history of sclerotherapy and surgery not been recorded, differential misclassification would have remained unknown. Differences in proportions of missing values. In Table 8.6. some variables have a large number of missing values although the information should readily be available. This observation applies to the number of pregnancies. smoking status and educational level. This information was collected by phone call. Missing values are explained in part by the impossibility of reaching some subjects by phone despite several attempts. by refusals to provide information on the phone (despite the specification of the telephone interview on the informed consent questionnaire) and perhaps by a lengthy questionnaire that some interviewees interrupted. The number of missing values is larger for the number of pregnancies than for the h o other variables because the question regarding the number of pregnancies was not always answered. The proportions of missing values also differed by country. It was higher in France because, due ta an operational problem, some patients could not be interviewed by phone. For
example. in Belgium. ltaly and Quebec. the proportion of missing values for parity were 4.5%. 11.8% and 0% respectively. In France, this proportion was 32.2%.
For these three variables, the proportion of missing values was higher among cases than among controls : 16.0% and 12.8% respectively for smoking status. 20.8% and 14.9% for pregnancy history and 17.9% and 10.6% for educational level. For each
variable, the 95% confidence interval for the difference of proportions included the nuIl value. The unequal distribution of missing values in cases and controls is unexpected and its cause is not uncierstood, because interviewers were not aware of the
interviewees' clinical status. These missing values have a consequence : a bias would be introduced if missing values were unequally distributed among categaries of the variables involved. Since the distribution of missing values cannot be verified. the result is uncertainty regarding the parameter estimates. Confoundinq In this context. confounding refers to the situation in which a measure of the effect of an exposure on risk is distorted because of the association of exposure with other factor(s) that influence the outcome being studied (Last. 1995. p. 35).
In this study, each
exposure variable of interest is a confounder of the relation between another variable and the risk of varicose veins. Moreover, other variables may be associated with that risk as a result of the sarnpling procedure, the distribution of the characteristics of the source population or by chance only. These potential cmfounders become actual confounders if they are also associated with exposure variable(s) of interest. In this study. information has been collected on a large number of variables which were suspected to be related to the risk of occurrence and detection of varicose veins. Variables possibly associated with the actual inclusion of subjects with varicose veins were also treated as potential confounders. There are nevertheless at least three variables which were not evaluated for their confounding effect. although they have been suspected by some authors to be related to varicose veins : Oestrogen use : this includes the use of contraceptives and hormone replacement therapy. The information was collected in the questionnaire filled in by general practitioners.
In the questionnaire to specialists. as questions were added
concerning iaboratory tests. some other questions had to be left out: it was considered that oestrogen use could be assessed only in the general practitioner based sample. Therefore. the rernoval of cases diagnosed by GPs impeded an assessment of oestrogen use. In any case. the validity of this evaluation would have been questionable due to the difficulty of ascertaining the duration of use before the occurrence of varicose veins.
Raised intra-abdominal pressure, including the effect of constipation and tight undergarments. This information was not collected because it was considered too difficult to ascertain with sufficient validity for this analysis of risk indicators. Working conditions, and specially standing position at work.
Information was
collected regarding previous employment. However, less than half of women were currently or previously employed outside the home, and it was judged difficult to reliably assess the duration of standing position in housewives. It was also assumed (but not substantiated) that socio-economic status was a better proxy variable than self-reporting in estimating duration of standing. Generalisation The choice of the study population does not necessarily affect the validity of the results. but the important question is : Does the choice of the study population affect the generalisation of the results ? The answer to this question is often a matter of judgement (Miettienen. 1985, p. 47), but this judgement must be guided by a clear understanding of the study design. Once the prerequisite of validity has been met. one should assess whether characteristics of the study population are relevant to the interpretation of results, namely if they are effect modifiers. In this study, the study population was restricted to women. The question is whether the "abstract relation" inferred from a female sample is applicable to a male one.
Since
number of
pregnancies was introduced into the model (and is a confounder of the association between varicose veins and other variables), it can be assumed that the results apply
only to women.
One could postulate, however. that some results (e.g., history of
varicose veins in mothers) apply also to men. but this inference is a matter of judgement. The source population of this study was represented by subjects consulting physicians
for venous problems. We have already alluded to the fact that the inclusion of subjects with symptoms or telangiectasia in the control group probably had no effect on the interpretation of results. given the high prevalence of these conditions found in
a
population-based surveys of "normal" subjects. However. consultation with physicians
is the result of a self-selection process. An atternpt was made to account for some of the variables associated with that self-selection, such as socio-economic status. associated morbidity and age. but there is no certainty that al1 variables were measured. Since these variables might also be effect modifiers. the safest attitude is to take into account that it was a patient population that was the source of cases and controls. The restriction of the case series to cases diagnosed by specialists adds another constraint on the extrapolation of results. The consequences can be judged from differences in the characteristics of individuals consulting general practitioners and specialist, for examples age and body mass index (BMI). In this sample, the mean ages of women consulting generalists and specialists were 56 years and 49 years respectively; the mean values for BMI were 26.7 and 24.2 respectively. Wornen consulting specialists also had a higher level of education. The restriction imposed on the case series increased the homogeneity of the case and control groups. Although this homogeneity affects the efficiency of the analysis rather than its validity, it can also obscure an association that is limited to one group because the effect is weaker in other groups or because of differences in the distribution of the exposure (Wacholder et al.. 1WZb). Since it cannot be postulated that unmeasured variables influencing consultations to specialists have no modifying effect. the restriction of the cases series should be taken into account when generalising the results.
It should also be
considered that the source population results from a random selection of consecutive subjects consulting these specialists and that much care has been taken to measure and control possible selection biases.
The source population of the VEINES study comprised S
U ~ J ~ C consulting ~ S
physicians
in Belgium. France. ltaly and Quebec. Strictly speaking, the results apply to individuals coming from an 'average' of these countries. The main implication is the fact that results are not biased by specific characteristics of a local population.
8.5.2 Cornparison with other studies Many authors have concluded that the etiology of varicose veins is multifactorial. but. when this project started. there was only one study which adjusted for the effect of several variables sirnultaneously, the Framingham Study (Brand et al., 1988). Two other studies were published in 1995 (Scott et al.; Sisto et al.). In order to assess the relative merits of this study. a synopsis of the study designs of the three available studies is useful. The results were presented in Chapter 3.2.3. Brand et al.. 1988 [Framinaham Studv) The clinical status of 5209 men and women 30-62 years of age was determined at each biennial visit to the Framingham Study clinic. Varicose veins were defined according to the presence "of distended and tortuous veins, clearly visible in the lower limb with the subject standing". Abnormalities of venules were excluded. The case definition did not specify the inclusion or exclusion of patients with other signs of venous disorders. Follow-up began at the gthbiennial examination and continued for 16 years. Of the 1720 men and 2102 women who were free of varicose veins at the time of the initial examination, 366 men and 629 women developed varicose veins during the 16 years of follow-up. Two-year incidence rates of varicose veins by sex were calculated. Risk factors investigated included : systolic blood pressure. body mass. number of cigarettes smoked each day. total cholesterol. diabetes. physical activity index. number of pregnancies, and age at menopause. Statistical analysis used logistic regression models for grouped event times occurring in a survival data analysis. Scott et al.. 1995 Patients who attended at the Vascular Surgery or General Surgery Clinics at two Boston hospitals were examined and intewiewed. Cases were 129 patients with a diagnosis of varicose veins based on "physical appearance of the leg" (no definition was provided). The control group comprised 113 persons from the general surgery clinic who did not report any history of varicose vein or leg ulcer. Interviewers were blinded to the patient's diagnosis. A large number of variables were investigated.
Logistic regression was used in multivariate regression. retain,
Only four variables were
in the model comparing cases and controls : age, gender. history of
phlebitislclot, family history of varicose veins. Sisto et al., 1995 This study was based on the Mini Finland Health Survey. The study population of 8000 people was a stratified two-stage sampling cluster drawn from the national
population register. Al1 people received a questionnaire and were invited to participate in a health examination (7217 attended). The presence of varicose veins was based on the information supplied in the self-administered questionnaire. lt was assessed in two ways : i) by questions on whether a physician had ever made a diagnosis of varicose veins in the lower extremities: ii) by questions on history of varicose vein surgery. A large number of determinants were collected at the clinical examination. The prevalence of patients with varicose veins diagnosed by physicians was 6.8% in men and 24.6% in women. Logistic regression compared subjects with and without diagnosed varicose veins. Adjustment was made for al1 the following variables simultaneously : age group. place of residence. income. height. weight, subscapular skinfold. standing at work. osteoarthritis of the knee. diabetes. surgery on myoma (women only) and parity (women only). Stren~thsand weaknesses of this studv In comparison with the three studies described above. the major strength of this study is to have isolated two groups of subjects without signs of venous disorders other than telangiectasia or varicose veins. In the three studies. oedema and skin changes could be present in both groups. but the actual proportions of subjects with and without these signs is unknown. Therefore, the results of the three studies could be influenced by determinants of more severe venous disorders. In comparison to the case-control
studies carried out by Scott and by Sisto. the main advantages of this analysis are in the identification of cases based on physical examination, the care taken to explore the
temporality of events and the misclassification of exposure and disease status, and the large number of suspected deteminants that have been investigated and adjusted for. The main weakness of this study relates to its design. The use of prevalent cases in a case-control study c m lead to information bias and this design is inferior to the population-based cohort of Framingham. The two other case-control studies are not better in this respect. Another important problem is the insufficient power for detecting a relative risk of 2. Both the Framingham study and the Finnish study were superior in this aspect. The proportions of missing values for important variables such as parity and education were not negligible in this study and add uncertainty to the validity of the
estimate. A future limitation of this study is that it provides information for women only.
8.5.3 Discussion of results
The estimates found for variables kept in the main model and cited in the list of specific objectives are discussed hereafter. The discussion will focus on the point estimates and their 95% confidence intervals rather than on statistical significance.
Age The study shows an association between age and the prevalence of varicose veins. In cornparison to the 18-39 age group. the prevalence in women aged 40-59 years is approxirnately doubled and the risk in women aged 60 to 75 years is approximately four times higher. However. the 95% confidence interval around these values includes the nuIl value. A consistent relationship of increasing prevalence of varicose veins has been found with increasing age almost in al1 epidemiological studies. The prevalence odds ratio found in this study is consistent with prevalence figures reported in population-based surveys. for example those pubiished by Abramson et al. (1981) and Coon et al. (1973) (Figure 3.2).
The results of the Framingham Study (Brand et al.. 1988) suggest that age is not associated with an increase in incidence density of varicose veins (Figure 3.3). Age
differences are therefore best explained by an accumulation of new cases as people prow oldw (Evans et al., 1994). Number of oreanancies The association of varicose veins with the number of pregnancies is the strongest effect found in this study. In comparison to the nulliparous. women having experienced at least one pregnancy have a risk about three times higher (OR: 2.96. 95% CI: 1.028.64). Several observations are worthy of note :
1) The point estimates were similar in univariate and multivariate analyses. and the association was not affected by the adjustment of other variables; this observation suggests that the effect of pregnancies is independent of the effect of other variables and is consistent with the fact that comparable estimates were found in studies where results are adjusted for different numbers of variables (Table 3.1 1). 2) Our results show that the risk does not increase beyond one pregnancy. This conclusion has been reached previously by several authors (Beaglehole et al.. 1976: Brand et al.. 1988; Weddel. 1966; Widmer et al.. 1987). To our knowledge. the oddsratio increased with the number of pregnancies in only two studies. with threshold values of 2 (Cornu-Thénard et al.. 1989) and 4 (Sisto et al.. 1995) pregnancies. The reason why the risk of varicose veins would not increase beyond one pregnancy has not been clearly demonstrated. This pattern favours the hypothesis of pregnancy being
an exacerbating factor in those already predisposed. as suggested by several authors in the seventies (Guberan et al.. 1973; Burkitt. 1976: Phillips and Burkitt. 1976). In any case. it does not favour the hypothesis of an effect mediated via the occurrence of thrombophlebitis. The risk of thrombophlebitis in pregnant women was estimated at about 4 per 10.000 women (James et al.. 1996) and should not normally not be
influenced by pregnancy order. The hypothesis of the pregnant uterus obstructing venous return from the legs and favouring the occurrence of thrombophlebitis is examined in Chapter 9 in which patterns of venous reflux are examined.
@
Given the high validity of its design, the Framingham study is generally considered as the reference source of information regarding the epidemiology of cardio-vascular diseases. However, it cannot help resolve the question of whether the risk increases with the number of pregnancies (and particularly if the risk is determined by the first pregnancy) because women with O and 1 pregnancy were grouped together. The reason for this grouping is unclear. When the same grouping was used in this study. the odds ratios were comparable to relative risks calculated from the data reported by Brand et al. (1988) (1.6 vs. 1.2 for 2-3 pregnancies and 1.1 vs. 1.3 for > 3 pregnancies). The same calculation performed on results published by Cornu-Thénard et al. (1989). Sisto et al. (1995) and Weddel et al. (1978) also shows relative risk estimates between 1.7 and 1.8 when the reference category is 0-1 pregnancy. 3) The estimates found in this study are in general higher than those found in other
studies (Table 3.1 1). The explanation could lie in a selection bias : women with varicose veins might consult a specialist more frequently when they have a history of pregnancy. This explanation is not in agreement with the fact that large odds-ratios were also found in studies using population-based statistical samples. for example surveys of New-Zealanders of European descent (Beaglehole, 1976) and of inhabitants of the City of Cardiff (Weddel. 1969). It is also possible that the lower odds ratios found in other studies result from a bias towards the null. e.g. the non-exclusion of subjects with other venous disorders in the case and control groups. The possible etiology of pregnancy-related varicose vein occurrence is discussed in Chapter 9 after venous reflux in women with varicose veins has been examined. 4) In this study. the risk decreases with nurnber of pregnancies greater than 2.
Although this decrease can be explained by chance alone. it could also be due to a selection factor : women with varicose veins could have avoided further pregnancies after experiencing bad outcomes while pregnant. such as leg pain or heaviness.
@
Obesity (body mass index
27)
Divergent resulls have k e n published in the iiterature regarding the association between obesity and varicose veins. Several studies suggest that a higher risk exists in women only, with a small increased risk in obese women (Table 3.12).
In
Frarningham, an odds-ratio of 1.3 (95% CI: 1.1-1-6) was found. In a follow-up of 565 overweight women (BMI > 27 kglm2) and 627 control women over 6-17 years in the Netherlands, Seidell et al. (1986) found an incidence density ratio of 1.4 for the occurrence of varicose veins. In contrast, no increased risk was found in this study in bi- or multivariate analyses in overweight (BMI 25) or obese (BMI 2 27) women. The risk is below 1 in both cases and it is only moderately increased in the enlarged model (Table 8.10). Guberan et al. (1973) did not find any difference between the lightest and the heaviest women once age was adjusted for. but this conclusion was not confirrned by several studies published later where the differences in BMI were independent of
age (Table 3.12). In the majority of studies, the small relative risk estimates could be affected by misclassification or lack of control for confounding effects. In this study. the assessment of the effect of BMI is based on small numbers. but this limitation influences the efficiency of the analysis rather than its validity.
It is
noteworthy that history of sclerotherapy is a confounder of the association between body mass index and presence of varicose veins. as women without varicose veins on admission. yet who experienced varicose veins in the past, are lighter than those who did not.
Therefore. the exclusion of women with previous sclerotherapy from the
control group influenced the effect of BMI. (In an unplanned analysis, we fitted a model where control patients with past sclerotherapy were not excluded. The odds-ratio for BMI 2 25 was 1.4). In none of the studies was a physiologicat explanation of the relationship between excess weight and varicose veins proposed. Metabolic changes could impact on the distensibility of the vein wall. but varicose veins were not linked to total cholesterol in the Framingham Study. The relative risk estimate for the association between BMI and varicose veins is much lower than that found for atherosclerosis or other cardiavascular diseases (CVD). In the Dutch study (Seidell et al.. 1986). the incidence
density ratio in obese women was 5.1 for atherosclerosis and only 1.4 for varicose veins. This result militates againrt the same risk factors being concerned with varicose veins and CVD. Wood gr ou^ A In a casetontrol study based on patients consulting two specialists, Cornu-Thénard et al. (1989) investigated whether A B 0 genes could play a rote in the pathogenesis of varicose veins. Their study was prompted by the relationship that exists between blood groups of the A B 0 system and deep venous thrombosis of the lower limb. The main result was a crude odds ratio of 2.00 (95% CI : 1.36-2.95) for patients with blood group A in comparison to patients with blood group O, B or AB. The high risk of varicose veins persisted after individual adjustments for age (OR: 2-13), gender (OR: 1.96). number of pregnancies (1.98), paternal history of varicose veins (OR: 1.98). personal history of deep vein thrombosis (OR: 1.87) and materna1 history of varicose veins (OR: 1.86).
We are not aware of another study having performed the same investigation. In this study, we found an OR of 0.95 (95% CI: 0.37-2.43).There were a large number of missing values because physicians were not required to take blood samples to determine the blood group. However. it is unlikely that knowledge of blood group affected the diagnosis of varicose veins. The crude proportion of control patients with blood group A in this sample (45.2%) is identical to the proportion of group A (45%) in Arnericans of Western European descent (Giblet. 1991). The crude proportion in cases was 52%. According to Cornu-Thénard et al.. the distribution of A B 0 blood groups in their study was identical to that of the general population in France. as estimated by the "Centre National de Transfusion Sanguine". but it is not known whether this statement refers to the control group (33.9%). the case group (50.6%)or the mix of both. The odds ratio of 2.0 found by
Cornu-Thénard et al. could be due to the low prevalence of group A in their control
group.
Familv histow of varicose veins Despite the fact that most authors of epiderniological reviews have followed Weddelrs conclusion that information from a proband is invalid (Weddel. 1969). many studies have used this method to investigate family history of varicose veins. In section 3.1.4. of this thesis, it has been shown that Weddel's results are hadequate to refute or confirm the lack of validity of a family history obtained from a patient. All studies that have investigated family history of varicose veins have found a positive association. There is no doubt that recall bias can increase the relative risk estimates. Still. studies presented in Table 3.13 are of various designs and some have been carefully planned and conducted. We doubt that recall bias would have acted similarly in al1 of them if the association was not real. Of particular interest are the studies by
Cornu-Thénard et al. (1994) and Weddel et al. (1969) which are based on direct observations of relatives. Both found a relative risk of 2. In this study. we attempted to address recall bias by using the presence of varicose veins in mothers as the variable of interest in the main model. In an alternative model, women who had disagreed with the physician regarding the presence of varicose veins on their own legs were excluded. The presence of varicose veins in other family members was also investigated. All results show an odds-ratio of 2 or more for the relationship between family history and varicose veins. The demonstration of a positive association with family history is not sutficient to demonstrate a genetic predisposition. because the association could also be due to a common environmental factor. This hypothesis requires that common environmental factors act during childhood. An interesting study in this respect is the Bochum study (Schultz-Ehrenburget al., 1989): examination of a cohort of children aged 10-12 years demonstrated a four-fold increase in the prevalence of discrete reticular varices four years later.
According to Cleave (1959).no true hereditary defect occurs with a
frequency as high as the prevalence of varicose veins. Building on this argument. Burkitt (1976) and Geeihoed and Burkitt (1991) contended that the common environmental factor favouring the occurrence of 'idiopathic', primary varicose veins (as opposed to secondary varicose veins caused by deep vein thrombosis) was raised
intra-abdominal pressure caused by fiber-depleted diet. Aggravating factors such as pregnancy would explain increased frequency of varicose veins in some individualsAlthough there is no doubt that high prevalence figures are found in countries with dietary deficiencies, the association between the two remains ecological and has never been demonstrated. Historv of supetficial or d e e thrombo~hlebitis ~
On theoretical grounds, thrornbosis can be followed by valve destruction.
The
importance of this phenornenon is difficult to appreciate in epidemiological studies because more than two thirds of venous thrombi are clinically silent (Weinmann, 1994).
Moreover. recall bias is certain1y acting towards a more complete ascertainment of previous thrombophlebitis in subjects with varicose veins than without. An important finding of this study is the lack of association between past thrombophlebitis and varicose veins in subjects with varicose veins alone (OR: 0.76. 95% CI: 0.21-2.74), and an increased risk once subjects with signs of severe venous disorders are included
(OR : 2.08, 95% CI: 0.73-5.87). A first observation is k a t recall bias would probably act for both analyses if it could explain part of the association found in the enlarged rnodel (Table 8.10). Secondly. the odds-ratio is substantially higher in bivariate (OR : 2.37. 95% CI: 0.94-6.00) than in multivariate analysis.
A search for variables
associated with thrombophlebitis showed that the reduced effect of thrombophlebitis in multivariate analysis is explained by a small effect of nearly each variable present in the model. resulting in a marked reduction of the odds-ratio when al1 variables were introduced simultaneously. More irnportantly, the effect of thrombophlebitis was little affected by the presence or absence of positive pregnancy history in the rnodel (OR in absence of pregnancy history : 1.08. 95% CI: 0.32-3-65), suggesting that both are independent from each other. The absence of effect of thrornbophlebitis on varicose veins would be compatible with results of the follow-up of a series of 61 patients with deep vein thrombosis examined every 6 rnonths for an average of 39 months (range : 1 to 144 months). Symptoms. pigmentation. ulcer and swelling were shown to be longterrn sequels of deep vein thrornbosis, but not varicose veins (Strandness et al.. 1983).
Past histow of non-venous lea suraerv or leq trauma
Due to the small number of events that could be identified prior to the onset date of venous disorders as identified by the patients, the association of these variables with varicose veins could not be reliably estimated and has been dropped from subsequent analyses. Smoking Following publication of the results of Çramtngham Study, interest was raised in the similarity of risk factors for varicose veins and for atherosclerotic cardio-vascular disease. In the Framingham study. the mean number of cigarettes smoked per day was 7.5 in men with varicose veins and 5.6 in those without varicose veins.
The
corresponding figures for women were 3.9 and 4.7. The authors hypothesised that possible mechanisms may include a decrease in fibrinolytic activity. The relationship
was not confirmed in the population-based study conducted in Finland (Sisto et al.. 1995) and this study is also negative.
The hypothesis of an association between height and varicose veins is based on height increasing the orthostatic pressure in the legs. This mechanism is in line with the valvular hypothesis of varicose vein occurrence : valvular incornpetence would start from valves at the sapheno-femoral junction and would progress downwards. The association between height and varicose veins has been supported by a few studies (Beaglehole. 1976. Sisto et al.. 1995: Stewart et al. 1955) but not by others (Guberan et al.. 1973; Da Silva et al.. 1974). In the Sisto study. the relationship was found only in
women. with odds-ratios of 1.3. 1.3 and 1.4 in the three upper height quintiles after adjusting for several variables including age. AH confidence intervals excluded the nuIl value. In the present study, the odds-ratio was 1.04 (95% CI: 0.37-2.93) for women with a height ranging from 160 to 164 cm in cornparison to < 165 cm. and 1.77 (95% CI:
@
0.61-5.15) for women with a height 2 165 cm. No definite conclusion can be drawn frorn the available evidence. If it exists. the relative risk however is likely to ba srnall.
8.5.4 Conclusions Results of this casetontfol study of women aged 18-75 y(rs
consulting a physician
for venous disorders shows that age, pregnancy and a family history of varicose veins are risk indicators of presenting with varicose veins alone. The most important results are however that the effect of pregnancy does not increase beyond one pregnancy and does not appear ro be mediated by thrombophlebitis. A height of over 165 cm is also associated with a slightly increased risk of presenting
with varicose veins. Body mass index and thrornbophlebitis were not associated with varicose veins while a slightly decreased risk was present for smoking status. All results are adjusted for the effect of the risk indicators tested and for educational level. time of day at consultation. month of year at consultation and country.
9. Varicose veins and venous reflux
Controversy still exists about the pathogenesis of varicose veins, which could be a consequence as well as a cause of venous incompetence (Hobbs, 1991; Laing, 1992. Labropoulos et a1.,1997; Evans et al., 1998). Once a vein has started to dilate, its wall becomes weak and its valves incompetent. The weight of blood pressing down tends to dilate the vein further and thus render other valves incompetent (Laing, 1992). People with varicose veins can develop chronic venous insufficiency, a condition associated wlh incornpetence of the valves in the perforator veins, and with venous hypertension. The cascade of physiological consequences resulting fiom venous hypertension could include oedema, skin changes and, ultimately. ulcer (Widmer et al.. 1992).
As shown in the literature review chapter, the prevalence of superficial and deep venous incompetence is associated with the severity of varicosities and with the severity of signs associated with varicose veins (Rosfors et al., 1990; Neglên and Raju. 1993; Vasdekis et al, 1980). The investigation of venous incornpetence in patients with varicose veins therefore has an important prognostic value. The study of the location of refiux in subjects with early stages of varicose veins is also useful to clarify the aetiology of varicose veins : the theory of the primary cause of varicose veins being a weakening of the venous wall due to genetic predisposition would be supported if it could be shown that valve refiux can occur at distal levels of the venous system without involvement of the proximal systern (namely the sapheno-femoral junction) (Labropoulos et al., 1997). The iaterality of varicose veins and of valve incompetence is also important to investigate in order to examine the theory that raised intra-abdominal pressure. e.g. following constipation, might promote the development of varicose veins (Alexander,
1972; Burkitt. 1976; Geelhoed and Burkit. 1991). An increase in intra-abdominal pressure w u l d obstruct venous return fmm the legs and transmit pressure down the veins of the legs, leading to their dilation and to incornpetence of the valves. In the case of constipation, the sigmoïd colon would be heavier and exert a strong compressive effect on the left iliac vein. According to this theory, varicose veins would be more often left-sided than right-sided (Cornu-Thénard et al., 1986). Several studies have examined the laterality of varicose veins and answered this question negatively, but few of them have fonnally investigated whether this conclusion also applied to reflux. The laterality of reflux in relation to pregnancy history is also important to consider in order to examine the possible mechanisms of occurrence of varicose veins caused by pregnancy (Chapter 3, section 3.2.3). As venous refiux is routinely investigated to identify the anatomical site of reflux preoperatively. numerous cases series have been published on the location and patterns of reflux in patients undergoing surgery (Nicolaides et al., 1999). However. few of these studies can be used to determine the prevalence of venous reflux at different levels of the deep and superficial venous systems, to relate the patterns of reflux to the presence of varicose veins with or without other signs of venous disease, and to relate the presence of refiux in subjects with varicose veins to pregnancy history, genetic predisposition and orthostatic pressure .
9.2 Objectives
1. To estimate the prevalence of venous reflux in the deep venous system. the
superficial venous system and the perforators. in subjects without signs of venous disorders, with telangiectasia atone or with varicose veins.
2. To examine the distribution of venous reflux and investigate whether this distribution is compatible with a distal onset of venous reflux.
3. To examine the association between venous reflux and pregnancy, family histoiy of varicose veins and body mass index, which are variables suspected to be related to the occurrence of varicose veins.
9.3 Patients and methods 9.3.1 Ovewiew
Two sources of data were used : patients recruited in the VEINES study by specialists, and patients recruited by general practitioners and referred to a specialist for the Duplex study. O
Availability of results of echo-Doppler (duplex scan) for both legs was required to document reflux; in the Duplex study, the cutoff point to define reflux was a reflux duration > 1.O second. Clinical groups were defined by the presence of varicose veins and other signs of venous disorders. Location of reflux was defined as follows : superficial system alone, deep system alone (including perforaton), and superficial reflux combined with deep reflux or reflux in perforators. More precise location of superficial reflux was examined by taking the leg as the observation unit. Laterality of reflux was examined in relation to laterality of varicose veins and to pregnancy history. Unadjusted and adjusted logistic regressions were used to test the association of
.
age, excess weight-for-height family history and pregnancy history with the presence of superfcial or deep venous reflux. taking into account the effect of al1 other variables in the model. This analysis was perfomed in women only. Since patients could have a higher probability of referral to specialists because they had experienced thrombo-phlebitis in the past, the analysis of variables associated with reflux was restricted to the patients for whom no such history had been reported at the baseline visit.
9.3.2 Study population
The study population consisted in subjects included into one of the varicose vein categories presented in Chapter 6. This criterion excluded patients without varicose veins who had oedema, skin changes or ulcer. The study population was also restricted to patients recruited by specialists or referred to one of the specialists participating in the Duplex study. This restriction aimed to obtain an accurate diagnosis of both varicose veins and venous reflux. The general methodology of the VEINES and Duplex studies has been presented in Chapter 4. This analysis was initially based on 584 patients recruited by specialists and 146 patients referred to one of the seven specialists participating in the Duplex study.
Three of the 7 specialists who participated in the Duplex study were also participating in the VEINES study as investigaton.
9.3.3 Assessrnent of venous reflux
Demonstration of venous reflux based on duplex scanning was required to ascertain valve incornpetence because it is a non-invasive method that allows investigation of the
deep system and perforaton. Studies have shown it to be at least as good as descending phlebography in detecting venous reflux. The inter- and intrasbserver reproducibility in the measurement was qualified as 'reasonable' but was influenced by the position of the vein and by observer training and experience (Evans et al., 1993; Evans et al., 1995; Nicolaides et al.. 1999). Although information was available in the VEINES study regarding incompetence in the superficial system detected with the use
of continuous wave (c.w.) Doppler, these data were not used for this thesis, because an assessment of both the superficial and the deep systems in the same individual was required and could not be provided by C.W. Doppler alone.
The available information varied according to whether subjects were recruited by specialists or participated in the Duplex study.
Sample of patients recruited bv specialists participatina in the VEINES studv Venous reflux was assessed based on the information provided in the initial and each fcllow-up clinical examination. The questionnaire used by specialists included two sets of questions. In the fint one, the specialist was requested to specify the laboratory tests of venous function performeâ at any time in the past (baseline visit) or since the last visit (follow-up visits). If a test had been perfomed. information regarding the presence of reflux and its location (superficial system, perforating veins and deep system) had to be reported. In the second set of questions, detailed information was sought regarding the result of any continuous wave Doppler test or duplex scanning had been perfomed during the visit. The purpose of this was to provide investigators with forms that could be used to report test results in a standardised way. although the testing itself was not mandatory. For duplex scanning. the presence or absence of refiux in the following vein segments was elicited : superficial femoral vein, long saphenous vein. short saphenous vein, popliteal vein, gemellar veins, perforators. No cut-off point of reflux duration was recommended to define the presence of reflux. For this thesis, a subject was included in the analysis only if resuits about the presence or
absence of reflux were available for both legs.
When information was missing for a
vein segment. reflux for considered absent in this vein segment. Information collected in the three clinical exarninations perfomed in the course of the VEINES study was used to ascertain the presence of venous reflux.
Reflux was
considered present if it was recorded in at least one of the clinical examinations. Sample of ~atientsparticipatina in the Duplex study The protocol for the duplex investigation was based on recommendations issued by the specialists who participated as investigaton in the Duplex study.
The final
questionnaire derived from these recornmendations was approved by al1 investigators.
The duplex examination had to ascertain the presence of refiux in both legs in vein segments related to the superficial system. the deep system or the petforators. The following veins were investigated in each system :
supetficial system:
great saphenous vein : scarpa, mid-thigh, knees, legs short saphenous vein : poplitea fossa
deep system :
popliteal vein
perforator veins :
Hunter, Boyd, Cocket.
A cut-off of 1.O second for reflux duration (RD) was chosen according to the opinion of the expert panel in preference to a shorter duration because of a concern of a lack of specificity with short reflux. This concern was shared by the authors of a study published recently which compared reflux duration 2 0.5 and reflux duration > 1.O as the cut-off points to define significant reflux (Evans et al., 1998).
The authors
concluded, however, that using a RD >1.O as the cut-off point decreased the sensitivity and could define some incompetent veins as normal. It is therefore possible that the prevalence of venous incornpetence reported in the study is under-estirnated. When the study was planned, no attempt was made to standardise the technique, patient position and material used. A reliabiiity study published in the course of the
Duplex study showed that these conditions could have an effect on the reproducibility of the results of duplex scanning (Evans et al., 1995). The result of the duplex scan perfomed in the Duplex study was used, irrespective of the results of previous duplex tests reported by the general practitioner at the baseline examination.
9.3.4 Assessment of clinical status
The same questionnaire was used by specialists participating in the VEINES study and by those participating in the Duplex study. The presence or absence of the following
signs had to be recorded for each leg : hyperpigmented zone, dermatitis (eczema), lipodematosclerosis, atrophie blanche, ankle fiares, healed ulcer, open ulcer, oedema, varicose veins (if present : location and seventy on a 4-point scale), telangiectasia, reticular veins. The physical examination did not include objective measurements (kg-Ometer, calliper, ulcer size measurement).
In patients recruited in the VEINES study, the clinical status was defined at the time of the clinical examination and was not modified according to subsequent clinical examinations. In the Duplex study, clinical status was established according to the results of the physical examination recorded by the specialists, irrespective of the clinical status resulting from the general practitioners' examination.
9.3.5 Data analysis
An important difference between the two samples needs to be mentioned : duplex investigations perforrned in the course of the VEINES study were perforrned in the context of an obsewational study, and the indication of the duplex investigation impacted on the presence of reflux. white the Duplex study was interventional. i.e. the legs of al1 patients included in the study had to be investigated in full. In the first stage of analysis, the prevalence is therefore investigated separately. In the following stages, the two samples were merged because the pooled data offered a better representation of the patient mix of the source population than either of the two samples taken separately. Results of the pooled population are nevertheless influenced by the relative proportions of patients in the two samples. No attempt has been made to weight the prevalence found in each sample according to its representation in the source population : the available information was inadequate to compute weighting factors, and the differences in prevalence figures between the two samples were not large. All analyses used the patient as the statistical unit, except for the detailed analysis of vein segments, which used the leg as the statistical unit. Unadjusted and adjusted logistic regression analyses were used to estirnate the strength of association between several variables and reflux adjusted for the effect of other variables. A "case" was defined as a woman in whom supemcial or deep (including perforators) venous incompetence had been detected by duplex scanning in the course of the VEINES study or in the Duplex study; a "control"was defined as a
@
woman in whom a duplex scan was perfomed in either study and who was not classifled as a case WOrnen with a reported histoly of thrombophlebitis were excluded from the case and control groups.
The variables introduced in the model were selected according to prior knowledge of a possible association with varicose veins and reflux. The number of pregnancies was used to examine whether varicose veins occurring in the course or the aftennath of pregnancy could be associated with silent thrombophlebitis and valve destruction. Family history of varicose veins was used as an indicator of genetic predisposition acting, for example. on distensibility of vein wall and secondary val re incornpetence. Age and body mass index were introduced as they are considered to be associated with varicose veins and superficial reflux. These variables were defined as follows : a
age: 3 categories: 18-39.40-59.60-75 years
a
body mass index : 2 categories : 25.0
a
number of pregnancies : 2 categories :O, r 1;
family history of varicose veins in rnoaier : 2 categories: yeslno.
9.4 Results
9.4.1 Prevalence of venous reflux
Prevalence in specialist-base sample Duplex information was available for 225 of the 584 subjects included in the VEINES study by specialists (38.5 %). but detailed information for each leg was available for 170 of them (75.6%). Twenty-seven of these 170 subjects had aedema, skin changes or ulcer without varicose veins. The sample of the remaining 143 subjects included 32 men (22.4%). with a mean age of 52.6 years (range : 26-70), and 111 women (77.6%). with a rnean age of 49.6 years (range : 25-75). It included 106 subjects with varicose
veins and 37 without signs or with telangiectasia alone. Results of the duplex scan for these subjects are presented in Table 9.1 .A. Venous reflux was observed in 97.6% of patients with varicose veins alone and 64.9% of patients without signs or with
telangiectasia alone. In both cases, reflux in the deep system alone or in perforators alone was rare. and both categories were rnerged. The proportion of reflux restricted to the deep or perforating system was 11.3% in patients with varicose veins and was not considerably affected by the presence of signs of more severe venous disorders. An important difference between varicose vein categories arises frorn the proportion of patients with superficial and deep venous incornpetence combined. This proportion ranged from 26.2% (varicose veins alone) to 75.0% (healed or open ulcer combined with varicose veins). In the majority of patients with varicose veins alone, reflux is restricted to the superficial system (61.9%), whereas the deep system is involved in 100% of cases with open or healed ulcer.
Prevalence in the Duplex study Of the 146 patients included in the Duplex study (Table 9.1 .B),27 had oedema. skin changes or ulcer without varicose veins and were excluded.
Amongst the 119
remaining patients, 32 (26.9%) were males, with a mean age of 59 years (range: 2975), and 87 (73.1%) were females. with a mean age of 55 years (range: 23-74). Major
differences between the results of the Duplex study and the results from the specialistbased VEINES study are in the lower proportions of subjects without varicose veins who have venous refiux (27.3% vs. 64.9%), and the larger proportion of subjects with varicose veins who have reflux in the superficial system alone (56.0% vs. 44.3%).
Table 9.1 Prevalence (%) of venous reflux as ascertained by duplex scan in the specialist-based sample of the VEINES study (A). the Dupiex study (6) and both combined (C) (S= supenicial system. D= deep systern, P= perforaton. DIP : deep system or perforators, W=varicose veins) A Specialist-based sample of the VEINES study
Varicose veins atone (n=42)
Vaficose veins and oedema (n=12)
W and skin changes (n=40)
W and healed or active ulcer (n=12)
All patients with W (n=106)
(n=44)
Varicose veins alone (fi=28)
Varicose veins and oedema (n=6)
W and skin changes (n=39)
W and healed or active ulcer (n=2)
All patients with W (n=75)
none
72.7
7.1
33.3
10.3
0.0
10.7
S alone
17.4
75.0
33.3
46.2
50.0
56.0
DIP alone
6.8
0.0
0.0
2.6
0.0
1.3
S + DIP
9.1
17.9
33.3
41.O
50.0
32.0
27.3
92.9
66.7
89.7
100.0
89.3
No signs or telangiectasia alone (n=81)
Varicose veins alone (n=70)
Vancose veins and oedema (n=18)
W and skin changes (n=79)
W and healed or active ulcer (n=14)
All patients with W (n=181)
none
55.6
4.3
11.1
5.1
0.0
5.0
S alone
27.2
67.1
33.3
43.0
14.3
49.2
6.2
5.7
11.1
6.3
14.3
7.2
S + DIP
11.1
22.9
44.4
45.6
71.4
38.6
S or P or D
44.4
95.7
88.9
94.9
100.0
95.0
Levet of reflux
No signs or telangiectasia alone (n=37)
B Duplex study
Level of reflux
S or P or D
No signs or telangiectasia alone
C Pcoled data
Level of reflux
DIP alone
Prevalence in pooled data Table 9.1.C. presents the prevalence of superficial and deep reflux in 262 subjects representing a mix of patients included in the VEINES cohort and in whom a complete duplex scan was perfomed during the study period. The following conclusions can be drawn about the prevalence of venous reflux in the study population : In patients without signs or with telangiectasia alone. an estimated 44.4% had reflux in the superficial or the deep systems. including 38.3% of subjects with reflux in the superficial system and 17.3% of subjects with reflux in the deep system. O
Reflux in any system was present in about 95% of subjects with varicose veins; this proportion was 96% in individuals with varicose veins alone and 100% in subjects with varicose veins and healed or active ufcer.
a The proportion of reflux in the deep system alone was low and moderately affected
by the severity of signs associated with varicose veins. The prevalence of reflux in both systerns increased with the severity of signs associated with varicose veins, ranging from 22.9% in subjects with varicose veins alone to 71.4% in patients with varicose veins and ulcer.
9.4.2 Distribution of reflux. by Ieg
The location of reflux by leg and clinical category is presented for the pooled data in Table 9.2. Results of the duplex examination are available for 524 legs (DNo legs for each of the 262 subjects). Seventeen legs had oedema, skin changes or healedlopen ulceration without varicose veins and were not included in the analysis. In subjects with reflux in the supeficial system alone. reflux in the sapheno-femoral junction alone and in the long saphenous vein below the knee alone were found in
10.8% of legs or less. Reflux in the short saphenous vein alone was found in 16.2% of legs without varicose veins, but in less than 5% of legs with varicose veins. In the majority of legs. superfcial reflux was found in the long saphenous vein. alone or in combination with reflux in the short saphenous vein. Although this evaluation is based
on small numbers. a trend by clinical category was found for the combination of reflux at the sapheno-femoral junction and in the long and short saphenous veins (2.7%,
8.6% and 19.4%). Table 9.2 Location of reflux in 507 legs, by clinical category No signs or telangiectasia alone (n=207)
Varicose veins alone (n=146)
Varicose veins and other signs (n=7 54)
Location of reflux ' None Superficial alone SFJ alone LSV above knee alone LSV below knee alone LSV below and above knee SSV alone SSV and LSV alone SFJ + LSVSSV unknown
Deep systern/perforators alone Superficial + D/P S or D or P
'
SFJ = Sapheno-femoraljunction; LSV : long saphenous vein; SSV : short saphenous vein: LSVISSV : LSV or SSV; D/P : deep system or perforators
The main result of this investigation is that long and short saphenous vein reflux
exists in a large proportion of cases even where there is no reflux at the saphenofemoral junction.
9.43 La teraüty o f reflux
The presence or absence of reflux in both legs is known for 181 patients with varicose veins. Among the 61 patients with unilateral varicose veins, the location of varicose veins was in the right k g in 32 individuals (52.4%) and in the left leg in
29 individuals (47.6%). There is therefore no indication that varicose veins are locateâ mm frequently on the le* leg than on the right kg. Tables 9.3 and 9 4 present the laterality of superficial and deep venous refiux in subjects with unilateral
(Table 9.3) and bilateral (Table 9.4) varicose veins. In only a very small number of cases with unilateral varicose veins (Table 9.3), was supeficial and deep reflux located only on the contra-lateral leg. In the majority of patients, reflux was found only on the same leg, suggesting that a local rather than a systemic cause is linked to reflux and varicose veins. Results in the 120 subjects with bilateral varicose veins (Table 9.4) show that supeficial and deep reflux were absent in 19.2% and 46.7% of cases respectively. Among patients with reflux, unilateral superficial and deep reflux were found in 30% of cases. Table 9.3 Location of reflux in 61 patients with unitateral varicose veins Superficial reflux
Location of reflux :
n
Deep reflux
n
%
Oh
None Ipsilateral leg only Contralateral leg only Bilateral
Table 9.4 Location of reflux in 120 patients with bilateral varicose veins - -
Superficial reflux Location of reflux : none unilateral bilaterat
n
Oh
- -
-
--- -
-
-- -
Deep reflux
n
%
9.4.4 Lateraiity of varicose veins and reflux according to prepancy
The hypothesis that pregnancy history is associated with varicose veins being more frequently left-sided than rîght-sided is examined in Table 9.5 (pooled data).
Information on reproductive history was missing for 30 of the 198
women. Amongst the remaining 168 women, 66 (39.3%) had no varicose veins, 16 (9.5%) had right-sided varicose veins only, 15 (8.9%) had left-sided varicose veins only and 71 (42.3%) had bilateral varicose veins. Amongst the 146 women who had had at least one pregnancy in the past. a total of 78
(53.4%) had left-sided varicose veins and a total of 80 (54.8%) had right-sided varicose veins.
Thus, in both groups there is a trends towards a larger
proportion of patients with right-sided varicose veins than left-sided varicose veins, and the hypothesis of a left predominance of varicose veins in pregnant women is not confirmed. The same conclusion can be drawn from the analysis of the laterality of supeficial (Table 9.6) and deep (Table 9.7) venous reflux. In Table 9.7, the proportions of women with 21 pregnancy with reflux in the deep system on the left and the right legs only are 13.7% and 9.6% respectively. The difference of 4.1% is not statistically different from the nuIl value at the 5% level. Therefore,
there is no left predominance of reflux neither in the superficial nor in the deep system in relation to pregnancy history. Table 9.5 Laterality of varicose veins according to pregnancy history history in 198 women '
Location of varicose veins
Number of pregnancies 21 none (n=22) (n=146) n % n %
None Right kg only
Left ieg only Both legs
' Information O pregnancy history is missing for 30 women.
Table 9.6 Location of superficial venous reflux according to pregnancy history in 198 women '
Number of pregnancies
1 Location of superhcial reflux
n
%
n
%
None Right leg only Left leg only
Both legs l
Information O pregnancy history is missing for 30 wornen.
Table 9.7 Location of deep venous reflux according to pregnancy history in 198 women '
Location of deep reflux
Number of pregnancies none 21 (n=22) (n=146) n O h n %
None Right k g only Left kg only
60th legs Information O pregnancy history is missing for 30 women.
9.4.5 Variables associated with venous reflux
Frorn the 240 wornen. 52 had a history of past thrornbophlebitis and were exciuded from the analysis (26 in each sample). Therefore, the analysis of variables associated with venous reffux used data on 188 wornen (104 included in the specialist-based
VEINE study and 84 included in the Duplex study). Results of the unadjusted and adjusted analyses are presented in Table 9.8. In unadjusted analysis, age and pregnancy history were the variables most strongly
associated with reflux, with 95% confidence intervals excluding the nuIl value only for
@
age categories. ln multivariate analysis. age rernained associated with reflux with a
95% confidence interval excluding unity. Odds-ratios of 1.37 and 1.29 were found for pregnsncy history and maternai history of vaiicose wins, with a 95% CI insluding unity.
Body mass index was not associated with the presence of reflux..
Table 9.8 Association betweem venous reflux and suspected determinants in women without a history of thrornbophlebitis Distribution of deteminant in controls (n=58) -
cases (n=130)
unadjusted odds-ratio
adjusted odds-ratio '
-
Aae arouP (years) 18-39 40-59 60-75 Bodv mass index < 25 2 25 unknown Preanancv history none r 1 pregnancies unknown Varicose veins in mother no Yes Sample sue in final mode1 due to missing values : 48 controls. 109 cases; odds ratio adjusteâ for al1 variables in the table
9.5.1 Prevalence of venous reflux
There are few available data on the prevalence of venous reflux in individuals presenting to physicians with varicose veins (W)at an early stage of development. and particularly with W alone. Three studies allow estimation of the prevalence of reflux in subjects with whom cornparison to those without signs or with telangiectasia alone could be performed. In a group of patients consulting for cosmetic complaints and rated either as 'clinically normal' or as patients with spider veins only, the prevalence of superficial and deep reflux (including perforaton) in 102 legs was 25.5% and 1.0% respectively (Thibault et al., 1990). In a total of 120 legs of 'normal' volunteers or individuals with prominent non-varicose veins referred for mild symptoms, the prevalence of superficial and deep reflux were 31.7% and 6.7% respectively (Labropoulos et al.. 1997).
In a survey of 630 "venous disease-free" participants
sampled from the general population in Scotland, 18% of subjects had evidence of reflux duration (RD) > 1.O second in at least one vein segment of the supeficial system, and 4%-5% showed R D 4 .O sec in the popliteal vein of the deep system (Evans et al., 1998). Figures of 38.3% and 17.3% respectively were found in this study (pooled data). Whereas data for the superficial system are comparable to those reported in other studies, those conceming the deep system show higher figures. The results of this study regarding the group of patients with varicose veins (W) alone can be compared to those presented in Table 3.8. In 28 limbs with W and symptoms,
Vasdekis et al. (1980) found reflux in 60% in the long saphenous veins (LSV), reflux in 2 9 2 in the short saphenous vein (SSV) and reflux in 11% in the deep systern. In
another study on a total of 398 limbs with W alone or W and spider veins (Thibault et al., 1990). 52.8% had superficial reflux and 12.6% had deep incornpetence. These proportions were 87.0% and 13.0% in the study by Labropoulos et al. (1997) on patients with W alone consulting for venous disorders. In this study, superficial and deep reflux were found in respectively 90.0% and 28.6% of the legs with W alone. Given the differences in study design and rnethods, the proportion of lirnbs with
superficial refiux found in this study is thus comparable to that reported in the other studies. The proportion of limbs with deep reflux is higher, even when only the results fom the Duplex study are considered. The explanation of the higher prevalence of reflux in the deep systern is not clear but may be found in the study population originating from patients consulting with complaints related to venous disorders. Moreover, only subjects with a complete evaluation of the vein systems were included in this study, and the above percentages refer to subjects with reflux obsewed in at least one segment. As shown in Chapter 11 below, the proportion of patients with leg symptoms is high. In some patients, these symptoms may be related to deep venous incompetence.
9.5.2 Distribution of venous reflux
Because the venous pressure in the lower limbs is increased due to hydrostatic reasons in the upright posture, it has traditionally been believed that reflux develops in a retrograde fashion (Labropoulos et al., 1997). According to this theory, incompetence of valves in proximal vein segments would be the initiating factor for a development of reflux. Many physiological, morphologie and biochemical studies have shown that venous wall changes can occut irrespective of the function of the valves (Laborpoulos et al., 1997). The investigation of the vein segments presenting with refiux at early stages of the development of varicose veins was an important contribution to the understanding of the role of the 'retrograde hypothesis' in their aetiology. Few Duplex studies aiming to investigate this hypothesis were available at the start of this project. The results of the study published by Labropoulos et al. in 1997 provided a detailed account of the distribution and extent of reflux in a large number of vein segments, but was based on a limited sample sire.
The results of this study reach the same
conclusions as those reported by Laborpoulos et al.: O
In patients with early stages of valve dysfunction, the development of the reflux appears to be most often a local process that can develop in any part of the lower limb venous system, and parbicularly in the long saphenous vein domain.
Deep venous reflux alone is uncornmon in early stages of varicose vein development. The extent of reflux is associated with the clinical seventy of the venous disease. Labropoulos et al. also concluded that reflux can often develop in the below-knee segment of the long saphenous vein. The term 'often' does not apply to this study, because reflux was exclusively found in this location in 5 of 507 legs only. The findings related to ths location of valve incornpetence in subjects with early stages of varicose vein development have important implications regarding the aetiology of varicose veins. They exclude that hydrostatic pressure and obstruction of the venous reflux by intra-abdominal pressure are necessary to cause varicose veins. The findings also provide a rationale for failures of surgical treatment : ligation of the saphenofemoral junction may not be sufficient to eradicate varicose veins and can result in their persistence or recurrence (Royle, 1986; Sarin et al., l992b; Labropoulos et al., 1996). The examination of reflux in the non-affected contra-lateral leg provides an additional argument against the theory that the primary cause of varicose veins is intra-abdominal pressure. Superficial and deep reflux were infrequently found in the contra-lateral leg of subjects with unilateral varicose veins (3.3%),suggesting that varicose veins and reflux originate hom a leg specific cause. Moreover, unilateral reflux was found in only 30% of patients with bilateral varicose veins. This finding supports the vein wall defect theory that dilation of the vein precedes the developrnent of valve incornpetence. In line with the theory of varicose veins development based on increased intraabdominal pressure, some authors have suggested that pregnancy-related varicose veins and reflux would be more frequently left-sided than right-sided. This assertion is refuted by a simple analysis of the distribution of varicose veins and reflux by categories of pregnancy history. No preferential laterality was found in the results of this study.
9.5.3 Variables associated with reflux
Varicose veins are usually classified as 'ptimary' or 'secondary' according to whether the underlying cause is thrombophlebitis or an initial defect of the vein wall or valves.
The only way to differentiate between these two groups is thrombophlebitis history. It is assumed that varicose veins are of 'primary cause' in subjects without a history of thrombophlebitis.
The only way to differentiate between these two groups is by
thrombophlebitis history. In this study. a multivariate risk model was built only for women without such as a history because it was considered that thrombophlebitis could be an indication for duplex scanning. It is therefore postulated that varicose veins in
women included in the analysis of detemiinants of reflux are of primary origin. Primary varicose veins are often familial and caused by weakness of the vein wall or valvular incornpetence. In this analysis. history of materna1 varicose veins was used as
a 'proxy' variable for genetic disposition to venous reflux. In multivariate analysis. the relative risk estimate is 1.29 (95% CI: 0.48-1.96). This moderate association is perhaps explained by the fact that reflux is not a prerequisite for the development of varicose veins. The sample used in the analysis also includes women who do not have varicose veins. Although pregnancy history was strongly associated with varicose veins (OR of 2.96). it is weakly associated with reflux. either in unadjusted cr in multivariate analyses
(OR : 1.37). This finding does not help in detemining the cause of pregnancy-related varicose veins. However, it suggests that development of reflux is not necessarily an intemediate factor in the pathogenesis of varicose veins occurring in pregnant women. Excess weight-for-height (BMI 2 25) was not associated with the presence of reflux (OR
of 0.93, 0.44-1.96). This result is in accordance with the results of the analysis of rkk indicators of varicose veins presented in Chapter 8, where no association with the presence of varicose veins was found.
10. Complications f0.1 introduction Complications addressed in this chapter are adverse outcornes associated with venous disorders and possibly precipitated by the presence of varicose veins. They include the occurrence of signs that generally indicate an aggravation of the disease. such of oedema, skin changes and ulceration. Few studies have examined their incidence in patients with varicose veins. The main source of information currently available is the 11-year follow-up of the chemical workers in the Basle Study (Widmer et al., 1992). The
percentages of subjects developing CVI, oedema or ulcer depended on the severity of varicose veins.
For example, the cumulative incidence of ulcer in patients without
varicose veins and with mild. moderate and severe varicose veins were 0.8%, 1.O% and
20.0% respectively. Other data are presented in Chapter 3, section 3.2.4, Table 3.13. In chapten 3 and 8, it has been alluded to the much debated question of whether varicose veins are associated with a higher risk of developing thrornbophlebitis. Based on an editorial of the Lancet (Campbell. 1996), it has been concluded that varicose veins are normally not accompanied by a higher risk of occurrence of thrombophlebitis. This chapter examines the hypothesis that patients with varicose veins alone have a higher risk of thrornbophlebitis than patients without varicose veins and without signs of more severe venous disordem. There are very few medical complications that can be directly attributed to the presence of varicose veins. The most significant one is a bleeding of varicose veins due to
trauma or to bumping into things. The frequency of this outcome is also examined in
this chapter. Undesirable effects due to treatments (e.g.. sensorimotor impairment below knee following vein stripping) are not considered in this thesis.
10.2 Objectives To measure the cumulative incidence and the incidence density of oedema, skin changes and ulceration during follow-up. a
To measure the cumulative incidence and incidence density of thrombophlebitis and detemine whether it occurs at a higher rate in patients with varicose veins alone than in patients without varicose veins and without signs of more severe venous disorders.
a
To estimate the cumulative incidence of bleeding of varicose veins.
IOJ. 1 Study population The study population comprised 1313 patients included in one of the categories of the
classification of varicose veins, i.e. patients without any signs, patients with telangiectasia only, and patients with varicose veins alone or associated with other signs. The numbers of patients included in various analyses varied according to the sign whose rate of occurrence was examined, as incidence rates were estimated in patients free of that sign at the baseline examination. Patients with healed ulcer were also excluded from the population at risk of developing a new ulceration.
10.3.2 Source of data
Results presented in this chapter are based on data recorded by investigators at each of the follow-up visits. Oedema and ulceration were assessed on each leg and recorded as absent or present . 'Skin changes' is a composite variable created at the data analysis phase. It was equal to 1 in presence of at least one of the following sign : hyperpigmentation, eczerna or denatitis along the venous pathway, lipodenatosclerosis, and atrophie blanche. It took the value O if al1 of these signs were absent. The only time referent for the information on presence of oederna, skin changes and ulceration was the
date of the clinical examination during which it was assessed. lnvestigators also had to record whether the patient had presented superficial or deep thrombophlebitir vein bleeding and ulceration since the previous vise made in the context of the VEINES study. This information included the date on which these events had occurred, and the kg.
10.3.3 Data analysis
lncident cases of oedema, skin change and ulceration were defined as individuals free of the sign of interest at the baseline examination and who presented with it on any leg at any of the two follow-up visits. For ulceration. the definition also included the record that ulceration had occurred during the period since the previous VEINES visit, even no active ulceration was obsenred by physical examination. For non case patients, the observation time was calculated as the difference between the dates of the last and first clinical examinations. For incident cases, the observation time was calculated as the difference between the date of the clinical examination at which the patient presented with the sign and the date of the first clinical examination. Incident cases of thrornbophlebitis and bleeding were defined based on the record that such events had occurred on the left or the right leg since the previous VEINES visit. The obsewation time was calculated as the difference between the date of the occurrence of the event and the date of the first clinical examination. In case of missing information on the date of occurrence, the date of the follow-up clinical examination was used instead. When either thrombophlebitis or bleeding were recorded in the two clinical exarninations, only the earliest of the two events was considered. Due to the small number of deep thrombophlebitis recorded, superficial and deep thrombophlebitis were pooled into a single category. Cumulative incidences and incidence densities (with their 95% confidence interval) were calculated in each varicose vein category using standard formulas (Walker, 1991, p. 99). Incidence densities were reported per 100 person-yean of observation.
Due to possible diagnostic problems in the general practitioner-based sample, separate
analyses for GPs and specialists were perfomed.
10.4.1 Oedema
Very large incidence rates of oederna were found in al1 varicose veins categories, both in the GP based- and the specialist-based samples (Tables 10.1 and 10.2). In the specialist-based sample, oederna was reported in more than 90% of subjects free of oedema at the baseline examination, with an incidence density ranging from 208 to 230 per 100 person-yean.
It was as frequent in subjects without signs or with
telangiectasia alone as in the other categories. In the GP sample, the proportion of subjects developing oedema was 4540%. with an average incidence density of 71.8 per 100 person-years in patients with varicose veins. There was no statistically significant difference between the group of patients with varicose veins alone and the group without signs or with telangiectasia only.
Table 10.1 Cumulative incidence (%) and incidence density (per 100 person-years) of oedema during follow-up, by clinical category general practitioner-based sarnple (VV=varicose veins)
-
--
N patients at nsk
Observation time (years) lncident cases cumulative incidence estimate 95% CI incidence density estimate 95% CI
No signs or telangiectasia
Varicose veins alone
96 72.8 57
232 180.3 123
59.4 49.5-69.2
53.0 46.9-59.4
78.3 58.0-98.6
68.2 56.2-80.3
-.
W and skin changes
-
W and Al! healdopen patients with W ulcer
Table 10.2 Cumulative incidence (%) and incidence density (per 100 person-years) of oederna during follow-up, by clinical category specialist-based sample (W=varicose veins)
-
No signs or telangiectasio
Vancose veins alone
W and skin changes
141 58.1 124
115 46.4 107
69 30.2 63
N patients at risk Observation tirne (years) Incident cases
W and All healedlopen patients with ulcer W
29 12.3 27
213 88.9 197
cumulative incidence estimate 95% CI incidence density estimate 95% CI
10.4.2 Skin changes
Skin changes developed in about 70% of patients with varicose veins seen by general practitioners and in 50% of patients seen by specialists (Tables 10.3 and 10.4). Incidence densities are however comparable. with figures ranging from about 80 to 100 incident cases per 100 person-years of observation. Table 10.3 Cumuiative incidence (%) and incidence density ( per 100 person-years) of skin changes during follow-up, by clinical category general practitioner-based sample (W=varicose veins)
-
N patients at risk Observation time (years) lncident cases
cumulative incidence estimate 95% CI incidence density estimate 95% CI
No signs or telangiectasia
Varicose veins alone
Varicose veins and oedema
Ali patients with W
96 70.8 57
232 175.0 158
88
63.0 65
320 238.0 223
59.4 49.5-69.2
68.1 62.1-74.1
80.5 59.6-10 t .4
90.3 76.2-104.4
Table 10.4 Cumulative incidence (Oh)and incidence density (per 100 person-years) of skin changes dunng follow-up. by clinical category specialist-bas& sample
-
No signs or telangiectasia
Varicase veins alone
Varicose veins and oedema
Al1 patients with
141 88.9 73
115 66.0 54
33 20.1 15
148 86.1 69
51.8 43.5-60.0
47.0 37.8-56.1
45.4 28.5-62.4
46.6 38.6-54.7
82.1 63.3-100.9
81-8 60.0-103.6
746 36.9-1 12.4
80.1 61.2-99.1
N patients at risk Obsewation time (years) lncident cases cumulative incidence estimate 95% Ct incidence density estimate 95% CI
w
10.4.3 Ulceration
A total of 53 new ulcers were detected during the follow-up of the VEINES study. 35 in
the GP-based sample and 13 in the specialist-based sample (Tables 10.5 and 10.6). The cumulative incidence rates and incidence densities were on average higher in the GP-based sample than in the specialist-based sample, but did not Vary by category of venous disorden in bath samples (Tables 10.5 and 10.6). Due to the small numbers of events, al1 confidence interval are large. in the specialist-based sample, the lower and upper limits of the 95% confidence interval of the incidence density were 1.0 and 6.5 respectively per 100 penon-years of observation. Table 10.5 Cumulative incidence (%) and incidence density (per 100 person-years) of ulceration during follow-up. by clinical category general practitioner-based sarnple (W=varicose veins)
-
-
N patients at risk Observation time (years) Incident cases cumulative incidence estirnate 95% Cl incidence density estirnate 95% CI
No signs or telangiectasia
Varicose veins alone
Varicose veins and oedema
96 83.8 5
232 209.6 13
88 81.7 5
5.2 0.8-9.6
5.6 2.6-8.6
5.7 2.6-8.5
6.0 0.7-1 1.2
6.2 2.8-9.6
6.1 0.8-11.5
W and skin changes
patients with
6.1 3.2-9.0
6.2 4.1-8.2
All
w
Table 10.6 Cumulative incidence (%) and incidence density (per 100 person-years) of ulcefation during follow-up, by clinical categocy specialist-based sample (W=vancose veins)
-
N patients at nsk Observation time (years) Incident cases
cumulative incidence estimate 95% CI incidence density estimate 95% CI
w
No signs or telangiectasia
Varicose veins alone
Varicose veins and oedema
and skin changes
All patients with W
141 106.9
5
115 78.0 5
33 23.0 1
106 86.0 1
254 187.0 7
3.5 0.54.6
4.3 0.6-3.7
3.0 0.04.9
0.9 0.0-2.8
2.8 0.7-4.8
4.7 0.6-8.8
6.4 0.8-1 2.0
4.3 0.0-1 2.9
1.2 0.0-3.4
3.7 1. O 6 5
10.1.1 Thrombophlebitis
Tables 10.7 and 10.8 below show nearly identical incidence densities thrombophlebitis diagnosed by generalists and specialists. except in the 'Varicose veins alone' category. In both samples incidence densities increase with the severity of signs associateci with varicose veins. In the specialist-based sample. they range from 1.3 per 100 person-years in patients with varicose veins alone to 12.9 per 100 person-yean in patients with varicose veins and healed or open ulcer. The comparison of incidence densities observed in patients with varicose veins alone and in patients without signs or with telangiectasia alone showed a incidence density ratio of 3.24 (95% CI: 0.40-25.93) in the GP-based sample and of 0.68 (0.06-7.62) in the specialist-based sample. These results do not indicate that varicose veins are on their own associated with a higher risk of developing thrombophlebitis.
Table 10.7 Cumulative incidence (%) and incidence density (par 100 person-yean) of thrombophlebitis dunng follow-up. by clinical category - general practitioner-basadsample (W=varicose veins)
N patients at risk
Observation time (years) Incident cases cumulative incidence estimate 95% Cl incidence density estimate 95% CI
W and W and skin healedlop changes en ulcer
All patients with W
No signs or tebngiectasia
Varicose veins atone
Varicose veins and oedema
96 84.9 1
232 209.4 8
88 81.5 4
305 275.3 17
80 73.3 9
705 639.4 39
1.O 0.0-3.1
3.4 1.1-5.8
4.5 0.2-8.9
5.6 3.0-8.9
11.2 4.3-18.2
5.5 3.8-7.2
1.2 0.0-3.5
3.8 1.2-6.5
4.9 0.1-9.7
6.2 3.2-9.1
12.3 4.3-20.3
6.1 4.2-7.2
Table 10.8 Cumulative incidence (%) and incidence density (per 100 person-years) of thrombophlebitis during follow-up, by clinical category specialist-based sample (W=varicose ~eins)
-
W and All healedlop patients en ulcer with W
No signs or telangiectasia
Varicose veins alone
Varicose veins and Oedema
W and skin changes
N patients at risk Observation time (years) Incident cases
141 107.6 2
115 77.9 1
33 22.3 1
106 83.8 7
cumulative incidence estimate 95% CI
1.4 0.0-3.4
0.87 0.0-2.6
3.03 0.0-8.9
4.4 2-1-67
incidence density estimate 95% CI
1.9 0.0-4.4
1.3 0.0-3.8
4.5 0.0-1 3.4
2.8-9.3
42 30.9 4
295 214.9 13
6.0
10.1.5 Bleeding of varicose veins
A total of 10 cases of bleeding of varicose veins were reported during the study, 7 by general practitionen and 3 by specialists. One case in each sample was reported in patients with telangiectasia alone. A total of 8 cases were therefore reported in 1054
subjects with varkose veins, with a cumulative incidence rate of 0.76% and a 95% confidence interval of 0.24-1.28%. This figure represents the number of patients who
experienced one event and not the total number of events. Bleeding of varicose veins was experienced by 2 patients with varicose veins alone ( O S % ) , 4 subjects with varicose veins and skin changes (0.93%) and 2 patients with varicose veins and open ulcer (6.5%).
10.5. Discussion
High incidence rates were observed in this study. both for signs of aggravation of the venous disorder and for thrornbophleblis. The high rates observed for oedema and skin changes are particulariy striking. The
only comparative data on the incidence of oedema and skin changes in patients with varicose veins were collected in the Basle Study in the Il-year follow-up of workers of the chernical industry mdrner et al., 1992). The highest cumulative incidence figures of oedema and chronic venous insufficiency were 50% and 30% respectively in patients
with severe varicose veins. Several factors could at least partly explain of the high rates of oedema and skin changes found in the VEINES study : Problems of definition. The definition of skin changes included signs with low specificity.
such as 'hyperpigmentation'. a sign that is also included in the international CEAP classification (Porter et al.. 1995). Misclassification. Btankfield et al. (1998) investigated the validity of a clinical diagnosis
of bilateral leg oedema made by family physicians. They found that venous insufficiency was the initial clinical impression in 71% of patients. although only 22% of subjects actually had this condition after case validation. However. the incidence rate of oedema in this study was higher in patients consulting specialist, who. in pr~nciple.should be more accurate in diagnosing oedema of venous origin. Referral Mas. Oedema (and perhaps also skin changes) is a transient condition that
many people could present at least once a year. Since the date of the first follow-up
clinical examination was not fixed, it is possible that the date of retum to the specialist was detemined by the occurrence of oedema. Clearly, further investigations are needed to evaluate the incidence of oedema and skin changes (and its deteminants) in patients with venous disorden. These investigations are beyond the scope of this thesis. However. the data suggest that the occurrence of these N o signs are not specifically associated with the presence of varicose veins. About 6% of patients in the GP-based sample and 3% in the specialist-based sample developed new ulceration in the course of the VEINES study. These figures corresponded to incidence densities of about 6 and 4 per 100 person-yean. They are unexpectedly high in view of the Il-year incidence rate reported by Widmer et al. (1992): 0.8% in patients with mild varicose veins, 1.0% in patients with moderate varicose veins and 20.0% in patients with severe varicose veins. Part of the difference may certainly be explained by the characteristics of the populations surveyed. In one study based in Newcastle, England, in which district nurses retrospectively reviewed new cases coming to their attention, the annual incidence in the population over 45 years of age was estimated to 3.5 per 1000 (Lees and Lambert. 1992); this figure is about 10 times lower than the rate found in the VEINES study in the group of patients without varicose veins. The method used in the study carried out in Newcastle is certainly prone to underestimation of the incidence. Moreover, the VEINES study population includes people with complaints related to venous disorden and a large proportion of venous reflux.
Cumulative incidence rates and incidence densities
reported by specialists are therefore thought to reflect the real incidence in patients consulting for venous disorders. The higher rates reported by general practitioners may be due to a mix of ulcers of various origins. Factors associated with the occurrence of ulcen have not been examined because it required consideration of a large nurnber of specific issues outside the scope of this thesis.
The examination of the incidence rate of thrombophlebitis was primarily aimed at a comparison between subjects with varicose veins alone and subjects without varicose veins and other signs of more severe venous disorders. The results do not support the hypothesis that varicose veins induce a higher risk of developing thrombophlebitis. The data also show an increase of the incidence rate of thrombophlebitis with the severity of signs associated with varicose veins, and this result is found in both general practitionerand specialist-based sample. The cumulative incidence rate of bleeding of varicose veins was 0.76%. This figure is not negligible and may have relevance for further studies.
Il Symptoms and quality of life
1 .f Introduction
Several conclusions have been drawn fiom the literature review presented in section 3.3. They can be summarised as follows.
Svm~torns There has been no standardised and psychometrically-sound way of summarising the burden of symptoms; various symptoms have been measured and scales that have been used have not been psychometrically evaluated in ternis of reliability, validity and responsiveness. Symptoms reported by patients with venous disease are also reported by a large proportion of non affected individuak. Few studies have systematically examined the prevalence of symptoms in
individuals with varicose veins alone, the conditions in which these symptoms are alleviated or aggravated, the association with seventy of varicose veins and the determinants of symptoms in patients with vaflcose veins. Qualitv of life Only few studies have evaluated quality-of-life in patients with varicose veins. Disease-specific quality of life scales have generally not been comprehensively evaluated in ternis of reliabilq, validity and responsiveness. rn
In one study, SF-36 has been found responsive in detecting changes in quality of life after surgery for varicose veins.
There are no data on the impact of varicose veins on quality of life after adjusting for the effect of other quality of life deteminants.
This chapter aims at evaluating the usefulness of symptom reports and quality of life measures for assessing consequences of varicose veins and clinical benefits to be expected from varicose vein treatrnents.
V . 2 Objectives 11.2.1 Cenenl objective
To determine the association between varicose veins and venous symptoms and quality of Me, taking into account hypothesised deterrninants of outcorne. 11.2.2 Specific objectives
Svm~toms 1) To estimate the prevalence of individual symptoms in men and women with
varicose veins consulting general practitioners and specialists. 2) To investigate whether varicose veins alone are associated with symptoms in ternis
of :
- specificity of symptoms in patients presenting with varicose veins
- effect of varicose veins alone on a global symptom score. Qualitv of life 4) To assess quality of life in men and women with varicose veins consulting general
practitioners and specialists, using generic and disease-specific measures. 5) To measure the effect of varicose veins on qualdy of life.
11.3.1 Overview
Svmotoms 1. The prevalence of individual symptoms was estimated in six clinical categories :
no signs or telangiedasia alone, varicose veins alone, varicose veins with oedema, varicose veins with skin changes, varicose veins with healed ulcer, varicose vein with open ulcer. The estimation was based on patient-reports of individual symptoms. Gender-specific prevalence figures were standardised to the age distribution in al1 registered patients. The prevalence was estimated for al1 patients with varîcose veins using standardisation by the expected distribution of varicose vein categories in the registration pool. Statistical testing was appiied to differences in syrnptom prevalence between patients without any sign or with telangiectasia alone, and subjects with varicose veins alone. 2. The investigation of the specificity of symptoms associated with varicose veins was based on physician-rated symptom data; this evaluation used the leg as the unit of obsewation and took place at the time of the clinical examination. The question was addressed by :
a) estimating, in individuals wlh varicose veins reported on one leg only, the prevalence difference and ratio that symptoms are experienced more frequently on the affeded kg. In this analysis, legs with varicose veins were matched to legs without varicose veins on patient characteristics. As the association could be cunfounded by observation bias if oedema. skin changes or healedlopen ulcer were
present on the contralateral kg. subjects with any of these
clinical conditions
were excluded.
b) testing whether there was a linear trend between the prevalence of symptorns and
the seventy of varicose veins.
Only subjects with varicose veins alone were
included. C)wmparing conditions that alleviate or exacerbate symptoms in patients with and without varicose veins. especially conditions that may influence venous blooâ flow e.g. lying down. leg elevation, first walking, standing, sitang, end of day and menstnial period in wornen. 3. A symptom score (VEINES-SYM) was developed to assess the burden of
symptoms for patients. Analysis of covariance was used to measure the effect of varicose veins on VEINES-SYM adjusted for potential confounders. Qualitv of life 1. Quality of life was measured using standardised generic (SF-36) and disease-
specific (VEINES-QOL) patient-reported quality of life questionnaires. Genderspecific age standardised scores were reported in each clinical category and in al1 patients with varicose veins. 2. Analysis of covariance was used to rneasure the effect of varicose veins on the
SF-36 Physical Component Summary score and the VEINES-QOL taking into account the effect of potential confounders.
11.3.2 Study population
Of the 1531 subjects included in the VEINES study, 259 did not have objective signs of venous disease or had telangiectasia alone, 1054 had varicose veins and 218 had oedema, skin changes or healedlopen ulcer without varicose veins. The sample size used in this chapter varies according to the type of analyses. The total sample size of 1531 was used in the analyses of covariance of VEINES-SYM, VEINES-QOL and SF-36 ?CS because it allows a test of the interaction between varicose veins and
signs of more severe venous disease. A sample size of 1313 (259+1054) was used in the other analyses where the patient was the statistical unit. A total of 1252 legs
from 626 patients without any sign other than telangiectasia or varicose veins were used in analyses based on the leg as the statistical unit. The analysis of the association of symptom prevalence and severity of varicose veins was based on 734 legs of 379 subjects with varicose veins alone. The matched analysis involved 145 subjects with unilateral varicose veins and no signs of more severe venous disease.
113.3 Measurement of the severity of varicose vein
The severity of varicose veins was measured differently by specialists (SP) and by general practitioners (GP).
SP-based assessment of severity relied on clinical
judgement measured on a Cpoint scale: rnild. moderate, severe, very severe (the 'severe' and 'very severe' categories of the questionnaire were merged at the analysis phase). GPs used a slide calliper calibrateci to 0.1 mm to rneasure the maximum diameter of the largest varicose vein (Cornu-Thénard et al., 1991). Severity was defined by four categories represented by the four quartiles of the distribution of al1 measurernents. All subjects with venous oederna, skin changes and healed or open ulcer were excluded from this analysis.
1 13.4 Measurement of symptoms i) Source of data
The syrnptoms measured were based on venous disease-related complaints reported in published epidemiological surveys and clinical trials and expert opinion about symptoms experienced by patients in clinical practice. Symptom data were collected by both patients and physicians. Patient selfassessrnent was based on a self-administered quality of life questionnaire (VEINES-SYM), whereas physician assessments were done at the time of the
clinical examination.
Only data gathered at the first clinical examination are
included here. Physicians rated the presence or absence. during the last 30 days, of nine symptoms elicited by closed questions : hoaviness, pain or aching, swelling sensation, night cramps. hot or buming sensation, restless feeling, shooting or throbbing pain, itchiness and pins and needles. Response categories for the evaluation of the frequency of symptoms by patients were O=never, i=less than once a week, 2-about once a week. 3-several times a week, 4=every day. Of the nine symptoms, heavy legs, achinglpainful legs and swelling sensation were
defined as 'primarv svmptoms' as they were considered on the basis of expert opinion to be the symptoms rnost typical of venous disease.
The other six
symptoms were defined as 'secondant svm~toms'as they were thought to be
associated with venous disorden but are not generally characteristic of these disorciers. Patient-based and physician-based reports differed in several aspects : In the physician-based questionnaire. symptoms were measured for each leg, whereas they were measured globally in the patient-based questionnaire. a
In the physician-based questionnaire, information was also recorded on whether symptorns were alleviated when lying down. with leg elevation. when first beginning to walk, in cold weather, and with support stockings or elastic bandages. or aggravated with prolonged standing, prolonged sitting, at the end of day, in hot weather or during the pre-menstnial period. This information was recorded globally because the conditions investigated could generally not apply to one leg alone. The list of these conditions was established in a consensus meeting with several experts in the field. The patient questionnaire elicited information about the frequency of these nine symptoms.
ii) Development of the symptom score
In order to surnmarise the overall burden of symptorns for the patient, a global patient-based symptom score (VEINES-SYM) was developed by Lamping and Schroter (1998) based on the data collected in the VEINES study. The VEINES-SYM score is the surn of the frequency weights for the nine symptoms plus a tenth item on the intenstty of k g pain during the last 30 days scored on a 6 pt s a l e : O=none, 1= very mild, 2=mild. 3=moderate, 4=severe. 5=very severe. The score obtained using the ten items was translated to a range from O to 100 using an algorithm briefly described below for the VEINES-QOL. The VEINES-SYM was shown to be psychornetrically sound (Lamping and Schroter. 1998).
11.3.5 Measurement of quaiity of lifc
Generic measures i) Source cf date
The generic health-related quality of life measure used in the VEINES study was the MOS Short Fonn Health Survey (SF-36; Ware. 1993; Ware et al.. 1994). It is a self-
adrninistered questionnaire that includes 36 items measuring eight health dimensions in three areas. The areas and dimensions covered by the SF-36 are presented in
Table 11.1. Table 1l. 1 Areas and dimensions covered by the SF-36
Dimansion Functional status
Well-being
Overall evaluation of health
Physical functioning Social functioning Role limitations due to physical problems Role limitations due to emotional problems Mental health Vitality/energy Bodiiy pain General health perception Health change
Number of questions
Several features of the SF-36 have made 1 the current gold standard in generic measures of quality of life. lt has undergone extensive psychometric testing and provides good evidence for reliability, validity and responsiveness (Ware et al., 1994). It has been translated and revalidated for use in several languages (Ware et al., l995), and has been adopted worldwide.
This widespread use allows
comparisons to be made between patients with different diseases and conditions. For the VEINES study, the following venions were used : French (Belgium), French (France), French (Canadian), English (Canadian) and Italian. ii) Summary scores
The SF-36 is scored to produce eight scale scores (corresponding to each dimension) and two sumrnary scores. the Physical (PCS) and Mental (MCS) Component Summary scores (Ware. 1993; Ware et al., 1994). Higher scores indicate a better health status. The PCS and MCS summary scores allow direct comparison with normative values based on the population of studies in the UnitedStates, United Kingdom and other countries. The computation of SF-36 scale and component scores was perfomied by S. Schroter and D. Lamping at the Heath Services Research Unit, London School of Hygiene and Tropical Medicine, London, UK. Disease-s~ecificciualitv of life i) Source of data
Disease-specific quality of life was assessed with a disease-specific scale (VEINESQOL questionnaire) developed dufing the VEINES study. The development and comprehensive psychometric evaluation of the questionnaire was conducted according to standard procedures and criteria (Lamping and Schroter, 1998). The content of the scale was developed by a 7-member multidisciplinary group of clinicians and methodologists. Three sources of information were used : i) an extensive review of the literature (Abenhaim et al., 1999; Lamping, 1997); ii) a review of existing scales;
iii) expert clinical opinion regarding the problems comrnonly
reported by patients during consultations for chronic venous disorders of the leg. The VElNESQOL includw the information arscssed in VElNESSYM (Table 11.2). After pre-testing, four language versions were developed wlh translations perfonned according to standard translation-back translation procedures, including English. French (Belgian and French), French (Canadian) and
ltalian versions.
Responsiveness was assessed in a sub-sample of the VEINES population. Testretest reliability was evaluated in a sub-sarnple of 135 patients in the cohort. Results of psychometrîc evaluation based on field testing in al1 countries confirmed the acceptability, reliability, validity and responsiveness of the VEINES-00 WEINESSYM (Lamping 8 Schroter, 1998).
Table 11.2 Summary Scales and Items of the VEINES-QOUVEINES-SYM (adapted from Lamping and Schroter, t 998) )utcorne measured1Abbreviated Questionnaire Item Content resmnse categories are not included) bquency of symptoms teavy legs khing legs Swelling Wght cramps 4eat or burning sensation qestiess legs ïhrobbing tching iÏngling sensation Global nting of change in k g problern ovar 1 year Sompareâ to one year aga, how would you rate your teg problem n general now ? Phyakal limitations cauaeâ by log problem Daily activities at work Daily activities at home Social or leisure activities in which you are standing for long periods Social or leisure activities in which you are sitting for long periods Role limitations due to log problems Cut down the arnount of time you spent on work or other activities Accomplished less than you would have liked Were lirnited in kind of work or other activities Had difficulty perforrning work or other activities Interferencewith normal social activities Intensity of symptomr lntensity of leg pain Psychological impact Have you felt concerned about the appearance of your leg(s) ? Have you felt imtable ? Have you felt a burden ta family and fnends ? Have you been wonied about burnping into things ? Has the appearance of your leg(s) influenced your choice 01 clothing ? Frequency of symptorns (same symptoms as for VEINES-QOL) Intensity (ntensity of k g pain
ii) Cornputation of scores
The algorithm for computing the VEINES-QOL and VEINES-SYM summary scores
was based on the algorithm used for computing SF-36 summary scores. It includes three steps (Lamping and Schroter, 1998) : i) transformation of raw scores to zscores; ii) computation of a mean z-score, accomplished by dividing the sum of the zscores by the total number of items with valid information; iii) transformation of
z-
scores to T-scores. based on a mean of 50 and a standard deviation of 10. A questionnaire was included in the computation of a summary score if it contains 50% or less of missing values.
113.6 Data analysis
Descri~tiveanalvses Percentages were used to report prevalence data. Descriptive analyses for VEINESSYM, VEINES-QOL and the summary scores computed from the SF-36 questionnaire include means. medians. standard deviations. standard errors and the ranges of values. Descriptive data are reported in six categories defined by the presence of varicose veins and other signs of venous disorders : no signs or telangiectasia alone. varicose veins alone. varicose veins with oedema, varicose veins with skin changes. varicose veins with healed ulcer. varicose ~ e i n swith open ulcer. Stratification and standardisation Prevalence of symptoms and mean quality-of-life were parameten which needed to be described at the level of the source population. Their report at the level of the
study population was not scientifically interesting because this population was not meaningful as a basis of inferences. According to the protocol. the VEINES study population was biased since male subjects and subjects with ulcer were intentionally over-sampled. In addition, investigators did not entirely comply with the sampling
procedure, resulting in age differences between the source population of registered patients and the VEINES study population. Due to gender differences in the prevalence of symptoms and in qualitysf-life scores. the estimates for these parameten were stratified by gender. The objectives of this study also required calculation of means and proportions in each category of varicose veins.
Standardisation according to age was also necessary following
reports of an increasing prevalence of venous-related symptoms with age (Bradbury et a1..1999). A standardisation method was consequently used to estimate in each category of
varicose veins the proportions of symptoms and the mean scores that would have been observed in a hypothetical cohort whose age distribution in that category followed the distribution of the source population. In addlion. it was also considered useful to estimate the age-standardised prevalence of symptoms and quality of life scores in al1 patients with varicose veins presenting to physicians. The standardisation process thus implied estimation of the distribution of the categories of varicose veins in the source population.
However, a physical
examination was not required to report the patients' clinical status at the registration phase. and the result of the clinical examination performed at the baseline examination sometimes disagreed with the clinical status recorded at registration. Moreover. registration data of the GP sample did not allow the computation of the same varicose vein categories as in the study population since varicosities and oedema were not registered. In order to overcome this problem, standardisation was performed in four steps. In a first step. a classification of clinical signs at the registration phase was created based on the available clinical information. Since different numben of signs had been collected in the specialist-based and the GP-based samples. this classification had 6 and 5 categories respectively. The classification for specialists was identical ta the 'VEINES' classification presented in Chapter 6. whereas the classification for
generalists was slightly modified to take into account the lack of information on telangiedasia and oedema. In the second step. the 1467 subjects for whom data of the registration phase and data of the baseline examination could be matched were stratified according to the above-mention4 dinical categories computed on registration data and to varicose veins categones of the baseline examination.
This step allowed to compute a
probability distribution of the categones of the 'correct' classification in each clinical category of the registration phase. Using the age distributions (in two categories : 18-59 and 60-75 years) obsetved in the clinical categories computed for all
registered subjects, the estimated age distribution in each varicose vein category was derived from the source populat~on. This process assumed that errors in the clinical diagnosis made at the registration phase for the 1467 subjects reflected those existing in al1 registered patients. This procedure was performed separately for men and wornen. In the third step, weighted averages of proportions and means were calculated within each clinical category (Rothman and Greenland. 1998, p. 45).
The standard
distribution was the expected numben of subjects in the two age categories. while
the proportions and the means came from the study population. The fallowing formula was used :
where: P, k pk Nk
= age-standardised proportion in specific clinical category = each age category = crude proportion in specific age category (from study population) = expected number of subjects in each age category (standard derived from the source population)
Standardised numbers were also calculated in each category by applying the standardised proportions and means to the sample size of that category.
In the fourth step, age-standardised prevalence rates and means were calculated for
al1 patients with varicose veins using the corrected distribution of varicose vein categories of the source population (Le. varicose veins alone, vaficose veins with oedema, varicose vein with skin changes and varicose veins with ulcer) as the set of weights.
Hy~othesistestina and estimation A two-sided statistical test for the comparison of two proportions (Armitage and Berry, 1987, p. 123) was used to compare the proportions of individual symptoms in categories 'No signs or telangiectasia alone' and 'Varicose veins alone', to compare
the proportions of individuals in whorn symptoms were alleviated or exacerbated in these two categories, and to compare the prevalence of symptoms experienced by men and women with varicose veins. The nuIl hypothesis specified that the two proportions were equal; the type I error was set at 5%. A chi-square for trends was used to assess whether the proportion of legs in whom a
primary symptom was reported was compatible with a linear trend according to the severity of varicose veins (Amiitage and Berry, 1987, p. 372). A two-sided p-value of
0.05 on the chi-square distribution was used to indicate departure from the nuIl hypothesis of absence of a linear trend. A valid estimation of the association of the presence of symptoms associated with
varicose veins required adjustrnent for potentially confounding variables, such as gender, age, CO-morbidityand other unmeasured patient characteristics. For this purpose, the principle of individually matched series was applied to the comparison
of symptom prevalence in legs with and without varicose veins (Miettinen, 1985, p. 149; Moses, 1992). All 145 subjects with unilateral varicose veins (and without other
signs of venous disorder) were selected and pairs of legs were considered in a matched analysis. The Mantel-Haenszel point and variance estimates for one-tosne matched data were used to estimate the prevalence ratio and the 95% confidence interval around it (Rothman and Greenland, 1998. p. 284). The 95% confidence
interval for difference between two proportions based on paired sample was also computed (Amitage and Berry, 1987, p. 121). Only primary symptoms were considered in this analysis (heaviness, pain, swelling sensation and any of these) Estimation of adiusted effects Analysis of covariance (Anderson et al., 1980, p.140;Kleinbaum et al.. 1988, p.297) was used to compare mean symptorn and quality of life scores in patients with and without varicose veins, adjusted for the effect of potential confounders.
Three
summary scores were used in this analysis : SF-36 ?CS, VEINES-SYM and VEINES-QOL (MCS was not associated with varicose veins in univariate analyses).
The adjusted 'effect' was used as the primary outcome of interest and estimated from the adjusted mean difference (indicated &y A) between subjects with and without the exposed level of a variable. The 9S0h confidence interval around the adjusted effect was used to estirnate its precision. Standard techniques of regression diagnostics were applied to check the appropriateness of the model (Kleinbaum et al.. 1988, p.181).
In order to test the assumptions of the model, residuals (standardised
residuals, studentized residuals and jackknife residuals) were plotted in univariate analysis and in bivariate analyses against predicted values. Predicted values were plotted against observed values. Jackknife residuals, leverages and Cook's distance
were computed and plotted in order to detect outliers and rneasure their influence. The following critical values for a type I error of 0.10 were identifieci from reference tables (Kleinbaum et al.. 1988, p. 661) in order to detect an outlier of significant influence : jackknife residual : 3.86; leverage: 0.122; Cook's distance : 0.97.
Collinearity was assessed by computing the condition number (CN) in a principal component analysis. It was calculated as the square root of the ratio of the largest
and the srnallest eigenvalues. A CN 2 30 was used to reflect the presence of moderate to severe collinearity (Kleinbaum et al., 1988. p.213). Interaction ternis were defined by associating varicose veins and each of the following signs : oedema. skin change, healed or open ulcer. Each interaction terni
was tested one at a time in a model containing varicose veins and the other variable present in the interaction term. The criterion adopted to introduce an interaction term 0 the F test. Since none of the interaction terms into the model was a p-value ~ 0 . 1 at
were retained, no further analyses were requested. As the main objective was to assess the effect of vancose veins on symptoms and quality of life, the selection of variables to be included in the models was based on pnor knowledge of their potential confounding effect. All potential confounders were forced into the model and none of them were rernoved on the basis of statistical criteria. Except for two variables (country of origin and type of physician). the effects of al1 variables in the model are documented for descriptive purpose. All variables
were categorised as 011 variables.
When a variable contained more than two
categories (e.g. age). indicator variables were created for each non-referent category of the variable.
The following variables were included in the model : 0
varicose veins (exposure of interest): 2 categories : absent (O) or present (1) at
the baseline clinical examination. telangiectasia : 011. This variable was only used in models with VEINES-QOL as
the dependent variable because it was felt that telangiectasia would not affect symptoms (VEINES-SYM) or physical functioning (PCS). venous oedema : 011.
skin changes : 011. 0
healed or open ulcer : 011.
a
gender : malelfemale.
a
age : 5 categor~es(18-39.4049. 50-59.60-69. 70-75 years old);
O
body mass index : 4 categories (< 25.0, 25.0-26.9, 27.0-29.9, 2 30.0); three
categories have been used in some analyses : c 27.0, 27.0-29.9. r 30.0.
educational level : 3 categories : low (lower secondary school or less). moderate (high secondary school), high (university or college).
physicirn-nported conioibidity : 2 categories : Ozabsence of any major morbidw (including diabetes, hypertension. cardiac disease, renal insufficiency, chronic rsspiratory disease, osteoarthritis, hepatitis, hypothyroidism, peripheral arterial insufficiency of the lower legs. lymphoedema); 1 = presence of any of the above.
country: 4 categories : Belgium. France. Italy, Quebec. type of physician : 2 categories : general practitioner or specialist.
Sensitivitv analyses Given the low specificity of the diagnosis of varicose veins by general practitioners discussed in Chapter 7, analysis of covariance was carried out with the sub-sample of patients recruited by specialists. In these analysis, age was divided into two categories (18-59 and 60-75 yean old) and variables 'country' and 'type of physician' were not included in the model.
11.4 Results 11.4.1 Symptoms
1) Aae standardised prevalence of svm~tomsbv clinical cateaorv
The age-standardised distributions of primary and secondary symptoms are reported separately for men and women in Tables 11.3 and 11.4. The standardised estirnates for al1 subjects with vaflcose veins are presented in Table 11S. A high prevalence of symptoms, particularly heaviness and pain, was found in al1
clinical categories and in both genden. In both men and women, more than 75% of individuals presented with at least one of the primary symptoms.
For most
symptoms, prevalence increased with the severity of signs associated with varicose veins, except in the category defined by active ulcer. Many observations were consistent with clinical experience, such as a high prevalence of swelling in subjects
with oedema and a high prevalence of itching in subjects with skin changes, which often includes euema along veins. In men, no consistent pattern was found in difFerences obsewed between categories 'No signs or telangiectasia atone' and 'Vancose veins alone', and none of the differences reach statistical significance at the 5% level.
The difference in the
prevalence of symptoms between women with and without varicose veins is statistically significant at the 5% level for al1 primary symptoms, but none of the secondary symptoms were associated with the presence of varicose veins. Women with varicose veins were more likely than men to present primary symptoms, but the difference was statistically significant only for pain (Table 11S).
Table 11.3 Age-standardised numben and proportions of male patients with symptoms (last 30 days), by dinical category (W=varicoseveins) Patientreported symptom
No signs or telangiectasia (n=23) n %
Prirnarv svmotoms Heaviness 13 Pain 15
Swelling At least one
9 17
Secondam svmotoms Cramps 11 Burning 6 Restlessness 17 Shooting 8 ltching 12 Pins and needles 14
Varicose veins alone (n=90) n %
W and
oedema (n=21) n %
W and skin changes (n=120) n %
W and healed ulcer (n=31) n %
W and active ulcer (n=15) n %
Table 11.4 Age-standardised numôers and proportions of fernale patients with symptoms (last 30 days), by clinical category (VV=varicose veins)
r
-
Patientreported sy mptom
Heaviness Pain Swelling ' At least one
-- -
No signs or telangiectasia (n=236) n %
Varicose veins alone (n=277) n %
W and oedema (n=104) n %
W and skin changes (n=311) n %
W and healed ulcer (n=69) n %
W and active ulcer (n=l6) n %
152 133 85 180
Secondarv svmptoms Crarnps 90 Burning 98 Restlessness 115 Shooting 85 ltching 101 Pins and needles 61 pc0.05 for the test of a difference of the proportions of subjects with the specific symptorn between categories 'No signs or telangiectasia' and 'Varicose veins atone'
Table 11.5 Gender-specific, age standardised numbers and proportions of patients with symptorns (last 30 days) among al1 patients with varicose veins Patient- reported syrnptom
Males (n=277)
Females (n=777)
Primaw svmPtoms Heavîness Pain ' Swelling At least one of the above
Crarnps Buming sensation ' Restiess legs Shooting ltching ' Pins and needles l
~ ~ 0 . for 0 5the test of a difference of the proportions of males and femaîes with the specific symptom.
2) Do varicose veins have an impact on svm~toms? a- Paired analysis of subjects with unilateral varicose veins
In the matched-pair analysis, a prevalence difference of 20% was found for heaviness and pain and the highest prevalence ratio found was 1.67 for pain (Table 11.6). All confidence intervals excluded the nuIl value.
Table 11-6 One-to-one matched analysis of physician-rated sy mptom occurrence in 145 individuals with unilateral varicose veins (W)and no signs of more severe venous disorder
N discordant pairs
Prevalence with W
Heaviness
34
0.63
0.43
0.20
O. 13-0.29
1.48
1.28-1.73
Pain
35
O. 50
0.30
0.20
O. 12-0.28
1.67
1.36-2.06
Swelling
14
0.30
0.21
0.09
0.03-0.13
1.38
1.13-1.69
At least one
36
0.77
0.54
0.23
O. 15-0.32
1.43
1.27-1-63
Physicianrated sy mptom
Prevalence without W
Prevalence difference (PD) PD 95% CI
Prevalence ratio (PR) PR 95% CI
It is noteworthy that the prevalence of heaviness, aching, swelling and any primary symptom in the legs without varicose veins (43%. 30%, 21% and 54% respectively) was lower than that found in the 518 legs of 259 subjects free of varicose veins on both legs (prevalence of 53.5%, 43.1%, 33.6% and 71.4%). This suggests that the presence of symptoms in one k g is not increased when symptoms are present on the contralateral kg. b- Association between prevalence of symptoms and sevedty of varieose veins
Ratings of severity were available for 585 of the 589 legs with varicose veins alone,
In the specialist-based study, the varicose veins were assessed as mild in 44% of the There was no consistent trend in legs and moderate in 47% of the legs (Table 117). the distribution of symptoms across severity of categories, and the nuIl hypothesis of
absence of a trend between the 'mild' and the 'severe' categories could not be rejected ( ~ ~ 0 . 0 5A) . similar conclusion could be drawn from the analyses based on the GP-based sample (Table 11.8): none of the three primary symptom were
associated with the size of varicose veins in individuals with varicose veins alone. Table 11.7 Association between specialist-reported severity of varicose veins (VV) and prevalence of symptoms in the sarne leg in subjects with varicose veins alone
Physician-rated symptom
Heaviness Painlache Swelfing Any of the above
Varicose vein severity (number of legs) Miid Moderate Severe (n=83) (n=89) (n='i8) n % n % n % 54 65.1 41 49.4 36 43.4 68 81.9
65 73.0 47 52.8 24 27.0 72 80.9
9 50.0 9 50.0 4 22.2 12 66.7
Table 11.8 Association between general practitioner-reportcdseverity of varicose veins (VV). measured in puartiles of the maximum diameter of the largest vein. and prevalence of symptoms in Me same leg in subjects with varicose veins alone Varicose veins severity in quartiles of maximum diameter (number of legs) Physician-rated sy mptom
Heaviness Paidache Swelling Anyoftheabove
I 5 4.3 mm (n=104) n % 64 49 37 77
61.5 47.1 35.6 74.0
II 4.4-6.8 mm (n=98) n % 62 45 32 77
63.3 47.1 32.7 78.6
111 6.9-9.6 mm (n=96) n O h 44
39 27 66
45.8 40.6 28.1 68.8
IV 2 9.7 mm
(n=97) n % 57 46 19 69
58.8 47.4 19.6 71.1
c- Conditions alleviating or exacerbating symptoms Conditions which alleviate or exacerbate physician-rated symptoms in subjects with and without varicose veins are presented in Table 11.9. The table presents the number and percentage of patients experiencing an alleviating or an exacerbating condition among those presenting with at least one of the three primary symptorns (heavy legs. leg pain and swelling). In about 50 to 70% of subjects. symptorns were improved by conditions that favour venous retum (lying down, leg elevation, first walking), but no statistically significant difference was found between the two clinical categories. Among conditions known to generally exacerbate symptoms, results suggest that the impact of menstrual pefiod on primary symptoms experienced by women with varicose veins is less than that reported for women without varicose veins.
Table 11.9 Conditions which alleviate or exacerbate physician-reportedsymptoms in 477 subjects presenting at least one of the primary syrnptoms (heaviness, kg pain, swelling) on any le9
No signs or telangiecîasia (n=191) n % Alleviatincr conditions Lying down 116 60.7 Leg elevation 124 65.2 92 48.4 First walking 106 55.8
Cold weather
Exacerbatino conditions Standing 144 Sitting 100 End of day 147 Hot weather 46 58 Menstrual period
75,8 53.4 77.4
24.2 33.3
Varicose veins alone (n=286) n %
187 199 149 157
65.6 70.3 52.3 54.9
218 163 231 56 59
76,5 57.0 81.3 19.7 26.6
3) VEINES-SYM a) Distribution of VEINES-SYM scores The distribution of VEINES-SYM scores in the VEINES study population is presented
in Table 1 1.10. Missing values for 111 subjects reduced the sample site to 1422 subjects. Table 11.10 Patient-reported disease-specific quality of life: crude descriptive statistics in study population
1
I
Variable Venous-related symptom score WEI NES-SYMl
N 1422
Mean 49.9
Median 50.6
Std. dev. 6.7
Std.
enor
0.2
Minimum 31.5
Max-( murn 60.7
1
1
The distribution of VEINES-SYM scores by clinical category (Table 11.l1) shows decreasing scores with seventy of the associated signs. The main scores of patients with varicose veins alone are similar (and non statistically different at the 5% level) to those of patients without signs or with telangiectasia only. Age-standardised mean scores by gender are shown in Table 11.12. The standardised mean for al1 subjects with varicose veins was 50.3 in 262 men and 49.1 in 720 women. Table 11.11 Crude patient-reported venous-related symptoms score in study population as measured by VEINES-SYM. in different clinical categories '
Variable
No signs or telangiectasia (n=259)
Varicose veins alone (n=367)
Vancose veins and oedema (n=125)
Varicose veins and skin changes (n431)
Varicose veins and healed ulcer (n=i00)
Varicose veins and active ulcer (n=31)
n mean(s.e.) range
242 52.4(0.4) 35.4-60.7
347 51.3(0.3) 33.1-60.7
114 48.0(0.6) 33.7-60.0
398 48.9(0.3) 31.5-60.7
96 47.1 (0.7) 33.1-60.0
27 46.9 (1.3) 32.4-58.5
-- - -
-
-
' p< 0.05 for the test of a difference between the means of categories aNo signs or telangiectasia aione~and avariase veins al on el^ (Student's t-test)
Table 11.12 Gender-specific, age-standardised patient-reportecl mean VEINES-SYM score -
-
No signs or telangiectasia
Varicose veins alone
Vancose veins and oedema
Varicose veins and skin changes
Varicose veins and healed ulcer
Vancose veins and active ulcer
Men
52.2
52.3
48.9
49.4
48.8
50.3
Women
52.1
50.9
47.6
48.7
45.9
45.4
Ï
b) Multivariate analyses of characteristics associated with VEINES-SYM scores
The crude mean values of VEINES-SYM by subgroups are presented in Table 11.13.
The crude difference is large for almost al1 the variables tested in the model. In multivariate analysis, the largest difference (and the lowest VEINES-SYM score) was seen for a body mass index higher than 30 kg/m2,with a A of - 2.2 units (95% confidence interval: -3.1, -1.4) in comparison to a BMI < 27.0. Venous oedema (A:
-
1.9; 95% CI : -2.8, -1.1), healed or open ulcer (A: -1.1; 95% CI : -1.9, -0.4) and
presence of a serious co-morbid condition (A: -1.5; 95% CI : -2.5, -0.4) also had a large impact on the VEINES-SYM, with confidence intervals not including the nuIl value. Regression diagnostics (residual analyses, detection of collinearity, detection
of outliers) did not detect any particular problem. Table 11-13 Effect of patient characteristics and clinical status on VEINES-SYM (n=1422) Sample Variable
sire
Unadjusted Unadjusted mean effect
Adjusted mean '
Adjusted effect ' estimate 95% CI
Varicose veins
no Venous oedema no Skin chancies no Healed/open ulcer no Yes Gender male
female
Body mass index
~25.0 25.0-26.9 27.0-29.9 230 unknown Educational level low intemediate high unknown Co-morbidity no Ye* l
Adjusted for al1 other variables included in the table. and for tvpe of physician and country
In contrast to bivariate analysis, these results indicate a lack of association between varicose veins and VEINES-SYM scores when other predictors of venous diseaserelated symptoms are considered. Similar result is seen for al1 age categories and for educational level, where 95% confidence inteivals around the effect of each category in cornparison to the reference category includes the nuIl value. Women tend to have more symptoms than males after adjusting for the signs presented. 4) Summarv of results on svmtoms
A high prevalence of symptoms, particularly heaviness and pain. was found in al1
clinical categories and in both men and women. Specific symptorns for varicose veins were not found. For most symptoms, prevalence increased with the severity of signs associated with varicose veins. Swelling was specific to oedema and itching seemed to be specific to skin changes in men. Women presented primary symptoms more frequently than men.
Primary
symptoms are reported more frequently by women with varicose veins than by women without varicose veins. In matched paired analysis comparing legs with and without varicose veins, the largest prevalence differences (20%) and ratios (1.67) were found for pain. Prevalence of symptoms was not associated with the size of varicose veins. There were no differences among symptomatic subjects with or without varicose veins in tems of the alleviation or aggravation of syrnptoms by conditions affecting the venous blood fiow. Varicose veins had no effect on VEINES-SYM after adjusting for other suspected determinants of symptoms.
11J.2 Quality of life
1) Generic aualitv of life
a) Distribution of SF-36 scores
Descriptive statistics for the SF36 dimension and summary scores for the VEINES population are presented in Table 11.14. Higher scores indicate
better quality of life. Let us note that the scores for the eight scales are not standardised but that they have been transfonned to represent the percentage of the total possible score achieved for that scale, with a range of O to 100. The two summary scores have been standardised to the US
population. They have an expected mean (standard deviation) of 50.0 (10.0). Table 11.14 Crude descriptive statistics of patient-reportedgeneric quality -of life rneasured by the SF-36 in the VEINES study population
I
Variable Bodily pain Emotionai role iimit General health Mental health Physical functioning Physical role limit Social functioning Vitaiity Mental sumrnary score ~ h y s i c asurnmary ~ score
N
Mean
Median
Std. dev.
Std. erm
Minimum
Maximum
Table 11.14 shows that, on average, subjects with venous-related leg complaints included in this study experienced lower scores for mental (meant45.3) and physical (mean=45.3) quality of life ?han the general US population.
These scores are
comparable for US patients with low back pain (Ware et al., 1994). The distribution of SF-36 summary scores by cfinical category (Table 11.15) suggests that PCS scores decrease with varicose veins and with the severity of associated signs. In contrast. the MCS scores are similar in al1 categories. A test for
the comparison of mean MCS scores in patients with varicose veins alone and subjects without signs or with telangiectasia only does not allow rejection of the nuIl hypothesis of sirnilar scores both categories.
The age-standardised mean scores by gender are shown in Table 11.16. The standardised mean scores for al1 subjects with varicose veins were 46.1 and 45.6 for MCS and PCS respectively for 242 men; the MCS and PCS scores were 43.6 and 44.2 respectively for women. The trend of PCS scores is mainly observed in women.
with scores ranging from 48.3 to 35.5. Table 11.15 Cnide patient-related generic quality of life as measured by SF-36, by varicose vein CategOrY No signs or telangiectasia Mental summaw score 224 n rnean (s.e.) 43.2 ( 0.7) range 12.3-61.9 Phvsicai summaw score ' n 224 rnean(s.e.) 49.9(0.6) range 13.8-66.9
Vaficose veins aione
Varicose veins and oederna
Varicose veins and skin changes
Varicose veins and healed ulcer
Varicose veins and active ulcer
329 44.3 (0.6) 13.4-64.3
107 43.1 (1.1) 14.5-65.0
360 44.7 (0.6) 8.8-66.0
90 42.9 (1.3) 15.1-65.4
25
329 47.1(0.5) 19.8-65.0
107 43.7(0.9) 19.2-63.4
360 43.3(0.5) 14.8-64.2
90 40.1 (1.0) 18.4-61.2
25 36.4 (1.8) 18.5-53.5
-
48.0 (2.2) 25.3-63.2
' pc0.05 for the test of a difference between the means of categories aNo signs or telangiectasia alone~and ~Varicoseveins alone n (Student's t-test)
Table 10.16 Gender-specific, agestandardised SF-36 Mental Cornponent Score (MCS) and Physical Cornponent Score (PCS) in registered patients
Men -
n
MCS mean PCS mean Women n MCS mean PCS mean
No signs or telangiectasia
Varicose veins alone
Varicose veins and oedema
Varicose veins and skin changes
Varicose veins and healed ulcer
Varicose veins and active ulcer
19 47.2 45.6
80 46.2 47.6
18 46.8 48.0
106 45.2 45.4
27 46.2 42.9
11 53.8 37.3
205 43.0 48.3
249 43.9 46.5
89 41.8 43.6
254 44.5 43.6
63 40.9 39.4
14 43.4 35.5
b) Multivariate analyses of characteristics associated with physical QOL (PCS)
Table 11.17 displays results of analysis of several variables on physical QOL. As bivariate analysis did not find an effect of varicose veins on MCS, only PCS is included in this analysis.
Analyses of collineanty perfomed to examine the
appropriateness of the multivariate regression model did not detect any particular problem. However, for one outlier, the values of jackknife (3.94) and studentized residuals (3.92) overcame the critical value of 3.86. Although values of Cook's distance and leverage for this observation were within normal values, the deletion of the observation was associated with a change of almost al1 regression coefficients by 0.2.
This observation was therefore removed frorn the study population for al1
analyses using PCS. It concerned a 37-year old Belgian man (ID 3502) with telangiectasia only but with a chronic respiratory disease. Observed and predicted physical summary scores for this patient were 13.8 and 46.3 respectively. The low score for the physical summary score was associated with scores of O for vitality and for role function due to physical limitations. Except for gender, alrnost al1 the variables introduced into the model had an important effect on the PCS in univariate analysis. In multivariate analysis. age over 60 years old (A: -3.2 to -4.0), BMI 2-30 (A: 4.4), CO-morbidity(A: -4.0), presence of
ulcer (A: -2.5) and female gendet (A: -1.6)' significantly affected physical component scores (PCS), with 95% confidence intervals not including the nuIl value. The analysis showed that varicose veins have a very small adjusted effect (A: 0.5).
The result of the sensitivity analysis carried-out in patients recruited by specialists
showed an adjusted effect of -0.87 (95% confidence interval : -2.52 to 0.78) for varicose veins.
246 Table 11.1 7 Effect of patients' characteristics on SF-36 PCS (n=1319) -
-
Variable
Sample sire
Unadjusted Unadjusted mean effect
Adjusted
man '
Adjusted effect ' estimate 95% CI
Vadcose veins no Yes Venous oedema no Yes Skin changes no Y es Healed/o~enülcer no Y= Gender male female Acre (y=* 18-39 40-49 50-59 60-69 70-75 Bodv mass index ~25.0 25.0-26.9 27.0-29.9
230 Educational level low intemediate high Co-morbidity no Yes
' Adjusted for al1 other variables included in the table. and for type of physician and country
21 Disease-specific auality of life a- Distribution of VEINES-QOL scom
VEINES-QOL scores were computed for 1436 subjects. Descriptive statistics are presented in Table 11.18.
The crude (Table 11.19) and gender-specific age-
standardised (Table 11-20) VEINES-QOL scores in registered patients suggest that the VEINES-QOL scores decrease with increasing severity of signs associated with varicose veins. The mean score for al1 subjects with varicose veins was estimated at 50.7 in 266 men and 49.0 in 725 women. As for PCS, VEINES-QOL is more dependent on the clinical category in women than in men. Table 11.18 Patient-reported disease-specific quality of life: descriptive statistics in study population Variable Venous-related QOL score (VEINES-QOL)
N
1436
Mean 49.9
Median 50.8
Std. dev. 6.0
Std. enor
Minimurn
Maximum
0.2
29.6
60.3
1
1
Table 11.19 Patient-reporteclvenous-related symptoms and quality-of-life scores as measured by VRLD-QOL, in different clinical categories '
Variable
No signs or telangiectasia (n=259)
Varicose veins alone (n=367)
Varicose veins and oedema (n=125)
Varicase veins and skin changes (n=431)
Varicose veins and healed ulcer (n=100)
Vancose veins and active ulcer (n=31)
VEINES-QOL ' n mean(s.e.) range
243 52.3(0.3) 36.4-59.8
346 51.3(0.3) 33.7-60.3
115 48.5(0.6) 34.0-59.7
404 49.0(0.3) 29.6-59.8
97 46.6(0.6) 32.1-59.5
29 45.4(1.2) 32.9-55.5
' ~ ~ 0 . for 0 5the test of a difference between the means of categofles UNOsigns or telangiectasia alonen and ~Varicoseveins alonen (Student's t-test)
Table 11.20 Gender-specific, patient-reported venous-related symptorns score as measured by VEINES-QOL, standardised for the age distributions in registered patients of the sarne gender
Males Females
No signs or telangiectasia
Varicose veins alone
Varicose veins and oedema
Varicose veins and skin changes
Variccrse veins and healed ulcer
Varicose veins and active uker
51.9 52.0
52.6 50.8
51.5 47.8
50.20 48.7
48.5 45.5
46.9 43.1
1
b) Multivariate analysis of characteristics associated with VEINES-QOL
In the multivariate analysis of the effect associated with venous- and non-
venous-related variables (Table 11.21), the largest effects were observed for obesity (A: -2.2; 95% CI: -3.1, -1.3) and the presence of healedlopen ulcer (A: -1.9; 95% CI: -2.9, -0.9). A difference was also observed due to gender. with women having a mean score that is 1.5 point lower than males. In the sensitivity analysis carried out in specialist patients, the adjusted effect is practically unchanged 4 . 5 9 (95% confidence interval: -1.64 to 0.45). It is noteworthy that variables found to affect SF-36 PCS, such as age, obesity and CO-morbidityhave a much weaker effect on VEINES-QOL.
Venous
oedema affects VEINES-QOL but not SF-36 PCS, which probably reflects the impact of oedema on swelling sensation. In contrast, ulcer affects almost equally the two measures. 3) Summarv of investigations on ~ualitvof life a) Genenc quality of life O
Mean SF-36 MCS and PCS scores in patients with varicose veins were 46.1 and 45.6, respectively, in men. and 43.6 and 44.2, respectively, in women.
O
PCS scores decreased with the presence of varicose veins and with the severity
of associated signs. In contrast. MCS scores were similar in al1 categories. The decrease of PCS scores with the severity of associated signs was mainly seen in women. Multivariate analysis showed that varicose veins have a negligible effect on PCS. Age, BMI r 30. presence of ulcer and female gender had the largest effect on PCS. b) Disease-specific qoality of life
VEINES-QOL scores decreased with increasing severity of signs associated with varicose veins, and this trend was more apparent in women.
Venous oedema, ulcer, gender, BMI 2 30 and CO-morbidityaffected VEINES-QOL, but varicose veins had a negligible effect.
fable 11.21 Effect of patients' characteristics on VEINES-QOL (n=1436) Variable
Sample size
Unadjusted Unadjusted mean effect
Adjusted mean '
Adjusted effect ' estimate 95% CI
daricose veins no
denous oedema no Skin chanqes
no Y es Healed/open ulcer
no Gender male female
Bodv mass index ~25.0 25.0-26.9 27.0-29.9
230 Educational level low intermediate high
l
Adjusted for al1 other variables inchded in the table. for type of physician and for country
f 1.5. Discussion 11.5.1. Prevalence of symptoms
According to expert opinion, lower limb symptoms of heaviness, aching and swelling were consideted as 'primary symptoms'. i.e. symptoms more specifmlly related to venous disorders in clinical pradce.
From our data. they are indeed extremely
common in patients consulting general practitioners and specialists with cornplaints relateâ to venous disorders : 83.0% of men and 86.9% of women with varicose veins mentioned that they suffered from leg heaviness, k g pain or swelling. The prevalence of the three symptoms increased with the severity of signs associated with varicose veins. Aithough they are generally reported less frequently than "primary symptoms", "secondary symptoms" are also observed in high proportions of men and women. Large nurnbers of symptoms were also reporteâ in other studies (Bradbury et al., 1999; Jakobsen, 1979; Sarh et al., 1W 2 a ; Struckman. 1986a; Stnickman, l986b; Stnickman et al., 1986~).For exarnple. Struckman (1986b) reported a prevalence of sensation of swelling in 88% of patients. As signs of severe venous disorders (oedema, hyperpigrnentation, eczema, induration) were also present in large proportions in these studies, it is impossible to ascertain whether the large prevalence of symptoms is specifically associated to varicose veins or to more severe venous disease. The onginality of this study comes from showing that symptoms are also highly prevalent in patients with varicose veins only : at least one primary symptom was obseived in 76% of males and in 84% of females. Interestingly. the prevalence of primary symptoms was also ver- high (30-60 %) in patients without signs of venous disorders or with telangiectasia alone. These figures
are consistent w l h those of the Edinburgh Veins Survey using a sample of subjects from the general population: seFreported symptoms of aching heaviness and swelling sensation were reported respedively by 18%, 22% and 45% of women without varicose veins (Bradbury et al, 1999).
Three reasons may be considered to explain the high prevalence of symptoms in nonaffected individuals or in individuals affected by telangiectasia or variase veins only : i) symptoms could be associated with venous incompetence and there is a large proportion of venous incornpetence in these groups of patients; this hypothesis is supported by the prevalence of reflux found in the Duplex study (25Oh in subjects without signs or with telangiectasia alone, and 93Oh in subjects wlh varicose vein akne); ii) symptoms have a non-venous origin; iii) symptoms were promoted as an attempt to obtain treatment for cosmetic reasons; this hypothesis cannot be verified because cosmetic complaints and appearance of the k g were not investigated in this study.
11.5.2. Impact of varicose veins on symptoms
From the list of symptoms investigated, Wo show some degree of specificity : swelling with oedema and itching with skin changes. Symptoms specific to varicose veins could not be identified. ftom either primary and sewndary groups. From the matched pair analysis by leg. the prevalence difference comparing the prevalence of symptoms with and without varicose veins rangeâ from 9% for swelling to 20% for heaviness and pain, with prevalence ratios of respectively 1.38, 1.48 and
1.67. Although al1 95% confidence intervals excluded the nuIl value, the estimates are small and not consistent with a specific association of symptoms with varicose veins. They could be explained by possible observation and reporting biases : as mentioned above. symptoms may be given as a pretext by patients seeking cosmetic treatment for varicose veins, and symptoms mentioned by patients may be ascribed by the physician to the kg with varicose veins.
In further analysis, prevalence of symptoms was not associated with the seventy of varicose veins ascertained by general practitionen or specialists, and there was no
statistical difference between the proportions of patients with and without varicose veins in whom symptoms were alleviatecl by conditions operathg on the venous retum. These Cndings confimi the lack of association behnreen symptoms and varicose veins. Effect of varicose veins on svrn~tomscore (VEINES-SYM) In cnide (non standardised) analyses, the symptom score showed a declining trend with the seventy of signs associated with varicose veins, from a score of 51.1 in
subjects with varicose veins alone to 46.9 in subjects with varicose veins and ulcer. However, no statistically significant difference was found between the score found in subjects with varicose veins alone and subjects without signs or with telangiectasia alone (52.4). After standardisation for age. the score was lower by 1.2 in women. An effect of -1.7 on VEINES-SYM was observed in bivariate analysis of covariance perfomed on the total study population. but this effect disappeared after adjustment for other predictors of symptorns. Venous oedema, ulcer, body mass index and
CO-
morbidrty were the variables with the strongest effect on the score. These results conoborate the lack of impact of varicose veins on symptoms as well as on their frequency.
11.5.3. Impact of varicose veins on quality of life
Generic ~ualitvof Me Two summary scores were computed from the SF-36 sale : the Physical Component Score (PCS) and the Mental Component Saxe (MCS). The first important finding of this study was the lack of effect of varicose veins and other signs of venous disorden on MCS, irrespective of gender and type of analysis (crude or age-standardised). It has been hypothesised that the reason why so many persons se& treatrnent for varicose veins is related to concerns about the appearance of their legs (Robbins et al.. 1992; Ganat et al.. 1993). The difference of 1.O between age-standardised scores for mental quaMy of life found in patients with W alone and in patients without signs or with telangiectasia alone is not statisticalfy significant, and the hypothesis is therefore
@
not substantiated by the results of this study. Mental status was not affected by venous ploblems Ming long-standing disorders either, but th6 situation is not specfic to varicose veins : the population healh survey in Lothian region (Southeast Scotland) showed that limiting long terni illness was strongly associated with SF 36 physical limitation scores and much less associated with scores on scales of mental and social well-being (Cohen et al., 1995). There is no doubt that the mean MCS (44.0) is lower in the population examined in this study than in the general US population (50.0). Variables associated with this low MCS were not investigated because this was not an objective of the thesis and few variables relating to mental status were measured in the study . The second important finding is the lack of association between varicose veins and the Physical Summary Score in multivariate analysis alhough a strong effect was found in bivariate analysis. especially in women. Results presented in table 11-17 show that venous oedema, healedlopen ulcer. female gender, high age, obesity and co-rnorbidity are variables associated with a low PCS. The low effect of varicose veins found in this study (-0.5) is in contrast to the results of Baker et al. (1995) who reported that all SF-36 dimensions except social function and health perception were improved 6 months post operation for varicose veins (?CS and MCS were not available at the time of their study) . However, they did not consider the possibilw that at hast part of the improvement could be explained by placebo effect. An adjusted effect of -0.5 can be compared with the normative sample of 2474 people in the U.S. general population (Ware. 1994). In comparison to subjects without varicose veins, subjects with varicose veins would have on average an estimated 1.5% increase of limitation in climbing stain, a 3.0% increased difficulty at work and a 1.5% restriction of work time. Other indicators of limitations would change by 1% or less.
Disease-s~ecificaualitv of life NEINES-QOL) The analysis of VEINES-QOL demonstrates the specificity of this measure for venous disorden. There was a clear decrease of the score with the presence of varicose veins alone when compared to the category of patients without signs or with telangiectasia alone; there was a further decrease with the severity of the signs associated with varicose veins. This result was more apparent in women than in men. In multivariate analysis, however, varicose veins had no effect on quality of life. In this model, the main predictors differed to those found for ?CS, with a stronger effect from venous disorden (venous oedema, skin changes and ulcer).
11.5.4. Implications for clinical practice The findings presented in this chapter have important implications for clinical practice. The lack of association between varicose veins and symptoms has several consequences for the management of patients presenting with varicose veins : In patients presenting with varicose veins alone and symptoms, symptoms are unlikely to be caused by varicose veins. It may be necessary to investigate other aetiology. In the presenœ of varicose veins and other signs of venous disorders, symptoms are more likely to be associated with the other signs than with varicose veins. a
In patients presenting with varicose veins alone and with symptoms. the different
options of treatment should be clearly defined and presented according to the objectives pursued. Cosrnetic improvement and improvement of quality of life (including relief of symptoms) are different objectives which should be considered
separately. O
Outcornes of varicose vein treatment should also be defined. Removal of varicose veins will not necessarily improve symptoms or quality of Me. Evaluation criteria of interventions perfomed on varicose veins should therefore include cosmetic improvernent.
In addition, the findings presented in this thesis support the use of the CEAP
classPcation (Porter et al., 1995) for the description of a population of subjects with
venous disorden, since it makes a clear distinction between signs and symptoms.
12. Health services utilisation in Belgium 12.1 Introduction
Health services utilisation refiects a need (real or perceived) for treatment and is an objective indicator of the clinical importance of a medical condition to the patient. Health resource utilisation is also a means of summarising the medical consequences of the disease, its complications and its impact on the patients' wellbeing. Current estirnates of the use of health resources by patients with varicose veins are inadequate for several reasons. Firstly. several studies have based their estimates on the collection of administrative and statistical data. These data include prescription statistics. total spending for ambulatory care, hospital care and venous-related drugs. and compensation fees for lost work days.
These data do not generally allow
conelation of spending with individual variables like severity of symptoms or socioeconomic level. Moreover, they generally concern data on spending related to subsidised health care services. The utilisation of non-prescribed drugs (e.g. pain killers) or non-reimbuned drugs (e.g. venoactive drugs in some countries), or the purchase of non-reimbursed material (e.g., compression material) cannot be reliably estimated from these sources. Secondly. studies that collected individual-based data have generally restricted their scope to the measurement of specific treatments. For exarnple, sorne authors (Beresford et al., 1978; Belcaro et al., 1991) have calculated the cost associated with sclerotherapy for varicose veins in patients consulting at their own clinics in order to undergo a particular treatment. This type of design questions the extent to which such patients are representative of patients attending other practices. Thirdy. the use of monetary values to provide a single value of the utilisation of
various health resources is affected by the year of data collection and the physical units rneasured. The realisation of cost estirnates is not aiways possible in some
studies due to the lack of information on the numben of various physical units used by patients. W~ththe purpose of addressing some of these shortwmings, this chapter presents an estimation of health resource use for varicose veins in Belgium, using individualbased data collected prospectively. Retrospective information pertaining to the prestudy pend is also used to assess the effect of study participation on resource use. Both the patient and hidher physician were surveyed independently, and data obtained from these fiNo sources were cross-checked. This study was restricted to Belgium because data collection was closely supervised in this country in order to obtain high qualrty data. Moreover, the knowledge of the health system allowed an accurate valuation of physical units of resource use.
1 2 2 Objectives 12.2.1 Genenl objective To assess the intensity of health services utilisation in subjects with var~coseveins
consulting general practitioners and specialists in Belgium, and to identify the effect of varicose veins on the cost of health services utilisation in patients with venous disorders.
12.2.2 Specific objectives
To estimate in subjects with varicose veins consulting general practitioners and specialists :
-
the number of visits to general practitioners and specialists related to the presence of a venous disorder
-
the use of laboratory diagnostic tests of venous incornpetence
-
the extent of drug treatment prescribed for venous disorders by physicians. and bought over the counter by patients
-
the use of compression therapy
-
the use of other types of treatment the number of hospitalisations and the number of days of hospital stay associated with venous disorders of the lower limb.
a
To estimate the effect of varicose veins on the cost associated with resource use in patients with venous disorden, adjusted for the effect of other deteninants of cost. To compare results obtained by two methods of collecting data on health services utilisation : direct interview of patients and questionnaire completed by physicians.
f2.3 Patients and methods 12.3.1 Ovurview
Utilisation of health services was assessed in Belgian patients recruited into the VEINES study. and for whom follow-up information was available. Results were calculated separately for individuals included by general practitioners or by specialists.
They were stratified according to the varicose vein classification
presented in Chapter 6. Data pertaining to patients not included in any of these categories (i.e. patients without varicose veins and with oedema, skin changes or healedlactive ulcer) were therefore excluded from the descriptive analyses. They were nevertheless included in the analysis assessing the impact of varicose veins on total expenses of venous disorden. For each item of utilisation, three estimates were provided when applicable : 1) estimates from retrospective physician-based evaluation; 2) estimates from prospective patient-based evaluation; 3) estimates from prospective physician-based evaluation. Calculations included the sum of ail units, the mean number of units per patient in each clinical category (rather than per patient having consumed at least one unit). and the range. The median and the standard enor of the mean were not included,
O
the former because the median was O in most cases, the latter because of the skewed distribution and large variability of data. Estimates of direct medical costs were computed by multiplying the number of physical units of health service utilisation by the corresponding cost estimates. Cost data were derived from official tariffs, from a sample of hospital bills, from expert opinion and frorn the patients themselves. For each item, the total cost (irrespective of the source of payment) and the cost for patients was computed. A semilog model was ftted in an analysis of covariance to assess the effect of
varicose veins on the total cost associated with resource use, taking the effect of other suspected deteninants of the costs into account.
12.3.2 Study population The study population for the analysis of health service utilisation included subjects
recniited in the VEINES study in Belgiurn, excluding those who presented with oedema, skin changes or ulcer (healed or active) without varicose veins at the baseline visit. The study population thus included 335 subjects recruited by 48 general practitionen and 98 subjects recruited by 15 specialists.
Follow-up
information was available for 334 (99.7%) and 97 (99.0%) of these patients respedively. The distribution of subjects by clinical category in each sample is presented in Table 12.1.
Table 12.1 Distribution of subjectr included in the rtudy of health rewices utiiisation in Belgiurn. by physician type (GP: general practitioners, SP: specialists) No signs or telangiectasia alone GP-based sam~le nurnber of patients % of patients SP-based sam~le , number of patients 1 W of patients
Varicose
veins alone
Varicose veins and oedema
25 7.5
81 24.2
35 10.4
31 31-6
26 26.5
3 3.1
W and skin changes
W and healed or active ulcer
All patients
43.0
50 14.9
335 100.0
23 23.5
15 15.3
98 100.0
144
In order to have a comparison group, the multivariate analysis of the effect of varicose veins on resource use was based on the total sample of VEINES patients recruited by general practitioners in Belgium, and for whom follow-up information was available in both patient-based and physician-based evaluations.
This
analysis thus included 355 subjects (96.2% of the 369 patients recruited).
f2.3.3 Methods of data collection
Schedule of data collection In order to obtain robust estirnates of resource use, two sources of information were used : the investigator and the patient himlhersetf. Since part of the direct medical costs is available without medical prescription (e.g. compression materials and medications), 1 was indeed necessary to obtain from the patient himselfBerself information about the use of items that general practitioners rnay not be made aware of. Moreover, informal services and consultations with general practitioners or specialists other than the study investigator may not be mentioned to himlher by the patient. Different tirne schedules were followed to obtain information from investigatos and from patients :
lnvestigators
Ewnomic data were wllected at each of the patient's follow-up visits to the investigators. On the first visit (inclusion visit), investigators were asked a series of questions about treatments used in the past for venous-related k g disorders, including medications, interventions, consultations and diagnostic procedures. Another set of questions asked for information on cunent treatrnents: compression. dnigs, sclerotherapy, physiotherapy, wound care and drug use. On each of the follow-up visits, econornic items used sinœ the last VEINES visit were re-assessed using the same questions as in the first evaluation. Medical events (including surgical procedures and number of prescriptions for defined drug classes) covered the period since the previous VEINES visl.
Patients Patients' questionnaires were administered by phone interviews. al1 interviews to a same patient being (in the majority of cases) perfomed by the same interviewer. The interview schedule included three interviews at three month intervals, the fint one taking place three months after the inclusion visl. Since the duration of follow-up was right-censored by a defined calendar time. some patients with a late inclusion date (e.g. in September 1995) were interviewed only twice. Data collected. Consultations
In the physician-based evaluation, investigaton were asked to report the total number of visits made to different physicians dunng the period covered by the questionnaire. Closed questions were asked about the number of consultations made with the following physicians : the investigator himself, any other general practitioner, angiologists or phlebologists, surgeons. cardiologists. dermatologists. An open question asked the investigator to identify any consultation with specialists not cited in the list above, and to report the number of consultations. For that purpose, the investigator could make use of the patient's medical records. In the
specialist-based sample, only the number of consultations with the investigator himself was requested. In the patient-based evaluation, more precise infonnation was sought : date, peson consulted (e.g . general practitioner,...), place of consultation (hospital, physician's office,...), interventions and treatments performed or prescribed in the consultation. This infonnation was requested for every consultation during which the problem of venous disorder was discussed with a physician, a paramedic or another healh professional. Diagnostic procedures
Information collected about diagnostic procedures consisted of the recording by physicians of the procedures undergone since the last visit. A closed question elicited information about the following laboratory tests : Doppler, duplex, plethysmography, rheography, phlebography, measure of venous pressure, scanner, RMI. An open question investigated whether any other test (to be specified) had been performed. There was no question about the number of possible procedures because it was considered that only one of each test would be perfoned over one year.
ofugs In the physician-based evaluation. the investigator had to state the name, number of prescriptions and mean duration of prescription of venoactive drugs prescribed during the observation period since the last vise. The number of prescriptions and mean duration of prescription was ako noted for any prescription of diuretics. analgesics and non-steroidal anti-infiammatory drugs used for venous disorders.
In another question, the physician was asked to detail the names and dates of administration of any product, compound preparation or topical drug used for venous disorders in use at the time of the consultation. The date of the last prescription and the number of days wvered by the last prescription was also
0
requested.
In the patient-based evaluation, the patient was asked to provide information on the purchase of any drugs prescribed for hisiher venous disorder. This information included the date of purchase, the brandname, the number of packages bought on that occasion and the price paid. Compression matetial In the physician-based evaluation, use of compression was only requested as a 'yes' or 'no' response, with specification of the type of compression used: bandages, stockings or tights. In the patient-based evaluation, the questions concerned every purchase of compression material. with information on the date of purchase, the type of compression material, the brandname, the number of units purchased. the price per unit, and whether the material was purchased in a phamacy or not. An additional question concerned the frequency of use of that material in five categories: everyday, at least once a week, about once a week, less than once a week, never. Hospitalisations
In the questionnaires completed in by investigaton, general practitioners were asked questions about hospitalisations associated with the venous disorder, with reason for hospitalisation, hospital units visited, numbei of days in each unit, interventions and treatments. This information included the reason for hospitalisation and the number of days of hospital stay. The question was not asked to specialists. Questions asked to the patient about each episode of hospitalisation concemed date of hospitalisation, reason, service(s) to which the patient was admitted, intervention(s), number of days of hospital stay.
0
12.3.4. Design options for cost valuation
In order to summarise in a single value the different units of health service utilisation and to estimate the cost of varicose vein disorders in the source population, a monetary value was assigned to each unit of health service utilisation. Several options needed to be considered in this process (Drummond et al., 1989; BESPE, 1995; IBES, 1998, p. 217). They are presented below.
In accordance to the study population of the VEINES study, the target population was represented by patients aged 18-75 years consult~nggeneral practitioners and specialists for venous disorders in Belgium. Since patients consulting general practitioners and specialists differed in many respects, they were considered separately. The estimation of cost estimates was performed only with subjects recniited by general practitionen because it provided a more global picture of the management of venous disorders, including refenal to specialists. Time horizon The time horizon for this economic analysis was limited by the duration of the follow-up in the VEINES cohort. The time horizon was therefore restricted to a
period of 12 months after inclusion into the study. The advantage of using a short time horizon is the possibility of maintaining a low drop-out rate during the observation period and also collection of detailed and unbiased patient-based data during that period. This horizon may however be too short ?O capture in a large number of subjects relevant clinical outcomes with signifiant economic impact
such as the occurrence of ulcer. The analysis is also unable to estimate long-tenn costs associated with a specific clinical presentation. Since the study was based on consecutive patients with a cornplaint of venous disorder consulting physicians. the results relate to a mix of subjeck at different stages of the varicose vein
disorder. They were standardised to the distribution of the different stages of
varicose veins in the source population in order to reflect average values for al1 subjects W h varimse veins, irrespective of the duration of their disease. Pers~ective Two perspectives were adopted when sumrnarising health services utilisation by a single cost estimate: a society's perspective and the patients' perspective. r In the society's perspective, the total costs of al1 units of
health services
utilisation are totalled up irrespective of their source of payment.
This
perspective is relevant since it represents the burden of disease for the society and therefore gives a suggestion for its importance to public health. r In the patients' perspective, only the fraction of the costs that is supported by the
patients is totalled up. This perspective provides an estimate of the economic burden of venous disorders for the patients thernselves.
This analysis is
appropriate since a large range of health services provided for the treatment of varicose veins are not reimbuned in Belgium. Tv~e of costs considered
Economic consequences may be divided into direct and indirect costs (Drummond et al., 1989). Dimct costs represent al1 medical costs that can be readily attribut& to the care given to the patient to treat the disease, such as dnigs, medical consultations, hospitalisations, or medical goods. They include costs borne by the patient and their families such as out-of-pocket expenses and patient and family input into treatrnent. (Drummond et al.. 1989, p. 22). indirect costs include psychic cost and production losses incurred to the patient andlor family as a result of seeking treatment or participation in a health care programme (Dnimmond et al., 1989, p. 21). Indirect costs thus include societal tangible costs, which can be easily counted and valued (e.g. workdays lost, transportation to hospital) and intangible costs that would represent the effect of impaired quality of life to both the patient and his relatives. This analysis indudes the estimation of direct medical costs only.
Method of cost valuation Several assumptions were needed to estimate the cost assigned to physical units of treatment. These assumptions and the information on which they are based are presented below. A summary of cost assumptions is presented in the Appendix 1 to this chapter. Consultation fees for general practitioners and specialists were derived from official tariffs published in the Nomenclature des Soins de Santé (INAMI, 1998). The Nomenclature provides two figures: the total cost and the cost reimbursed by the sickness fund to the patient. The difference represents the official price paid by the patient. although the actual price paid by the patient may Vary. Cost of diagnostic tests were atso derived from the Nomenclature. The cost of a specialist's visit was added to the cost of the procedure. Drug costs were taken from the Belgian medicines fomulary (CBIP. 1996). For each medicinal product, the formulary gives the recommended daily dose. the total cost per package and the rate of reimbursement by the sickness fund. At the start of the VEINES study, most authorised venoactive drugs were reimbursed in Belgium at a rate of 20%. This policy was modified in January 1997 when the reimbursement of venoadive drugs was withdrawn. A reimbursement rate of 20%
was nevertheless considered in this analysis since it was the rate applied in 1994 and 1995. For each patient, the cost of drug treatment was computed from the package sire, the strength and the recommended daily dose. In some cases, the information provided by the patient was insufficient to identify the cost of the purchases; examples are
lack of information on the package size or on the
strength of the dnig. In these cases, the available data were supplemented by the information reported by the investigator for the same patient (drug name, dose and duration of treatment covered by the prescription).
The cost of compression material varies considerably according to the type of compression (bandages. stockings or tights), the strength of compression provided. the quality of the material, the brandname and the place of purchase (pharmacy or large store). The main source for estimating the cost of the material was the patient himlherself. Checks were made by computing the range of costs mentioned for bandages. stockings and tights of the same brandame. Srnall differences were observed once the place of purchase was taken into account. When the patient stated that a specific compression material was bought in a phanacy, the unit cost for the material with the same brandname was checked with the manufacturers' price lists available in pharmacies, Since this verification was made one to two years after the patient's interview. differences could exist between these two sources of information because of changes in the manufacturers' tariffs. However, the costs found in these two sources of information were of the same order of magnitude and no correction was made to the costs provided by patients. When the patient did not provide a cost. the average cost for the same type of
compression material and brandname was applied. The costs of other types of matment (including sclerotherapy, vein stripping, other surgery, ulcer care, physiotherapy) were derived from official tariffs and from a sample of hospital bills (provided by the billing office of the University Hospital of
Liege) made out for patients who underwent stripping or other venous surgery. The assumption was made that treatments of sclerotherapy, stripping and other surgery were perfoned as ambulatory care, unless the patient specifically stated he had been hospitalised for these procedures. The cost of a treatment by sclerotherapy can be divided into three components : the consultation fee, the fee for the technical intervention and the price of the sclerosing agent. The fees for the visit and for the technical interventions are specified in the Nomencîature and the price of the agent is mentioned in the medicines formulary. The sclerosing agent (aethoxysclerol) is not reimbursed (but is not expensive). and the piice of the technical intervention benefits from a small
reimbursement rate. The price mentioned by patients however varies considerably because many specialists apply their own fees without consideration to the official tariff. For this analysis, only official tariffs were used. It was also assumed, based on expert opinion, that a treatment by sclerotherapy included an average of 4 sessions. The cost of the visit to the physician for sclerotherapy was not counted because it was assumed that it was already included in the cost of consultations. The cost of vein stripping and surgery was also divided into three components : the cost of the technical intervention, the medical expenses incurred by the procedure (anaesthesia, wound care, blood studies. diagnostic tests,...) and the hotel costs. The cost of each category was estimated from the sample of hospital bills.
12.3.5 Data analyses 1- Search for duplicates A search for duplicated infomation was perfomed on the database in order to
avoid counting the same data mice if they had been recorded in two consecutive interviews or questionnaires. This identification was perfomed on the date of purchase and the type of item bought. Two sets of infomation recorded at two visits were considered to be duplicates if both pertained to the same drug bought
on the same day. The earliest of duplicated records was kept in the database.
2- Observation time in the ~hvsician-basedeconomic evaluation The period covered by each economic questionnaire completed by physicians was 'since the last visit'. Consequently, the observation time window covered the entire time elapsed between the last follow-up visit and the inclusion visit.
The
observation time (in months) was calculated as the difference between the date of the last visit to the study investigator and the date of the baseline visl, divided by
30.4.
3- Observation time in the patient-based economic evaluation By protocol, each questionnaire asked questions relating to events (e.g.
consultations induced by venous disorders) having occurred durina the last three months. This time period was chosen as the longest period that could be explored by phone interview with acceptable validity. When phone calls were separated by
less than 3 months, the observation time for that interview was calculated as the difference between the dates of the two phone calls. An observation tirne of 3 months was taken when the two phone calls were separated by 3 months or more. Thus, by protocol, the maximum observation time was 9 months. Since the time window was restricted to the 3 months before each interview, items of economic evaluation recorded at a calendar date falling outside that time window were excluded. This procedure was not applied to the 13 episodes of hospitalisation. Hospitalisationwas an unique event in al1 cases and it was felt that using a calendar date as a criterion for exclusion would lead to missing important information relating to the follow-up period. 4- Com~utationof crude, adiusted and standardised estimates
i) crude estimates
For each item of health services utilisation, two types of estimate were calculated : a) number and proportion of patients having used at least once a specific economic item; for example the number of patients with varicose veins alone having consulted a surgeon ; the proportion of patients was calculated with reference to the total sample site in the same clinical category. b) number of units consumed (e.g. number of times the patient consulted a surgeon). The units considered differed according to the source of data (patient
or physician) and to the economic item examined. For example, drug utilisation was estimated by the nurnber of packages bought (patients' data) and by the number of treatment-days covered by prescriptions (physicians' data).
ii) one-year adjusted estimates
Annual estimates of health service utilisation are easier to interpret, allow comparison between patient-based and physician-based estimates and faciMate comparisons with other surveys. To this end, the crude numbers of events were divided by the number of days of actual observation and multiplied by 365. This extrapolation
required the assumption that resource use foilows an unifom
distribution over the year. The validity of this assumption needed to be verified, especially for patient-based interviews which covered an average of 8 months of the year. Cook et al. (1997) have in fact demonstrated a seasonal variation in referral pattern of patients to a specialist varicose vein clinic in the United Kingdom between January 1991 and December 1994, with more patients being referred in the sumrner months. Annual estimates would therefore be excessive if more than 33% of the total period of observation covered June, July. August or Septernber.
They wouM be underrated if these months were included in less than 33% of the observation time. Of the 3,461 patient-months experienced in the patient-based survey, 629 took place in June, July, August or Septernber, which represented a proportion of 18.2% (95% confidence interval : 15.5-20.9%). Therefore. estimates of health service utilisation are more likely to be undenated than overrated. This adjustrnent process was only made for patients who experienced an event during their observation time. It results in a possible underestimation of the yearly number of events, since the calculation assumes that patients who did not experience an event during the observation time also did not experience one over 12 months of follow-up.
iii) standardised estimates The Belgian study population used in this evaluation represents a biased sample of the source population since male subjects and subjects with ulcer were intentionally over-sampled. Standardised proportions and means were therefore calculated to
estimate the proportions and the means that would have been observed in a hypothetical cohort whose age, gender and clinical category distributions followed the distributions observed in the source population. The method of standardisation was explained in Chapter 11. Section 11.3.6. In this analysis. no stratification by gender was used and the set of weights was defined by four age and gendercategories. 5- Amount of drua utilisation i) coding of dnigs
All dmg use reported by physicians and patients was coded by a clinical research assistant according to the Anatornical Therapeutic Chernical (ATC) classification index (WHO,1996). A double check of the code assigned to each dnig was perfomied by a physician (myself). Dnigs reported by physicians and by patients were divided into four categories: 1.
Relation of indication to chronic venous disorden of the leg (CVDL) is certain
or
probable.
2.
Relation of indication to CVDL is possible.
3.
Relation of indication to CVDL is unlikely.
4.
Dnig is unknown or information is insufficient to assign an ATC code.
Table 12.2 ôelow indicates the ATC codes assigned to each category.
Table 12.2 Classification of dnigs according to relationship with venous disorders Category 1.
2.
Drug class Antivaricose therapy Capillary stabilising agents Emollients and protectives Preparations for wounds and ulcers Other antipniritics Antiseptics and desinfectants Medicated dressings Anti-thromhtic dwgs Other haernatologicalagents Diuretics Penpheral vasodilators NSAIDS NSAIDS, combination Non-opioid analgesics
Categories of ATC codes CO58 C05C DO2
DO3 DWAX DO8 DO9 BO1
Bo6 CO3 Co4 MO1A M02B N02B
3.
Drugs for other diseases
Other codes
4.
Drugs unknownfunspecified
Code missing
Only the two first categories (relation of indication to CVDL is certain, probable or possible) were considered in the econornic analysis. They were subdivided into four categories: Venoactive druas: Capillary stabilising agents Other vascular drugs: Antivaricose therapy, Anti-thrombotic drugs. Other haematological agents, Peripheral vasodilators Dressinas or to~icaldruas: Emollients and protectives, Preparations for the treatment of wounds and ulcer, Other antipruritics, Antiseptics and disinfectants, Medicated dressings Other druos ~ossiblvrelated to CVDL: Diuretics, Systemic and local antiinflammatory drugs (alone or in combination), Non opioid analgesics.
ii) assurnptions used in the calculation of intensity of drug tmatment The estimation of drug utilisation was assessed using the number of packages purchased in the patient-based economic evaluation, and the number of treatment-
days in the physician-based evaluation.
Specific methods of calculation or
assumptions were used in the estimation of drug utilisation as follows: Patient-basedevaluation When the purchase of a drug was recorded but the number of packages bought was not stated, this nurnber was estimated conservatively as one. Physicien-basedevaluation The number of treatment-days covered by prescriptions made durhg a defined time period was calculated by multiplying the number of prescriptions written during that period with the mean duration of treatment covered by a prescription; the total number of treatment-days in the entire observation period was the sum of four figures :
* the nurnber of treatment-days after the first visit covered by a previous prescription or by a prescription made at that visit; r the number of treatment-days covered by prescriptions made between the
first and the second visit (excluding those made at the first visit);
* the number of treatment-days covered by prescriptions made at the time of the second visit;
* the number of treatment-days covered by prescriptions made between the second and the third visit (excluding those made at the second visit). a
For sorne patients, the period covered by prescriptions was longer than the adual observation time defined according to the dates of consultations; in these cases. the number of treatmentdays was truncated to the duration of the actual observation time and adjusted within a 12 month-time window. When N o drugs of the same class were prescribed during the same period (for example, two venoactive drugs prescribed between the first and the second visit), the number of treatment-days was taken as the largest of the numbers of treatment-days estimated for each dnig.
iii)monetary unit All costs were estimated in Belgian francs. In order to make costs more easily understandable, and to compare costs calculated in this study with those presented in published studies, costs were translated into US dollars. An exchange rate of 37 Belgian francs to one dollar was used.
6. Identification of the effect of varicose veins on total costs Analysis of covariance was used to compare the mean total costs associated with health servises utilisation in patients with and without varicose veins, adjusted for the effect of potential confounden. The method used is identical to that presented in Chapter 11. section 11.3.6. The dependent
variable in this analysis was the total cost of varicose vein
treatment, computed as the sum of the societal costs of consultations, diagnostic testing, drug therapy. compression treatment, other types of treatment (surgery, sclerotherapy. physiotherapy, wound care) and hospitalisation. Since the distribution of the total cost was negatively skewed (more values were below the mean than above it), a log transformation was used to transfomi the dependent variable to the normal distribution (Amitage and Berry, 1987, p. 360; Kleinbaum et al., 1988, p. 220). In doing sol consideration was given to the fact that the log function impacts on the interpretation of the model parameten (Johnston, 1984. p.66). The semilog model implies the transformation of an additive mode1 to a multiplicative one, i.e., log Y, = a + PX, + E,
If we wish to compare the effects of X, and Xaon Y,
and
Y, = Y/ eP
ifx, - x0= 1.
The adjusted 'effect' was used as the primary outcome of interest and estimated as the cost ratio (eP)associated with the 'exposed' level of a variable in cornparison to the 'non-exposed' level. The 95% confidence interval around the adjusted effect was used to estimate its precision. For 51 subjects (14.4%), the total cost amounted to O. In order to include these patients in the analysis of covariance. a cost of 1 US $ was assigned to them. The sarne techniques of regression diagnostics used in Chapter 11 were applied in this analysis to check the appropriateness of the model (Kleinbaum et al., 1988, p.181). In order to test the assumptions of the model, residuals (standardised residuals, studentized residuals and jackknife residuals) were plotted in univariate analysis and in bivariate analyses against predicted values. Predicted values were plotted against observed values.
Jackknife residuals, leverages and Cook's
distance were computed and plotted in order to detect outliers and measure their influence. As the main objective was to assess the effect of varicose veins on costs
associated with varicose veins, the selection of variables included into the model
was based on prior knowledge of their potential confounding effect. Two variables (employment status and income) were tested in bivariate analysis. These variables were not associated with the dependent variable in bivariate analysis and they were not introduced into the multivariate model. Their effect would also be captured by the level of education All other variables were forced into the model and none were rernoved on the basis of statistical criteria. The rationale for this choice was based on several arguments. Varicose veins being the determinant of interest, clinical variables such as telangiectasia, venous oedema, skin changes and ulcer were introduced in order to separate their effect from the effect of varicose veins. Age, gender, educational level and CO-morbidityare confounding variables. Body mass index had to be introduced because an association between venous disease and BMI has been reported by several authors and would therefore also be associated with health
service utilisation for venous disorders. Past thrombophlebitis is a condition which can entail specific treatrnents and consultations, independently frorn b effen on varicose veins, skin changes or ulcer. The specific quality of life index (VEINES-
QOL) rneasures a different dimension than that accounted for by physical signs (varicose veins. oedema, skin changes, ulcer). It reflects both symptorns and the effect of venous disorders on someone's well-being. This dimension can affect the utilisation of semices for venous disorders and be associated with varicose veins. VEINES-QOL was therefore classified into 4 categories defined by quartiles and these categories were introduced into the model. In summary, the following variables were tested : varicose veins (exposure of interest): 2 categories : absent (O) or present (1) at
the baseline clinical examination. telangiectada : 2 categor~es: 011. venour, oedema : 2 categories : 011. skin changes : 2 categories : 011. healed or open ulcer : 2 categories : 011.
past history of thrombophlebitis : 2 categories : 011. gender : 2 categories : maleJfemale. age : 3 categories (18-39,40-59,60-75 yean old). body mas8 index : 4 categories (< 25.0, 25.0-26.9, 27.0-29.9. r 30.0); three
categories have been used in some analyses : < 27.0. 27.0-29.9, 2 30.0. specifit quality of life (VEINES-QOL) : 4 categories defined by quartiles of the
VEINES-QOL distribution. coniorbidity : 2 categories : O (no major morbidity as defined in Chapter 8) or 1
(presence of at least one of the morbid conditions). educational level : 2 categories : low (lower secondary school or less) or
rnoderate to high (high secondary school to collegeluniversity). employment status : 2 categories : O = no employment, 1 = paid employment income : 3 categories : low, intermediate, high.
12.4.1. Observation time 1- Obsenration time in the ~hvsician-basedeconomic evaluation
A total observation time of 4716 months was captured for the 416 subjects included
in this econornic analysis. The mean observation-time was 11.79 months in the general practitioner-based sample and 9.71 months in the specialist-based sample (Table 12.3). Table 12.3 Physician-based economic evaluation in Belgium: number of patients followed and number of person-months of observation between the first and the last clinical examination ----
Patients recruited by : general specialists al1 physicians practitioners (n=433) (n=335) (n=98) Phvsician-based evaluation n (Oh)patients followed 325 (97.0) patient-months 3832 mean (std. deviation) 11.79 (1.95) range 5.74-20.98 Patient-based evaluation n (%) patients followed 332 (99.1) patient-months 2700 mean (std. deviation) 8.13 (0.92) 3-9 range
91 (92.9) 884 9.71 (3.17) 1.61-44.43
416(96.1) 4716 11.34 (2.43) 1.61-20.98
97 (99.0) 760 7.84 (1.30) 3-9
429 (99.1) 3460 8.07 (1.03) 3-9
2- Observation time in the patient-based economic evaluation At least one follow-up questionnaire is available for 332 (99.1%) of the 335 patients included by general practitioners and for 97 (99.0%) of the 98 patients included by specialists (Table 12.3). The total observation-time covered by the patient-based economic assessrnent was 3460 patient-months. 78% of which were accounted for by subjects recruited by general practitioners.
About 10% of events relating to consultations, purchase of drug and compression material fell outside the three-month time periods and were excluded (Table 12.4.)
1
Table 12.4 Number and proportion of econornic events occurring outside the defined time window and excluded (patient-based evaluation) Number of items recorded
Economic item
r
Number outside the time window
O h excluded
.
Consultation for venous disorder
1549
Purchase of venous-related drug
1769
Purchase of compression material
303
12.4.2, Consultations
General ~ractitioner-basedsample The one-year adjusted age and gender standardised means of consultations per patient to general practitionen and specialists are presented according to the different sources of data in Appendix 2, Tables A.2.1 to A.2.3. Overall. patients made an average of 2 to 4 visits each per year to a general practitioner (GP) for venous disorden. The mean number of consultations for
venous disorders during the follow-up time tended to increase with the severity of signs associated with varicose veins (from 1.9 to 3.5 in the physician-based evaluation), but a comparable result was found in patients with telangiectasia alone or symptoms alone (2.3). At the aggregate level, good agreement was found between the different sources of information, although patients reported a higher number of visits than physicians for al1 patients with varicose veins (average of 3.9)
or in each clinical category (range: 3.0 to 4.8). Results from the retrospective evaluation were consistent with the est~matesobtained prospectively from patients and physicians. The number of visits to a specialist was estimated by both patients and physicians at about one visit per two patients with varicose veins ON) per year.
Large
differences across varicose vein categories can be explained by the small number
of consultations recorded. For this item, the analyses based on the patient and the physician data did not completely agree. Comparable proportions of patients consulting a physiotherapist were reported by physicians (3.6%) and patients (4.0%).
This proportion was doubled in the
retrospective evaluation. The average number of visits could be computed only for the patient-based evaiuation and was 0.06 per patient in patients with W. vs. 0.04 in patients without any signs or with telangiectasia only. Between 3 and 12% of patients reported a visit to another heakh professional, with an average of 4% for subjects with varicose veins.
The number of consultations to various types of physicians was requested only from patients. On average, a patient with W recruited by a specialist made 2.6 visits par year to a specialist and 1.6 visits to a general practitioner (Tables A.3.1 in Appendix 3). The number of consultations shows a sharp increase in patients with W and healedlopen ulcer (4.5 to specialists and 2.3 to GPs). 12.4.3 Diagnostic test. General ~ractitioner-basedsam~le The proportions of patients having had a diagnostic test performed at any time before the inclusion visit and during the follow-up period are reported in Tables A.2.4 and A.2.5.
Low percentages of patients with varicose veins had had a laboratory test at any time in the past : about 10% had had Doppler or duplex and 2.3% had had another test. These proportions were not influenced by the presence of signs of more severe venous disorden.
The proportions reported for the follow-up period
(Doppler : 4.6%, duplex : 7.0%) are higher in cornparison to those found in the retrospective evaluation, but laboratory testing may be related to the reason for
a
consultation and inclusion into the VEINES study.
It is striking that less than 10% of subjects with varicose veins associated with a severe disorder like ulcer or ski" changes had had a duplex test perfomed in the past, although this method of testing venous function is now considered standard in the investigation of these disorders.
As expected, more patients consulting specialists had had a diagnostic test performed in the past than patients enrolled by general practitioners: respectively 19% and 37% of patients with varicose veins already had a Doppler or duplex test
performed (Table A.3.2). In the follow-up phase, about 18% underwent a Doppler test and 3.2% a duplex test (Table A.3.3).
The difference in percentages for
Doppler and duplex tests performed in GPs' and specialists' subjects is striking. A duplex test requires expensive equipment and an intensive training, and is usually perfomed in specialised centres. The differences obsenred rnight be explained by specialists performing Doppler tests themselves, with general practitioners referring their patients to specialised centres where duplex was performed.
12.4.4 Drug utilisation
General ~ractitioner-basedsam~le The age and gender standardised proportions of subjects having used different classes of drugs for venous disorden are presented in Tables A.2.6 to A.2.9, according to the type of evaluation. The proportions of patients with varicose veins ON) treated at least once during the follow-up time by venoactive drugs, other
drugs with a vascular indication or topical drugs or dressings (other than locally applied venoactive drugs) are comparable in the two physician-based evaluations, amounting to approximately 60%, 20% and 3% respedively. Except for venoactive drugs, there is a marked increased in the proportions of patients treated with the severw of signs associated with W. The high level of use of venoactive drugs in al1 clinical categories (and especially when varicose veins are associated with
oedema) is consistent with the high proportion of subjects with syrnptoms in al1 categories (Chapter 11, Section 11.4.1).
Two factors may explain the lower
percentages of usage of venoactive drugs found in the patient-based prospective evaluation : i) by protocol, the observation time was restricted to a maximum of 9 months,
ii) the prevalence of drug utilisation was calculated from prescription data in the physician-based evaluation and from purchases of drugs in the patient-based evaluation. This difference can explain the higher estimates found in the physicianbased evaluation, as prescriptions are not necessarily followed by the purchase of the prescribed drug. On the other hand, patients may have also bought nonprescribed medications. The estimates regarding the usage of other drugs possibly indicated for chronic venous disorders (NSAIDS, analgesics or diuretics) also varied greatly from one source to another : it was 5.2% in al1 patients with W as ascertained from patients (Table A.2.8) and 52.7% as ascertained from physicians (Table A.2.7). The mean number of packages of any venoactive drug bought by patients over 12 months was about 5 (A.2.8), and about 8% of purchases of venoactive drugs were made without a prescription. Given a median duration of prescription of 28 days (figure computed from physicians' data), the mean duration of treatment covered by a venoactive drug would therefore be 140 days on average per year of treatment. Other drugs with vascular indications (including anti-thrombotic agents, other haematological agents, peripheral vasodilators and anti-vancose therapy) were purchased at a rate of about 1 package per patient per year with varicose veins, about twice as much as that obsewed in patients without signs or with telangiectasia only. Given the large variability of packages sizes in this drug class, translation to a number of treatmentdays was not possible. The mean number of persondays of treatment with venoactive drugs according to data provided by physicians was about 80 days per patient per year with W (Table
A.2.9). This figure may cover concomitant treatment with two or more venoactive drugs. The treatment duration for analgesics, NSAIDS and diuretics was estimated at approximately 6, 10 and 15 days per patient per year respectively. These figures are difficult to interpret due to the uncertainty attached to the venous indication of these drugs.
In general, the level of drug use is lower in patients consulting specialists than in patients consulting general practitioners (Tables A.3.4 to A.3.7). Smaller sarnple sizes rnay explain larger differences between the three methods of evaluation. About 24% of individuals with varicose veins used a venoactive drug at least once during the follow-up period, with an average of 1.5 packages bought as reported by the patients (Table A.3.6). The mean duration of treatment was 22 days per patient per year (Table A.3.7). Whereas the mean dutation of treatrnent by analgesics is comparable to the duration recorded in the general practitioner-based sample, diuretics and NSAIDS use was reported by specialists much less frequently than by generalists.
12.4.5 Use of compression material General ~ractitioner-basedsamole Data collected on the purchase and use of compression material according to the general practitioners are presented in Tables A.2.10 to A.2.12. Frequency of use of any type of compression was evaluated at the same level by physicians and by patients at approxirnately 19% in patients with varicose veins alone. Large differences were observed according to the signs associated with varicose veins: 34%-46% in patients with varicose veins and oedema, 19%-33% in patients with varicose veins and skin changes, and 28%-48% in patients with varicose veins and ulcer. Tights and stockings were the types of compression most
frequently used (by 13-15% of subjects), bandages being reported by only 3% of subjects with vaficose veins.
Specialist-based sample Use of compression material was higher in patients with vaficose veins recruited by specialists than in those recruited by general practitioners (Tables A.3.8 to A.3.10). Contrary to generalists' patients, the highest prevalence in specialists' patients was in the presence of varicose veins and healed or active ulcer (53%-71%). In parallel to this distribution, the frequency of use of bandages was much greater (9% to 12% of patients). These results are consistent with those found in the retrospective evaluation, although srnall numbers do not allow definite conclusions to be drawn about the frequency of use of compression in specialists' patients. Patterns of compression usa= Amongst all Belgian patients, a mean of about 7 units of compression material per patient using compression (and who provided that information) were bought. This value was 7 to 8 units for bandages and stockings. The proportion of compression material bought in pharmacies was 9.8% for tights, 21.9% for stockings and 51% for bandages. Stockings and tights were worn more than once a week by 70 to 80% of subjects, whereas regular use of bandages was reported by only 50% of subjects.
12.4.6 Use of sclerotherapy and suigery
General ~ractitioner-basedsarn~le Information on other types of treatment received by patients was requested from physicians (Tables A.2.12 to A.2.13). Respectively 22% and 16% of patients with varicose veins had a history of sclerotherapy and stripping any time before the patient's inclusion. These interventions were experienced by 7.6% and 5.6% respectively of patients during follow-up.
No episode of 'other surgeryt was
a
reported before follow-up but it was reported in 5.3% of subjects with varicose veins durhg follow-up. No large difference by clinical category was observed in the retrospective or prospective evaluations. Specialist-based sample As expected. the usage of sclerotherapy and surçery was higher in the specialistbased sample. both in the past and during the follow-up (Tables A.3.11 to A.3.12). More than one third of patients with varicose veins consulting specialists received
sclerotherapy, but the same proportion was found for patients with telangiectasia alone. About 8% of subjects were operated on for venous stripping during followUP-
f2.4.7 Hospitalisations In the general practitioner-based sample. the number of patients with varicose veins hospitalised during follow-up was 13 (4.2%). It only involved subjects with varicose veins and skin changes (7 patients) and subjects with varicose veins alone (6 subjects). The reasons for hospitalisations were ulcer treatment in 5 cases and surgery for the 8 other cases. It is noteworthy that the number of subjects admitted to hospital is lower than the number of subjects operated on because stripping is often performed as ambulatory surgery. The average number of days of hospital stay for the whole sample is 17 per 100 patients. and 4.3 days per patient hospitalised.
Only 5 patients reported a hospitalisation in the specialist-based sample. Four of them underwent surgery.
12.4.8 Cost estimates in genenl pnctitioner-basrd sample Consultations
The total cost of consultations for patients with varicose veins ON) amounted to
73.6 US), the rnajority of which was acmunted for by consukations with general
practitionen (GP) (Table 12.5). A shift in costs was observed in patients with W and skin changes or ulcer. but the average cost for subjects with W alone did not differ from the cost for subjects without signs or with telangiectasia alone. The largest differences across clinical categories were observed for the cost of visits to specialists (total cost ranging from 6.2 to 12.9 US$). About 30% of the consultation costs were out-of-pocket expenses. The estimation of the costs to patients was based on official tariffs and should therefore be considered as minimum costs. Table 12.5 One-year adjusted, age and gender-standardised mean cost (US$) of consultations perfonned for chronic venous disorders (CVDL) per patient. by varicose vein (W)category General practitioner-basedsample '
-
1
Physician consulted
No signs Varicose or telangi- Varicose veins ectasia veins and alone alone oedema
W and skin changes
W and healed or active ulcer
All patients with W . .---
Total costs
General practitioners Specialists All consultations
48.17
49.67
50.71
76.99
61.08
62.48
9.71
6.15
6.91
15.95
12.88
1 1.O8
57.88
55.82
57.62
92.94
73.96
73.56
13.88
14.3t
14.62
22.20
17.59
18.01
3.88
2.47
2.77
6.38
5.14
4.44
17.76
16.78
17.39
28.58
22.73
22.45
Costs to patients General practitioners Specialists Ali consultations
'
Estimates for the category 'All patients with varicose veins' are standardised according to the expected distribution of varicose vein categories in the registration pool
Diaunostic tests Due to the small proportion of patients referred for diagnostic tests in the general practitioner-based sample, diagnostic tests did not represent an important source of costs. The combination of varicose veins and oedema was the clinical condition that was associated with the largest expense. On average, no important difference
was observed between patients wlh varicose veins alone and patients without signs or with telangiectasia alone.
Table 12.6 One-year adjusted, age and gender-standardisedmean cost (SUS)of diagnostic tests perfonned for chronic venous disorden (CVDL) per patient , by varicose vein (W) category Genetal practitioner-based sample '
-
Perspective and physician consutted --
Total costs Costs for patients
'
No signs or telangiectasia alone
Varicose veins alone
Varicose veins and
oedema
and skin changes
W and healed or active ulcer
All patients with W
W
-
11.17
12.51
17.76
13.00
3.50
12.46
0.26
0.48
0.62
0.46
0.13
0.45
Estimates for the category 'Al! patients with varicose veins' are standardised according to the expected distribution of varicose vein categories in the registration pool
Druas The total cost of drugs with a probable or possible relationship with W was 52 US$, 75% of that amount being paid by the patient (Table 12.7). About 70% of the costs were associated with the cost of venoactive drugs. Expense for this class of medication was substantial in al1 clinical categories, especially for subjects with W and oedema (average of US$ 53 per person per year). However. the costs of drugs were not higher in subjects with varicose veins alone than in subjects without signs or with telangiectasia only. It is noteworthy that costs to patients are currently higher than the costs estimated for the years 1995-1996 because venoactive drugs were no longer reimbursed from January 1997 in Belgium. Compression The average costs (US$) of bandages. stockings and üghts for patients with varicose veins were 1.11, 11.81 and 17.44 respectively (Table 12.8). These costs were not distributed evenly across clinical categories : for bandages. the cost was mainly accounted for by patients with ulcer; the higher cost for stockings and tights was seen in presence of oedema or skin changes. All costs were assumed to be met by the patient.
Table 12.7 One-year adjusted, age and gender-standardised mean cost (US$) of drugs used for chronic venous disorders (CVDL) per patient by varicose vein (W)category - General practitioner-based sample ' -.--
-
Variable
No signs or telangiectasia alone
Vaticose veins alone
Varicose veins and oedema
W and skin changes
W and healed or active ulcer
All
patients with W
Total costs Venoactive drug Other drug with vascular indication
Dressings, topical drugs Other drug possibly related to CVDL All drugs
Costs to ~atients Venoactive drug Other drug with vascular indication Dressings, topical drugs Other drug possibly related to CVDL All drugs l
Estimates for the category 'All patients with varicose veins' are standardised according to the expected disttibution of varicose vein categories in the fegistration pool
Table 12.8 One-year adjusted. age and gender-standardised rnean cost ("SI)of compression material purchased for chronic venous disorden (CVDL) per patient. by varicose vein (W) category - General practitioner-based sample '.
'
Type of compression
No signs or telangiectasia alone
Varicose veins alone
Varicose veins and oederna
w and skin changes
W and healed or active ulcer
Al1 patients with W
bandages stockings tights ail types
'
Estimates for the category 'All patients with varicose veins' are standardised according to the expected distribution of varicose vein categories in the registration pool. All costs were met by the patients.
'
Other interventions Although experienced by a small proportion of patients. stripping and other venous surgery perfomed for varicose veins were associated with a large expense (Table 12.9). Among patients with varicose veins alone. there was an average cost of 50 US%for stripping and of 126 US$ accounted for by other venous surgery. The type
of surgery was not always specified by the physician. However, it is known that it was not stripping. Sclerotherapy was not an important source of costs. However. the expenses considered in this calculation only included the cost of the sclerosing agent and the official cost of the technical intervention. It was assumed that the consultation was already included elsewhere in the consultations to specialists costs. On average, about 11% of the costs of interventions perfomed in patients with varicose veins are met by patients themselves.
This proportion varies
according to the intervention concemed and the reimbursement rate.
Table 12.9 One-year adjurted, age and gender-standardiredmean cost p r patient (US)) for intenientions performed during follow-up. General practitioner-based sample '
Variable
No signs or telangiectasia atone
Varicose veins alone
Varicose veins and oedema
W and skin changes
W and heated or active ulcer
All patients with W
Total costs Scterotherapy Stripping Other venous surgery Wound m e Physiotherapy All interventions Costs for patients Sclerotherapy Stripping Other venous surgery Wound care Physiotherapy All interventions
'
Estimates for the category 'All patients with varicose veins' are standardised according to the expected distribution of varicose vein categories in the registration pool.
All costs Table 12.10 summaries the direct medical costs in patients with varicose veins, both from the society's perspective (total costs) and from the patients' perspective. An average cost of 311 US$ per patient per year was estimated in the group of patients with varicose veins consulting a generai practitioner with a cornplaint relating to their condition. The mean cost differed according to the association of varicose veins with other signs of venous disorders. Results showed that surgical interventions were the main deteminant of the costs. As a consequence, the clinical categories showing the highest expenses were those which were the least wmplicated by venous disorden for which maintaining the superficial venous system is important, especially with regards to ulcen (Abenhairn et al.. 1999).
Other types of treatment showed some degree of specificity, such as the use of compression in varicose veins with oedema, and drugs (including wound dressings) in patients with ulcer. Another striking result was the low average cost associated with diagnostic testing. One should expect a proper evaluation of venous function to be a requirement before choosing between different types of treatment. Doppler or duplex examinations were usually performed pre-operatively (and the cost of
such a test was counted in the cost of surgery), but only a small number of subjects were tested in such circumstances. When al1 costs were considered, about one third of expenses in patients with varicose veins (1O7 US$) were out-of-pockets expenses. It is noteworthy that this proportion increased with the severity of signs associated with varicose veins (27% in subjects with varicose veins alone, 30% in subjects with varicose veins and skin changes, about 56% in subjects without signs or with telangiectasia alone, 56% in subjects with varicose veins and oederna, and 59% in subjects with varicose veins and ulcer).
Table 12.10 One-year adjusted. age and gender-standardised rnean total cost pet patient (US category Geneal piactitioner$) of chronic venous disorders (CVDL) by varicose vein based sample '
-
Variable
No signs or telangiectasia alone
Varicose veins alone
Varicose veins and oedema
W and skin changes
W and healed or active ulcer
All patients with W
Total costs Consultations Diagnostic tests Drugs Compression Interventions TOTAL Costs to patients Consultations Diagnostic tests
Drugs Compression
Other interventions TOTAL Estimates for the category 'All patients with varicose veins' are standardised according to the expected distribution of varicose vein categories in the registration pool
12.1.9. Variables iassociated with total c o r b
The analysis of variables associated with costs was canied out in the total study population recruited in Belgium for which follow-up information was available from both the physician and the patient. This analysis thus included 355 subjects of the 369 who had been recruited by general practitioners.
The total one-year adjusted cost of treatments for venous disorders computed for these patients amounted to USB95715, with a mean of 269.6. a median of 117.6. a standard deviation of 529.3 and a range of 1 to 4007.5.
After logarithmic
transformation with natural loganthms, the antilog of the mean was 68.7, the standard deviation was 8.1 and the median was 117.9.
The logarithrnic
transformation thus nonnalised the cost distribution and was used in the analysis of covariance. The effect reported in Table 12.11 represents the cost ratio (CR) calculated as eP,p being a parameter provided by the covariance analysis. The CR indicates the relationship between the cost in individuals with and without the exposed level of a variable. Inspection of residuals specified in the Method section did not indicate any particular problem. The two variables having the strongest association with the total cost are the presence of varicose veins (CR:2.32, 95% CI: 1.254.30) and the venous diseasespecific quality of life (VEINES-QOL) (CR for 4 4 vs. Q I : 0.14, 95% CI: 0.07 to 0.27). Their effects were observed both in bivariate and rnultivariate analyses and the 95% confidence interval (CI) around their adjusted effect excludes the nul1 value. Results for VEINES-QOL also suggest a trend towards increasing costs with lower specific quality of life. For al1 other variables, inciuding the presence of skin changes and ulcer, the 95% CI includes 1.00. This result suggest that the high costs obseived in patients with varicose veins and skin changes (Table 12.10) are associated with the presence of varicose veins rather than the presence of skin changes.
Table 12.11 Variables associated with total costs in Belgiurn (n=355) (effect = cost ratio). Variable
Sample size
Unadjusted effect estirnate 95% CI
Adjusted effect ' estimate 95% CI
relanaectasia no Yes Varicose veins no Y= Venous oedema no Yes Skin chancles no Yes Healedlo~enulcer no Yes Thrombo~hlebitis no Yes VEINES-QOL QI (low scores) Q2 Q3 Q4 (high scores)
' Adjusted for al1 other variables included in the table, and for age, gender, body mass index, comorbidity and educational level; sample size: 305
f2.5 DISCUSSION l2.S.l Methodological issues A report of health service utilisation is inherently descriptive and aims to provide a representative picture of a situation in a defined population over a defined time period. This set of data may be useful for different purposes : to help define health care needs, organise health services and respond to the patients' demand; to assess the public health impact of a disease and compare it with that of other diseases, allowing definition of priorities for allocation of budgetary resources, e.g., for prevention; to estimate the additional cost incurred as a consequence of the disease, that is required for economic evaluation of health care programmes; to bring new knowledge about the disease, by estimating the sequence of diagnostic procedures and treatments needed; and to estimate the prevalence of the disease. Although a valuation of the units of health care utilisation couM mask important differences in the numbers of units measured, it also useful to summarise in a single estimate the economic impact of a disease (BESPE, 1995). The evaluation of health services utilisation in Belgiurn presented in this thesis is sufficiently detailed to be used for these purposes. Moreover, to our knowledge. it is the first study that assessed the economic impact of varicose veins on such a large sale based on data directly collected from patients and physicians. In practice, the usefulness of economic analyses depends on the fuliilment of several prerequisites. At least three of them can be identified : 1) the results obtained on a sample can be extrapolated with confidence to the population for which economic outcomes need to be assessed; 2) the units of utilisation have been accurately measured, i.e. without systematic bias; 3) the valuation of the costs is based on empirical data or on assumptions that reflect real costs. These three aspects are exarnined in turn.
1- Extra~olationof results
Data on resource use presented in this thesis concem patients consulting general practitioners and specialists for venous disorders in Belgium. They are therefore representative of individuals seeking treatment rather than of individuals with venous disorders in the general population. In such a situation, the possibility should certainly be envisaged that the inclusion of patients into the study and the recording of physicians' activities could have influenced the use of health resource. As an attempt to minimise the likelihood of this bias or. at least. to assess it, study subjects were recruited from consecutive patients consulting physicians participating in the study. In the second stage, the study subjects were selected by the study CO-ordinators. One of the aims of this procedure was to try to avoid the selection of the 'best' patient by the investigator, i.e. the patient who could be easily followed because helshe reliably attended hislher consultation. This aim was only partly achieved : it was shown in Chapter 4 that some patients who should have been included were excluded without fulfilling one of the exclusion criteria; this suggests that some investigators recruited their patients instead of adopting the procedure of actively contacthg the patients they had to recruit. However, the minimum information collected dunng the registration phase included age, gender and the clinical presentation.
This allowed to standardisation of the results
obtained from the study subjects to the source population of patients consulting physicians. Therefore, although the results presented in this thesis do not apply to al1 subjects suffering from varicose veins in the population, they can be assumed to give an accurate representation of the utilisation of health services in subjects consulting physicians for this condition.
2- Absence of svstematic bias Measurement biases cannot be totally avoided in a descriptive study based on the observation of medical practices. However. attempts have been made to assess the likelihood that results were distorted by constraints imposed by the protocol.
An important feature of the study was the use of an ambispective approach for data collection. The retrospective assessment of the utilisation of health services was not considered as a specific endpoint of the project but was mainly used for validation purposes. A correlation between retrospective and prospective assessments was not expected at the individual level because patients were consulting with the express purpose of receiving a treatment. However, the study was based on prevalent cases of a chronic disease. Patients entering the study were therefore at different stages of disease severity with a long history of treatment for some of them.
In such a situation, the evaluation based on
retrospective data would give comparable results at the aggregated level to those collected in a prospective evaluation. Given the imprecision of the time window surveyed in the retrospective evaluation (12 months), a good concordance was generally found between the retrospective and prospective evaluations based on physician-based questionnaires. This is particularly true for items which, in my opinion. were estimated with the greatest accuracy. i.e. the number of consultations, the intensity of drug treatment and the usage of compression. This result is important, because it provides some evidence about the robustness of the results and reassurance that these results are not under- or overestimated. A second feature was the concurrent use of two sources of information : the
investigator and the patient himlherself. It should be stressed that none of the data were based on objective information, e.g., medical notes or database. Therefore, it was necessary to find a way to cross-check the information provided by either source. Of course, these two sources were not independent from one another, because at least part of the treatment used by patients was prescribed by the physician. In spite of this, the cornparison of infonants allowed estimation of whether there is systematic bias in the recording of some items. A good agreement was found for most items, even if the agreement found at the aggregated level was not always found at the individual level or at the level of the clinical categories. Several explanations can be proposed to explain this discrepancy :
different methods of data collection : since patients' intewiews were performed
using phone calls, there was no objective information that could be used; therefore, by protocol, each unit of health services utilisation consumed in relation to venous disorder had to be specified and detailed by the patient, especially in relation to the date (month and year) of its occurrence; this level of detail was necessary to exclude items that were outside the defined time window of 3 months. Although missing data for any of these variables was not a cause of exclusion. it is possible that patients did not recall individual items of resource use. although they would have been able to mention an average value over a defined time period (as requested from the physician). ditferences in perception : the questionnaires requested that only lems prescribed or used in relation to the patient's 'leg problem' be specified. This question rnay have been undentood differently by physicians and by patients; for example, the reason for consultation may have been perceived as related or not to venous disorders by one of the infonnants; moreover, many patients may not have mentioned that a drug was prescribed for hidher venous disorder (in which case the drug would not have been recorded). differences in t h e window surveyed : the timings of interviews and follow-up
visits were independent from one another. For patients who did not have a complete follow-up, it is quite possible that the periods surveyed in the two evaluations did not overlap completely. Globally, the agreement found between the three types of evaluation give some degree of confidence that valid information about resource use has been collected
in the study. 3- Valuation of costs
Costs have been estimated as precisely as possible given the constraints of a study carried out on such a large scale. Data were nevertheless lacking for several parameters and assurnptïons were necersary in order to take them into account in
the valuation process. Given the impact of the cost of surgical procedures other than stripping on the total costs, the lack of knowledge about the nature of the procedure (and about its cost) is the main area of concern. The cost of these other surgical procedures was globally estimated at the same tariff as stripping. This cost may be underestimated because half of the individuals who exper~enced'other surgery' reported a range of 3 to 7 days of hospital stay, which could indicate that they were operated on with a surgical procedure that was more complex (and probably more expensive) than stripping. A more precise evaluation of the nature
of surgical interventions carried out for varicose veins should therefore be a priority in further economic evaluations of the disease. Assumptions were needed for other parameters and should therefore also be better evaluated in the future : the number of sessions of injection of sclerosing agents in
a cure by sclerotherapy, and the actual price paid by the patient for that cure; the number of sessions of physiotherapy included in prescriptions written for patients with varicose veins; the number of sessions of wound care prescribed. Globally, both the intensity of health care utilisation and the estimation of cost can be considered as accurate for the number of consultations, the use of venoactive drugs, drugs with other vascular indications and dressingsfiopical drugs, the use of compression material and the number and cost of stripping procedures. It can also be considered that the total costs estimated for surgery with a venous indication
and for sclerotherapy are underestimated. It should also be noted that only direct rnedical costs were estimated in the study.
Examples of indirect costs that were not considered are the cost of transport to hospitals and the cost of absence from work.
12.5.2 Total costs associated with varicose veins In 1995, the Belgian population aged 20-69 years comprised 6,215 million
inhabitants (IBES, 1999). According to findings of published studies (Chapter 3. Section 3.2.1), the prevalence of clinically significant varicose veins in this population can be estimated at 7% in males and 15% in fernales, with an average of I l % , or 683.650 inhabitants. An estirnate of the proportion of patients with varicose veins seeking treatment in Belgium is lacking. In France, on the basis of a medical consumption survey carried out by a university centre for economic studies (INSEUCESGE), the number of patients seeking treatment was estimated in 1994 at 1.5 to 2 million (Lafuma et al., 1994). This would represent 300,000 to 400,000 patients in Belgium. Approximately 350,000 patients (51% of the estimated number of prevalent cases) would therefore seek treatment at least once a year for varicose veins and incur an average cost of 311 US$, with a total of 109 million US$ per year. From Table 12.10., the cost to the health system would be 204 US$ (311 minus 107) per person. with
a total of 71 million US$ per year. The main
uncertainty in this figure is of course the proportion of individuals with varicose veins seeking treatment. The amount of 71 million US$ per year would have represented 0.64% of the national expenditure on health care in 1995. As the estimation of costs was based on the general practitioner-based sample only, it is likely that the use of resources such diagnostic tests and surgery was underestimated, which would affect the total cost. There are only a few available studies allowing an estimation of whether the results
of this study are consistent with previous ones. Moreover, comprehensive cost evaluations have been made on the basis of aggregate data (e.g. billing data from the compulsory health reimbursement system, drug sales). Unless record linkage between several databases can be perfomed (which is exceptional in Europe) the
@
number of physical units of resource use in these studies cannot be linked to the
indication at the individual level and, a fortion, to the severii of the indication.
On the basis of diagnostic and therapeutic procedures for venous disease obtained from the health insurance system in Belgium, Van den Oever et al. (1995) have estimated the medical cost of "severe varicosis" for the health system at 4.63 billion BEF (125 million US $) for an estimated prevalence of 2.25% in the population of 10 million inhabitants in Belgium. This figure included office visits, Doppler or US-
Doppler, phlebography, hospitalisation (total cost), and venotropic drugs. The cost of other drugs and compression materials were not estimated. Co-payment by patients would amount to 1.89 billion BEF, for a total cost of 6.52 billion BEF (176.2 million US$). On the basis of Van den Oever's prevalence figure (225,000 affected individuals in Belgium). the total cost estimated from the data presented in this thesis would amount to 70 million US$. However, patients included in the VEINES study had on average milder disease than those included in the estimation proposed by Van den Oever et ai. On the basis of the INSEE/CRESGE study, Lafuma et al. (1994) estimated the total cost of varicose vein treatment (not counting sick leaves) at 764 million US$ for the year 1991, with an average of 255 US$ per patient. a figure close to that found in the VEINES study.
4 2.5.3 Variables associated with cos& Table 12.11 gives evidence that the presence of varicose veins substantially increases the medical costs associated with venous disorders after taking into account the effect of several other determinants of costs. The analysis of health
service utilisation ctearly shows that costly procedures such as stripping and other surgery are mostly used in patients with vancose veins that are not complicated by oedema or ulcer. A large proportion of surgical interventions are also perfonned in patients with varicose veins and skin changes. These data may suggest that cosmetic appearance is not the sole reason for operating on patients. prevention of ulcer may also be an important indication.
The
The finding that VEINES-QOL is directly related to expense is a striking result. One could trivially interpret it by concluding that "People consulting their doctor are more likely to be ill". The result has much stronger implications. VEINES-QOL reflects the overall impact of venous disease on the patient's well-being. One important dimension is of course the presence and frequency of symptoms. It also measures the consequence of the disease on social activities and on role limitations (Chapter 11. Table 11.2). It is the combination of the rating of al1 scales of al1 dimensions
that results in a single score. This study found that people with a VEINES-QOL score included in the upper quartile of the distribution of the score consumed one sixth of the amount of resources consumed by people in the lowest quartile of the VEINES-QOL distribution.
More importantly. this resuit was obtained after
adjustment for the effect of other variables generally suspected to be associated with health service utilisation, such as educational status, age and comorbidity. The effect of VEINES-QOL is also independent from varicose veins. which suggests that these two variables measure something different. By suggesting a strong relationship between VEINES-QOL and medical costs in the following year, the result suggests that VEINES-QOL could be an important element to be included in the baseline valuation of clinical and pharmaco-economic studies. The association between VEINES-QOL and expense also adds to the validity of the
VEINES-QOL measure by providing a criterion-based interpretation of its effect. Ware (1994) has used this validation method by presenting analyses of relationships between the SF-36 Physical Component Score (PCS) and Mental Cornponent Score (MCS) and 'criteria'. measured in a survey of the US general population either concurrently or after some period of time. For PCS, these criteria included current work status, subsequent job loss, hospitai stays, doctor visits, probability of a chronic condition, and five-year survival.
The VEINES study
suggests a strong relationship between the VEINES-QOL score and the expense of treatment of the venous disease. This finding will be helpful in the Mure to analyse
in more detaii the relationships between the score and more specific criteria such
as drug treatment, compression or number of consultations.
A.1.1 Surnmary of cost assumptions used in the analysis of health service utilisation in Belgium. Physical Unit
Source
Consultation to generalist
Official tariff (INAMt , 1998)
Consultation to specialist
id.
Consultation to physiotherapist
id.
Day care hospital
biiling office CHU Liege
Hospital stay day
billing office CHU Liege
Doppler test
Official tariff
Duplex scanning
id.
Other diagnostic test
Expert
Stripping (except hospital stay)
Official tariff + billing office
Other venous surgery
id.
Sclerotherapy (4 sessions)
Official tariff + expert
Wound care
Official tariff
Total cost (US 8 ) '
Cost for patient (US$)
Exchange rate used : 100 BEF = 2.70 US$
A. 1.2 Assumptions used in the valuation of costs
1 Unit
Assumption
Drugs
Cost for each package purchased extracted from the medicines formulary (CBIP, 1996)
Compression
Cost per type of compression, brandname and place of purchase estirnated from patients and from manufacturers' price lists
Physiotherapy
According to expert opinion : average of 4 visits to physiotherapist per prescription
Wound care
According to expert opinion : average of 4 nurse visits to patient for wound a r e
Sclerotherapy
Average of 4 sessions per cure; cost of consultation is already considered in costs of consultations.
Other venous surgery
Same cost as stripping was applied.
1
Table A.2.1 One-year adjusted and and gender-standardizednumbers of consultations to different types of physicians for 12 months before the inclusion GP-based evaluationphysician-basedevaluation'
-
Variable
No signs or telangiectasia alone (n=25)
Varicose veins alone (n=81)
Varicose veins and edema (n=35)
W and skin changes (n=144)
W and healed or active ulcer (n=50)
All
patients with W (n-310)
Generat practitioner n patients with z 1 visit % patients with 2 1 visit N visits mean per patient Phlebo/anoeioloaist n patients with 2 1 visit % patients with 2 1 visit N visits mean per patient Surgeon n patients with r 1 visit O h patients with 2 1 visit N visits mean per patient Other swcialist n patients with r 1 visit % patients with 2 1 visit N visits mean per patient Anv specialist n patients with 2 1 visit % patients with 2 1 visit N visits mean per patient
'
Estimates for the category 'All subjects wiM varicose veins' are standardised according to the expected distribution of varicose vein categories in the registration pool
Table A.2.2 One-year adjusted, age and gender-standardized nurnbers of consultations to different types of physicians during follow-up - GP-based sample- physician-based evaluation '
1
Variable
No signs or telangiectasia alone
Varicose veins alone
Va ricose veins and edema (n=34)
W and skin changes (n=143)
W and heated or active ulcer (n=46)
Ali patients with W (n=302)
1 Genenl practitioner n patients with r 1 visit O h patients with 2 1 visit
N visits mean per patient
19 82.3 53 2.30
65 82.3 154 1.95
n patients with 2-1 visit % patients with 2 1 visit N visits mean per patient
Surgeon n patients with 2 1 visit % patients with r 7 visit N visits mean per patient Other smcialist n patients with 2 1 visit % patients with 2 1 visit N visits mean per patient Anv specialist n patients with 2 1 visit O h patients with 2 1 visit N visits mean per patient Phvsiothera~ist n patients with 2 1 visit O h patients with 2 1 visit
'
Estirnates for the category 'All wbjects with varicose veins' are standardiseci according to the expected distribution of varicose vein categories in the registration pool
Table A.2.3 Oneyear adjusted. age and gender-standardized numbers of consultations to different types of physicians dunng follow-up GP-based evaluation- patient-based data
-
Variable
Senerai practitioner 1 patients with 2 1 visit )/O patients with 2 1 visit N visits nean visits per patient Phlebo/anaeiolooist n patients with 2 1 visit % patients with r 1 visit N visits mean per patient Sumeon n patients with 2 1 visit % patients with r 1 visit N visits mean per patient Other specialist n patients with 2 1 visit % patients with z 1 visit N visits mean per patient Anv specialist n patients with 2 1 visit % patients with 2 1 visit N visits mean per patient Phvsiothera~ist n patients with 2 1 visit % patients with 2 1 visit N visits mean per patient Othet health professional n patients with r 1 visit % patients with 2 1 visit N visits mean per patient
No signs or telangi- Varicose ectasia veins alone alone (n=24) (n=8l)
Varicose veins
and
edema (n=35)
W and skin changes
W and healed or active ulcer (n=143) (n=49)
AI1 patients with W (n=308)
Table A-2.4 Age and gender-standardized numben and proportions of patients in whom diagnostic tests have been perfonned any time before the inclusion visit in Belgium GPs-based sample - Physician-based evaluation
-
No signs or telangiectasia alone (n=25)
Variable DoDD~~~ n patients with 2 1 test
Varicose veins alone (n=81)
Vancose veins and edema (n=35)
1 4.0
8 9.9
5 14.3
14 9.7
4
8.0
32 10.3
2 8.0
6 7.4
7 20.0
12 8.3
5 10.0
30 9.7
3 12.0
13 16.0
9 25.7
22 15.3
8 16.0
53 17.1
O
O
O
O
O
O
1
1 2.9
4 2.8
1 2.0
7 2.3
W and skin changes (n=144)
W and healed or active ulcer (n=50)
All patients with W (n=3lO)
C.W.
Yo duplex n patients with 2 1 test
'?40 D o ~ ~ lor e rduplex n patients with r t test O h phleôo~ra~hy n patients with 2 1 test O h other test n patients with 2 1 test
C.W.
Oh
3 12.0
1.2
Table A.2.5 Age and gender-standardired numbers and proportions of patients in whom diagnostic tests have b e n performed during follow-up in Belgium GPs-based sample Physician-basedevaluation '
-
-
- -- -
Variable c-W.Doppler n patients with r 1 test
-
%
-
--
- --
-- -
Variwse Varicose veins veins and alone edema (n=79) (n=34)
W and skin changes (n=143)
W and healed or active ulcer (n=46)
All patients with W (n=302)
4 5.1
2 5.9
7 4.9
1 2.2
14 4.6
O
5 6.3
3 8.8
11 7.7
1 2.2
21 7.0
O
6 7.6
3 8.8
16 11.2
1 2.2
27 8.9
3 13.0
2 2.5
O 0.0
3 2.1
%
other test n patients with 2 1 test
-
O
%
du~lex n patients witti r 1 test % C.W.Domter or du~lex n patients with 2 1 test
- ---
No signs or telangiectasia alone (n=23)
O
6 2.0
Table 142.6 Age and gendec-standardized numben and proportions of patients to whom dnigs have been prescribed for CVDL over 12 months before inclusion visit- Belgium GPs-based sample Physician-basedevaluation
-
Dnig ciass
No signs or telangiectasia alone (n=25)
Varicose veins alone (n=8l)
Varicose veins and edema (n=35)
W and skin changes (n=144)
W and healed or active ulcer (n=50)
All patients withW (n=U0)
1 Anv venoactive dniq n % Other vascular druq n Oh Dressinqs, local care n
12 48.0
45 55.6
24 68.6
85 59.0
29 58.0
183 59.0
4
10 12.3
5
16.0
14.3
23 16.0
20 40.0
53 17.1
O
1
O
1 0.7
8 16.0
7 2.3
1.2
O/o
1 Other dnia wssiblv related
l b'
to CVDL
Table A.2.7 Unadjusted, age and gender-standardired nurnbers and proportions of patients to whom drugs have been prescribed at least one during follow-up for CVDL GPs-based sample Physicianbased evaluation
-
Variable
Anv venoactive druq
Other vascular druq
Dressinas. to~icâldrucls
Other dmcr possiblv related ta CVDL n %
No signs or telangiectasia alone (n=23)
Varicose veins alone (n=79)
Varicose veins and edema (n=34)
W and skin changes (n=143)
-
W and healed or active
ulcer
(n=46)
All patients with W (n=302)
Table A.2.8 Unadjusted, age and gender-standardüed numbers and proportions of patients to whom dnigs have been prescribed for CVOL during follow-up- GPr-based sample - Patient-based evaluation
Variabie
No signs or telangiectasia alone (n=24)
Varicose veins alone (n=8l)
Varicose W veins and and skin edema changes (n=35) (n=143)
W and healed or active ulcer (n=49)
All patients with W (n=308)
Anv venoactive drug n patients with 21 punhase O h patients with r1 purchase n packages bought mean per patient Other vascuiar dniq n patients with 21 purchase % patients with 21 purchase n packages bought mean per patient Dressinos. to~icaldruqs n patients with 21 purchase % patients with 21 purchase n packages bought mean per patient Other druq ~ossiblvrelated to CVDL6 n patients with 21 purchase % patients with 21 purchase n packages bought mean per patient 1 One patient (ID 699) with extremely high value is excluded 2 One patient (ID 829) with extremely high value is excluded
Table A.2.9 On-year adjusted, age and gender-standardized number of person-days of treatment for several classes of drugs prescribed by investigators for CVDL GPs-based sampte - Physician-based evaluation ' --
Variable
n patients with data N person-ûays of treatment mean no of person-days
-
-
No signs or tetangiectasia atone (n=23)
9 1681 73.09
Analaesics n patients with data N person-days of treatrnent mean no persondays
0.04
Diuretics n patients with data N persondays of treatment mean no persondays
436 18.96
NSAIDS n patients with data N person-days of treatment mean no person-days
26 1.13
1 1
4
4
-
--
Varicose veins alone (n=79)
-
Varicose veins and edema (n=34)
W and skin changes (n=143)
W and healed or active ulcer (n=46)
All patients with W (n=302)
Table A.2.10 Unadjusted, age and gender-standardized numbers and proportions of patients using compression at the time of inclusion - Belgium - GPs-baseb sample - Physician-based evaluation
Drug class compression - bandage n ?40 compression - stockings n
Yo
compression - tights n O h
compression - any n %
No signs or teiangiectasia alone (n=25) O
1
4.0 3 12.0 3 12.0
VaBcose veins alone (n=81)
Varicose veins and ederna (n=35)
1 1.2
1 2.9
6 7.4
3 8.6
6 7.4 14 17.3
W and skin changes (n=144)
W and healed or active ulcer (n=50)
Ail
patients with W (n=3lO)
6 12.0
10 3.2
11 22.0
36
13.2
8 22.9
6 4.2
10
30
20.0
9.7
12 34.3
28 19.4
24 48.0
73 23.6
4
2.8 19
11.6
Table A.2.11 Unadjusted. age and gender-standardized numbers and proportions of patients having used compression during follow-up Belgium GPs-based sample Physician-basedevaluation
-
Drug class compression n % compression n % compression n % compression n
No signs or telangiectasia alone (n=23)
- bandage O
- stockings
- tights
1 4.3 O
- any 1
Yo
4.3
-
-
Varicose veins alone (n=79)
Varicose veins and edema (n=34)
W and skin changes (n=143)
W and healed or active ulcer (n=46)
2 2.5
1 2.9
6 4.2
5 10.9
12 4.0
9 11.4
7 20.6
21 14.7
8 17.4
45 14.9
6 7.6
11 32.4
22 15.4
5 10.9
44
15 19.0
14 41.2
42 29.4
16 34.8
84 27.8
Al l patients with W (n=302)
14.6
Table A.2.12 Unadjusted, age and gender-standardized nurnbers and proportions of patients having used compression during follow-up GPs-based sample Patient-based evaluation
-
-
Varicose veins alone (n=81)
Varicose veins and edema (n=35)
W and skin changes (n=143)
W and healed or active ulcer (n=49)
All patients with W (n=308)
O 0.0
1 1.2
2 5.7
3 2.1
3 6.1
8 2.6
compression stockings n
1
%
4.2
8 9.9
7 20.0
20 14.0
7 14.3
41 13.3
3 12.5
9 11.7
8 22.9
28 19.6
4 8.2
50 16.2
4 16.7
15 18.5
16 45.7
47 32.9
14 28.6
90 29.2
Drug class
No signs or telangiectasia alone (n=23)
-
compression bandage n Oo /
-
-
compression tights n O h compression any n O h
-
e
Table A.2.13 Age and gender-standardized numbers and proportions of patients having used other types of treatment before or at the time of the inclusion visit GPs-based sample - Physician-based evaluation
-
No signs or telangiectasia alone (n=25)
Variable sclerothera~v' n
15 30.0
O 0.0
O 0.0
O 0.0
O 0.0
O 0.0
8
9.9
2 5.7
6 12.0
25 8.1
O
O
12 24.0
8 2.6
9 11.1
O 0.0
O
phvsiothera~v* n patients
17
2t .O
Oh ulcer care n patients %
O
' Any time before inclusion visit;
Al1 patients with W (n=310)
25 17.4
3 12.0
%
W and healed or active ulcer (n=50)
5 14.3
Oh
'
W and skin changes (n=144)
10 20.0
5 20.0
other venous surqerv n
Varicose veins and edema (n=35)
35 24.3
Oh
strippinn ' n
;
Varicose veins alone (n=81)
6 17.0
10
6.9 O
68 21.9 51 16.5
At the time of the inclusion visit
Table A.2.14 Age and gender-standardued numben and proportions of patients having used other types of treatment dunng follow-up GP-based sample Physician-basedevaluation
-
Variable
sclerotherapy n
Yo strippinq n % other venous surgery n % phvsiothera~y n patients % ulcer m e n patients %
No signs or telangiectasia alone (n=23)
-
Varicose veins alone (n=79)
1 4.2
7 8.9
O
4 5.1
O
4 5.1
O
O
O
1 1.3
Varicose veins and edema (n=34)
W and skin changes (n=143)
W and healed or active ulcer (n=46)
All patients with W (n=302)
13 9.1
2 4.3
23 7.6
O
13 9.1
O
17 5.6
1 2.9
11 7.7
O
16 5.3
4 11.8
7 4.9
1 2.2
12 4.0
O
1 0.7
8 17.4
8 2.7
1
2.9
Table A.3.1 One-year adjusted age and gender-standardizednumbers of consultations to various types of health professionals during follow-up -0elgium - specialists-based sample patients-based evaluation
Variable
3eneral practitioner r 1 visit HI patients with r 1 visit U visits nean per patient 1 patients with
r 1 visit patients with 2 1 visit N visits mean per patient
1 patients with
)/o
Surqeon n patients with r 1 visit % patients with 1 1 visit N visits mean per patient Other s~ecialist n patients with 2 1 visit O/O patients with L 1 visit N visits mean per patient Anv specialist n patients with 2 1 visit O/O patients with 2 1 visit N visits mean per patient
Phvsiotheraoist n patients with 2 1 visit O h patients with 2 1 visit N visits mean per patient Other paramedic n patients with 2 1 visit % patients with 2 1 visit N visits mean per patient
No sign or telangectasia alone (n=31)
Varicose veins alone (n=26)
W and skin changes or edema (n=25)
W and healed or active ulcer (n=15)
-
All subjects with W (n=66)
Table A 3 2 Age and gender-standardized numben and proportions of patients in whorn diagnostic tests have been perfomed any time before the inclusion visit in Belgium Specialistsbased sample Physician-basedevaluation '
-
-
Variable
Do~pler n % du ~ l e x n % C.W. Doppler or duolex n
W and skin changes or edema (n=26)
No sign or telangectasia alone (n=31)
Varicose veins alone (n=26)
W and healed or active ulcer (n=15)
6 19.4
4 15.4
7 26.9
1 7.1
13 19.4
12 36.5
9 34.6
8 30.8
8 53.3
25 37.3
18 58.1
13
14
50.0
53.8
2 6.5
1 3.8
11.5
All subjects with W (n=67)
C.W.
%
other test n O/o
'
3
9
60.0
O
36
53.7 4 6.0
Estimates for the category 'al1 subjects withh varicose vens are standardised according to the cortected distribution of the varicose vein categories in the registration pool
Table A.3.3 Age and gender-standardizedand proportions of patients in whom diagnostic tests have been perfonned during follow-up -6elgium specialists-based sample -physician-based evaluation
-
Variable
Do~~ler n patients with r 1 test
No sign or telangectasia alone (n=29)
W and skin changes or edema (n=24)
Varicose veins alone (n=24)
W and healed or active ulcer (n=14)
Al l subjects with W (n=62)
C.W.
%
duplex n patients with 2 1 test O h C.W. D o ~ ~ lor e rdu~lex n patients with r t test O h other test n patients with r 1 test %
2 6.9
2 8.3
O
2
O
2 8.3
6.9 4
13.8
O
2 8.33
9 37.5
O -
2 8.3
10 33.3
11 17.7
O
2 3.2
1 7.14 O
13 21.O 2 3.2
Table A.3.4 Age and gender-standardized numben and proportions patients to whom dnigs have been prescribed for CVDL during a period of 12 months ûefore the inclusion visit in Belgium - Specialists-basad sample Physician-based evaluation '
-
Variable
Anv venoactive drug n %
Other vascular druq n O h Dressin~s.local care n
No sign or telangectasia alone (n=3 1) 9 29.0 O
O
W and healed or active ulcer (n=15)
Al1 subjects with W (n=67)
9 34.6
5 33.3
24 35.8
1 3.8
2
5 33.3
O
O
O
O
3 12.0
9 34.6
2 15.1
15 22.4
Varicose veins alone (n=26)
11 42.3
W and skin changes or edema (n=26)
7.7
7 10.5
%
Other dnia ~ossiblvrelated to CVDL n %
2 6.3
' Estimates for the category 'al1 subjects with varicose veins' are standardised according to the conected distribution of the varicose vein categories in Me registration pool
Table A.3.5 Unadjusted. age and gender-standardued numben and proportions of patients to whom drugs have been prescnbed for CVDL during follow-up- in Belgium Specialists-based sarnple Physician-based evaluation '
-
-
--
W and healed or active ulcer (n=14)
No sign or telangectasia alone (n=29)
Varicose veins alone (n=24)
Anv venoactive drug n patients with 21 presct. % patients with 21 prescr.
6 20.7
7 29.2
3 12.5
6 42.9
15 24.2
Other vascular druq n patients with 11 prescr. % patients with 21 prescr.
2 6.9
O
1 4.2
3 21.4
4
O
O
Variable
Dressinqs. local care n patients with 11 prescr. % patients with 21 prescr. Other drua ~ossiblv related to CVDL n patients with r l prescr. % patients with 21 prescr.
'
W and skin changes or edema (n=24)
O
3 21.4
1 3.4
2
8.3
4
16.7
4
28.6
Ali subjects with W (n=62)
6.5
3 4.8
10 16.1
Estirnates for the category 'al1 subjects withh varicose veinsnare standardised according to the corrected distribution of the varicose vein categories in the registration pool
Tabk A.3.6 Unadjusted. age and gender-standardized numben of packages purchased for CVDL by patients during follow-up in Belgium Specialists-based sample Patient-based evaluation'
-
-
No sign or telangecbsia alone (n=31)
Variable
-
-
Varicose veins alone (n=26)
W and skin changes or edema (n=25)
Anv venoactive druq n patients with 21 purch. O h patients with ri purch. n packages bought mean per patient
'
Other vascular dniq patients with r l purchase % with r i purchase n packages bought mean per patient Dressinas. local care patients with r 1 purchase % with 21 purchase n packages bought mean per patient Other druq ~ossiblk related to CVDL patients with rl purchase OO / with 21 purchase n packages bought mean per patient 1 One patient (ID 3555) was excluded because of extremely high values
W and healed or active ulcer (n=15)
AH subjects with W (n=66)
Table A.3.7 One-year adjusted, age and gender-standardized numbers of persondays of treatment for several classes of dnigs prescribed by investigators for CVDL during follow-up- in Belgium - Specialists-based sample Physician-basedevaluation '
-
Variable
1 1
No sign or telangectasia alone (n=29)
Varicose veins alone (n=24)
W and skin changes or edema (n=24)
W and healed or active ulcer (n=14)
All
subjects with W (n=62)
Anv venoactive dniq n patients with data persondays of treatment mean no of persondays Analaesics n patients with data persondays of treatment mean no person-days Diuretics n patierits treated person-days of treatment mean no persondays NSAIDS n patients treated person-days of treatment mean no persondays
'
Estimates for the category "al1 subjects withh varicose veins" are standardised according to the corrected distribution of the varicose vein categories in the registration pool
Tabie A.3.8 Age and gender-standardized numben and proportions of patients using compression a the time of the inclusion visit in Belgium - Specialists-based sample Physician-based evaluation '
-
Variable
No sign or telangectasia alone (n=31)
Varicose veins alone (n=26)
W and skin changes or edema (n=26)
W and healed or active ulcer (n=15)
All patients with W (n=67)
3 11.5
8 53.3
10 t 4.9
-
compression bandacies n O! 1 compression stockinas
-
1 3.2
O
1 corn~ression- tiahts % com~ression any n
-
a/0
3.2
11.5
11.5
20.0
13.4
1 3.2
4
9 34.6
12
15.4
80.O
25 37.3
Table A.3.9 Unadjusted, age and gender-standardized numbers and proportions of patients usinf compression during follow-uo in Belgiurn Specialists-based sample Physician-based evaluation '
-
Variable
No sign or telangectasia alone (n=29)
Vancose veins alone (n=24)
-
W and skin changes or edema (n=24)
W and healed or active ulcer (n=14)
All subjects with W (n=62)
-
compression bandages n
Yo
1 com~ression- stockinos 1 compression - tiahts
1 compression - anv
O
O
2 8.3
7 50.0
8 12.9
-
Table A.3.10 Unadjusted, age and gender-standardized numbers and proportions of patients usinf compression during follow-up in Belgium Specialists-based sample - Patientbased evaluation
-
Variable
No sign or telangectasia alone
(n=31) com~ression- bandacres n
O
%
com~ression- stockinas n
Varicose veins alone (n=26) 2
7.7
W and skin changes or edema (n=25)
1 4.0
W and healed or active ulcer (n=15)
3 20.0
O h
1 3.4
12.5
2 8.3
%
6 19.4
7 26.9
20.0
6.7
6
9 34.6
6 24.0
53.3
com~ression- tiahts n com~ression- anv n
19.4
3
5
3
All subjects with W (n=66) 6 9.7
8
21.4
12.1
1
14 21-2
8
22 33.3
Table A.3.11 Age and gender-standardized numben and proportions of patients having used other types of tfeatment before the inclusion visit in Belgiurn Spectalists-based sample Physician-basedevaluation '
-
-
--
No sign or telangectasia alone (n-31)
Variable
sclerothera~v n O h strippinq n O h other venous surqey n
Varicose veins alone (n=26)
W and skin changes or edema (n=26)
W and healed or active ulcer (n=15)
Al1 subjects with W (n=67)
*
%
phvsiotherapy n % ulcer care n % Estimates for the category 'al1 subjects withh varicose vens are standardised according to the conected distribution of the varicose vein categories in the registration pool. Any time before inclusion visit. At the time of the inclusion visit
'
Table A.3.12 Age and gender-standardizednumbers and proportions of patients having used other types of treatment during follow-up Belgium Specialists-basedsample Physician-based evaluation
-
Variable
1
sclerothera~v n patients % stripping n patients % other venous sume4 n patients O h ulcer care n patients O!
No sign or tetangectasia alone (n=29)
Varicose veins alone (n=24)
-
-
W and skin changes or edema (n=24)
11 37.9
8 33.3
10 41.7
O
O
2 8.3
W and healed or active ulcer (n=14)
All subjects with W (n=62)
5 35.7
23 37.1
3
5 8.1
21.4
@
13. Conclusions This thesis aimed to investigate several epidemiological aspects of varicose vein disorders and evaluate specific patient-based outcomes, including aspects relating to quality of life and economic impact. The etiology of varicose veins was assessed from different viewpoints. previously described risk indicators were either confimied or reassessed, important patient-based outcomes have been reported and determinants of these outcomes were identified. Although based on empirical data, this re-assessrnent of a pathology was carried out in a relatively short period of time : five years have elapsed between the start of the literature review and the completion of the report. The use of a cohort population was an essential component of this achievement. It allowed study of a patient population from different perspectives and using different rnethodologies, including investigations based on cross-sectional, case-control and cohort designs. However, the study also illustrated the difkulties that can be encountered when using a cohort based on prevalent cases, and with multiple endpoints for descriptive and etiological purposes. Before concluding on the main results reported in this thesis, it seerns to be appropriate to present several of the methodological aspects encountered in the VEINES cohort study.
13.1 Methodologicalaspects 13.1.1 Strengtbs of the VEINES cohort design
Com~rehensiveassessrnent of outcomes The VEINES study is unique due to the weaRh of data 1 offen on outcornes related to varicose veins. The follow-up period was a key element in this respect, as a crosssectional study would not have suffced to answer the questions raised by the objectives of the study. Many examples illustrate this argument. Follow-up was needed to assess complications of the disease, and to compute incidence rates of the more severe stages of venous disorden such as ulcer. The population experience was also needed to
develop and validate the specific quality of life instrument in a population-based sample. The study has also shown the correlation between quality of life scores at baseline and resource use in the following year. Needless to Say, a close population experience (with interviews every three months) was a prerequisite in the assessment of economic outcomes. In the field of venous disease, medical practice varies between countries and the availability of different forms of treatment influences patterns of health services utilisation. Outcornes were therefore also expected to Vary from country to country. Comprehensive assessment of outcornes and acquisition of knowledge on the disease required the presence of sufficient heterogeneity in the study population. The scope of the VEINES study necessitated the setting up of a multi-centre multicountry organisation. which is unprecedented for an epidemiological study on venous disorders. Reclister of ~atientsas source ~ o ~ u l a t i o n The selection of study subjects from a register of consecutive patients consulting physicians for venous complaints had several advantages. A major strength in the design was in the explicit definition of the source population which gave rise to groups of patients included in etiological studies. Therefore, there was no need to conceptualise a source population from which cases and controls (in a case-control study) would be valid representatives. By definition. both arose from the same population. Furthemiore. the source population could be directly used as a target population to which results could be inferred (provided the sampling fraction was taken into account), even if the extrapolation of the results needed not be restricted to that target population. This is analogous to using a primary base as a source of subjeds in a case-control study. In practice, the strength in this design was also to allow the assessment of possible selection biases introduced by investigaton in the recruitment of study subjects from the registration pool. It can be argued that the source population used in this work was not representative of the 'general' population. In the context of venous disorders. the issue of representativity
should not be overestimated due to the large proportion of subjects affected by venous disorders in the general population. Moreover, representativity is not an issue to be considered in drawing inferences. What matters is the validity of the association with respect to the source population. lnference to other populations is a matter of personal judgement. Ambis~ectivedata collection Data collection included the period that preceded inclusion into the study as well as the follow-up phase. The strength of this ambispective data collection was obvious for assessing medical practice and health services utilisation. Since participation in the study could have interfered with the usual patterns of treatment, retrospective information allowed validation of estimates of resource use collected prospectively. This validation was not amenable to statistical testing because differences between estimates pertaining to the study and the 'pre-study' periods were expected to differ. Therefore, the conclusion about the validity of prospective data is a matter of personal judgement on the expected size of the differences. For example, it can be expected that the proportions of patients using venoactive medication is comparable in the two periods. whereas the proportions of patients operated on or having veins sclerosed would be related to the reason for consultation and would increase during the follow-up period. The data give confidence to the idea that patterns of treatment were not severely influenced by the constraints imposed by the protocol.
13.1.2 Weaknesses of the VEINES cohort study
Descri~tivevs. etioloaical studv The development of a cohort study is often justified by the lack of essential descriptive data about a pathology. In many cases, as in the VEINES study, the objectives of a cohort study comprise of the description of a health phenornenon, as well as the search for its cause. In principle, these two objectives are contradictory, because the conditions needed to ensure validity of the results may differ for these two objectives.
Furthermore. the observation of outcomes such as quality of life or medical practice requires an undisrupted research environment. The difficulty was evident in the case of the VEINES study as the health phenomenon was a sïaïe rather than an evenï, with a slow onset time, a long-duration and little variation with time. The investigation of an event has the advantage of having a direct time referent allowing determination of the time sequence between event and exposure. This property is the cornerstone of an empirically-based conceptual model of causation. and consequently of the elaboration of robust epidemiological designs such as a nested case-control study. The lack of time referent in studies based on prevalent cases has a major untoward consequence regarding
etiological inference : information biases.
These have been alluded to above in several chapters where they had to be considered for interpretation of the data. These information biases took several foms. Misclassification of disease status- The presence of varicose veins at the baseline
physical examination could not be used to determine the 'case' status in a case-control study on risk indicaton of varicose veins. Firstly, the study of the validity of the diagnosis by general practitionen showed a poor specificity, indicating that half of individuals without varicose veins would be wrongly classified as cases; furthermore. validity depended on the presence of suspected deteminants of varicose veins. Secondly, some individuals without varicose veins had veins surgically removed or sclerosed in the preceding year and would therefore be wrongly classified as 'controls'; the probability of surgical intervention or sclerotherapy also depended on the presence of suspected detenninants of varicose veins.
This problem violated one of the
assurnptions needed for validation of the Prevabnce Odds Ratio to estimate the Incidence Density Ratio.
The only available remedies to disease misclassification were restriction of the study population to patients with varicose veins diagnosed by specialists, and restriction of the control series to individuals without previous venous surgery or sclerotherapy of the leg. No restriction was used in descriptive analyses based on cross-sectional relationships,
@
such as association of varicose veins wlh quality of or life health service utilisation. In
these cases, it is the simuHaneous absence or presence of varicose veins (Le. diagnosis of varicose veins) and of the outcome that rnattered.
.
Misclassification of exposure- Contrary to an exposure cohort where entry is conditional on exposure status, a cohort where entry is conditional on a health state shares the same methodological problems for the ascertainment of exposure as case-control studies. The two main difficulties for etiological research encountered in this work were the definition of a time window of exposure that preceded the onset of varicose veins and the accurate assessment of exposure. The first problem was complicated by the long period of time elapsed since the onset of varicose veins (a disease duration of more than 5 years was declared by more than 80% of subjects) and the dificulty of explorhg whether an exposure occurred before the onset date; this was the case, for example, in the investigation of the relationship between leg surgery and leg trauma. For other variables. such as smoking status, it was necessary to assume that current exposure reflected exposure before the onset of varicose veins. The second problem (accurate assessment of exposure) was complicated by the lack of objective information and the likelihood of recall bias for some variables such as past thrombophlebitis. Multiple endpoints meant extensive data collection Each outcome had its own requirement for data not only as prirnary endpoints but also as potential confounding variables which needed to be taken into account when investigating associations. Although the data collection was based on several sources, the accumulation of information needed had two conseguences : 1) The number of items collected for each endpoint had to be limited and some
hypotheses could not be investigated in depth. There is no doubt that some objectives could have been more thoroughly examined la specific epidemiological study had been carried out. Some examples can illustrate this point. Some variables, such as hormonal factors, could not be investigated in the multivariate analysis of risk indicaton because this information was only collected in the general practlioner-based sample. Furthemore, the uncertainty regarding the indication of sclerotherapy performed in the
past led to exclusion of a large number of subjects from the control group and decreased the power of the analysis. In the econornic analysis, limited information was requested from investigators about surgical procedures, which, although they were performed in only a small nurnber of patients, proved to be one of the major sources of costs. An ad-hoc study on the cost of varicose veins would have implied access to medical records for the computation of actual resource use. Limited information was also available about the number of sessions of sclerotherapy perforrned on patients who had varicose veins sclerosed, and cost estimates had to rely on assumptions regarding this parameter. 2) The length of questionnaires to be filled in by physicians and the duration of the
patient interviews led to incomplete data collection for some patients or physicians who could not sustain long interviews or questionnaires. It was explained in Chapter 8 (Risk indicators for varicose veins) that the number of missing values for important variables such as educational level differed by caselcontrol status; confounding could have been introduced if the distribution of missing values differed from that observed in subjects with complete data. Fixed cohort ~ o ~ u l a t i omeant n fixed sam~lesize Due to the multiplicity of endpoints and the issues of sample size raised by the assessrnent of econornic and quality-of- life outcomes, the sample size was estimated on the basis of assurnptions rather than on fixed rules of sample size determination (Chapter 4, Section 4.4.3). All analyses were therefore carried out with a fixed study population. This feature led to a lack of power when restriction of the study population proved necessary to ensure the validity of the results. This restriction was necessary in two situations :
in the analysis of risk indicators associated with varicose veins
(restriction of the case group to patients with varicose veins diagnosed by physicians) and in the analysis of variables associated with venous reflux (restriction of the study population to women without history of thrombophlebitis). In these situations, specific studies could have ensured that an adequate sample size was attained.
13.2 Mein findings
Conclusion #l. From the liteiature, prevalence of varicose veins in western populations is between 25% and 35% in women and between 10% and 20% in men. Current estimates of prevalence are based on Amoldi's definition of varicose veins. which is amenable to different interpretations, in particular as regards the inclusion or exclusion of small varicosities. Prevalence estimates of 25%-35% in women and 10%20% in men rely on population-based studies with comparable survey methodologies.
About 50% of these prevalence figures concern a disorder associated with clinicallysignificant symptoms and signs of more severe venous disorders. The increase of prevalence with age seems to be due to an accumulation of cases rather than on an increase of incidence. Conclusion #2. Diagnosis of varicose veins by genenl pnctitioners has a low specificity; it is associated with misclassification of telangiectasia and influenced by prior expectation.
Sensitivity and specificity estimates of 0.93 and 0.45 were found in comparison with the diagnosis made by specialists. An important source of misclassification appean to be in the recognition or reporting of telangiectasia which seem to be either unrecognised or misclassified as varicose veins. Misclassification is observed according to reproductive history, history of past thrombophlebitis, history of leg trauma and presence of skin changes (including ulcer). Conclusion #3. In women aged 18-15 yean consulting physicians, the main risk indicators of presenting with varicose veins slone are age, pregnancy and hmily history of varicose veins. A case-control analysis compared women with varicose veins alone diagnosed by
specialists and wornen without signs or with telangiectasia alone recruited by general
practitioners or specialists.
A large number of variables were introduced into a
multivariate logistic regression model either as variables of interest or as potential confounding variables. Association of varicose veins with pregnancy was the strongest effect found in this study. with an odds-ratio of 2.96 ( 95% CI : 1.02-8.64) in women with at least one pregnancy. It is also the most original contribution of this thesis regarding risk indicators of varicose veins. The odds-ratio of 3 suggests that the association of varicose veins with pregnancy may be stronger than generally thought on the basis of the Framingham Study. This thesis has also shown that the weak association found in Framingham could be due to an improper definition of exposure categories. association with varicose veins was found for obesity.
No
This in contrast to the
Framingham study where a small (1.4) but statistically significant association was reported only in women. Other results confin previous findings or are dimcult to interpret due to the possibility of recall bias. Family history was associated with an odds-ratio of 2 to 4 accotding to the variable used to indicate family history. The risk of varicose veins also increased with age. No association was found for blood group A, smoking or history of thrombophlebitis. An odds ratio of 2.08 (95%CI: 0.72-5.87) was found for past thrombophlebitis when patients with varicose veins. oedema, skin changes or ulcer were included in the case groups.
Conclusion M. Reflux assessed by duplex was present in 44.4Oh of subjects without signs or with t.langiectasia alone, in 95.7% of subjects with varicose vein alone and in 89-100°h of subjects with varicose veins with oaier signs of venous
disoidem. The distribution of superficial reflux is compatible with the hypothesis
of a distal onset of reflux. Deep venous reflux and superficial venous reflux were found in 17.3% and 38.3% respectively of subjects without signs or w l h telangiectasia alone. in 28.6% and 90% respectively of subjects with varicose veins alone and in 52.86% and 78.89% respectively of subjects with varicose veins associated with other sign of venous
disease. The proportions of patients with reflux in the deep system alone increased with
the severity of signs associated with varicose veins (6% to 14%), whereas an opposite trend was observed for superficial reflux alone (67% to 14%). In subjects with unilateral varicose veins, superficial and deep reflux were restricted to the ipsilateral leg in respectively 73.6% and 72.0% of patients with reflux. In the majority of legs. superficial reflux was found in the long saphenous vein alone or in combination
with reflux in the short saphenous vein. In subjects with varicose veins alone, reflux at the saphenous-femoral junction accounted for only 8.1% of legs with superficial reflux alone; this figure was 27.8% in subjects with varicose veins and other signs of venous disease. Conclusion #S.
Although pregnancy is associated with vsricose veins, this
association does not appear to be mediated by the increased iisk of developing thrornbophlebitis and venous incornpetence. Based on current evidence, the main etiological factor for varicose veins r e m s to be distension of the vein wall related to a genetic predisposition.
The odds ratio of the association between pregnancy history and presence of reflux was 1.37 (65% CI: 0.68-2.79). No iaterality of varicose veins or reflux was found, which contradicts the theory of an increased risk of thrombophlebitis in the left leg associated with a gravid uterus. The chain of events that would lead from pregnancy to increased risk of thrombophlebitis to onset of refiux and to the onset of varicose veins remains to be substantiated. This thesis did not identify an explanation for the increased risk of varicose veins in pregnancy. The role of genetic predisposition is supported by the association of varicose veins with family history of varicose veins, by the evidence that reflux has a distal origin in most cases and by the lack of association between pregnancy and reflux. Moreover. it has been shown that reflux is absent in 5% of patients wlh varicose veins. which supports the hypothesis that vein wall distension is the primary lesion leading to the development of varicose veins.
Conclusion #6. A high incidence rate of ulcention (4 per 100 person-years) was found in patients with varicose veins and was not influenced by the piesence of other signs of venous disorders. A high incidence of thrornbophlebitis was also obsewed (6.0 per 100 penon-yean) and increased with the severity of signs associated with varicose veins. The incidence density of ulceration in al1 subjects with varicose veins was 3.7 per 100 person-years in patients seen by specialists and 6.2 per 100 person-years in patients seen by general practitioners. There was no consistent trend with the severity of signs (oedema or skin changes) associated with varicose veins. High incidence rates were also found in patients without signs or with telangiectasia alone. The incidence rate of thrombophlebitis was similar in both general practitioner-based and specialist-based samples (6.0 per 100 person-years). In specialist patients. it ranged from 1.3 to 12.9 according to the severity of signs associated with varicose veins. Skin changes and oedema occurred at a very high rate in both samples, but the clinical significance of this finding is difficult to determine without further investigations.
Conclusion #7. A high prevalence of symptoms (especially heaviness and pain) was found in ail clinical categories and in both genders. It increosed with the severity of signs associated with varicose veins. Varicose veins had no effect of a symptom score after adjustment f o i other suspected determinants on
symptoms. Age-standardised prevalence of primary symptoms (heaviness. pain or swelling) in men and women was 78% and 76% respectively in subjects without signs or with telangiectasia alone, and 83% and 87% respectively in subjects with varicose veins. There was no association of symptoms with specialist-rated severity of varicose veins. Although varicose veins were associated with symptoms in bivariate analysis, this association disappeared when other variables were adjusted for, such as age and body mass index.
Conclusion W. Varicose veins alone had no impact on generic and diseasespecific quality of life scores when the effect of detenninants of quality of life was adjusted for. Mean SF-36 MCS and ?CS scores in patients with varicose veins were 46.1 and 45.6 respectively in men and 43.6 and 44.2 respectively in women. In contrast to MCS scores, ?CS scores decreased with the severity of signs associated with varicose veins.
No association was seen between varicose veins alone and PCS in multivariate analysis. In a disease-specific quality of life (QOL) measure developed in the course of the VEINES study, the mean score was 51.3 in subjects with varicose veins alone and decteased with the severity of signs associated with varicose veins, going down to 45.4 in subjects with varicose veins and active ulcer. In analysis of covariance, varicose veins had a very small effect on the score when adjusted for the presence of venous oedema. ulcer, body mass index r 30 and CO-morbidity. These results suggest that symptoms and presence of varicose veins are different outcornes. They have implications for identifying proper evaluation criteria for varicose vein treatrnent.
Conclusion #9. Except for the use of venoactive drugs, health service utilisation in Belgium increased with the severity of signs associated with varicose veins. One-year total direct medical costs for varicose veins in the genenl-pnctitioner based sample in Belgium avenged 311USS per patient. In the general practitioner-basedsample. a patient with varicose veins made an average
of 2 to 4 visitslyear to a general practitioner and 0.5 visitslyear to a specialist. About 5% and 7% had Doppler and duplex tests perfonned. The proportions of patients treated at ieast once during follow-up by venoactive drugs, other dmgs with a vascular indication and topical drugs or dressings were approximately 60%, 20% and 3% respectively. The
mean number of person-days of treatment with venoactive drugs over one year was about 80 days. Compression was used by about 19% of patients with varicose veins.
Sclerotherapy. vein stripping and other surgery were experienced by about 8%. 6% and
5% respectively of patients during followup. The main differences observed in the specialist-based sample were the larger proportions of subjects in whom diagnostic tests were perfonned, the lower level of drug use. the higher level of utilisation of compression, and the higher proportions of subjects who had sclerotherapy or venous surgery. As surgical interventions represented the main determinant of cost, patients with varicose veins alone (305 US$) and patients with varicose veins and skin changes (376s) are those with the highest costs. A cost of 135 US$ was found in subjects
without signs or with telangectasia alone. Within the total cost, the cost to patients tended to increase with the severity of signs associated with varicose veins. About one third of the costs were out-of-pocket expenses. Conclusion M O . Variables most strongly associated with direct medical costs are varicose veins and diseaseapecific quality of life.
Varicose veins therefore
represent a major source of costs, although they are weakly associated with symptoms and quality-of-life impainnent
After adjustment for several confounding factors, the cost ratio comparing subjects with and without varicose veins was 2.32 (95% CI: 1.25430). Costs were seven times lower in patients in the upper quartile of disease-specific QOL scores as compared to those in the lower quartile of scores. These data adds empirical evidence to the conclusion of a reœntly published article (Bradbury et al.. 1999) suggesting that management strategies for patients with varicose veins should probably be revised.
14. Refeiences Abenhaim L and Kun X for the VEINES Group. The VEINES Study (Venous lnsufficiency Epidemiologic and Economic Study): An international cohort study on chronic venous disorders of the leg. Angeiology 1997;48:59-66. Abenhaim L and The Veines Task Force. The Management of Chronic Venous Disorders of the Leg (CVDL): An Evidence-Based Report of an International Task Force. Phlebology 1999; vol 14 (Suppl.1). in press.
-
Abramson JH. Hopp Cl Epstein LM. The epidemiology of varicose veins A sutvey of Western Jenisalem. J Epiderniol Cornrnunity Heatth 1981;35:213-217. Aitken RJ, Matley PJ. lmmelman EJ. Lower limb trauma: a long term clinical and physiological assessment. Br J Surg 1989;76:585-588. Alexander CJ. Chair-sitting and varicose veins. Lancet 1972;1:822. Almgren B. Eriksson 1. Valvular incompetence in superficial, deep and perforator veins of limbs with varicose veins. Act Chir Scand 1990; 156:69-74. Anderson JH. Geraghty JG. Wilson YT. Murray GD et al. Paroven and graduated compression hosiery for superficial venous insufficiency. Phlebology 1990;5:271-276. Anderson S. Auquier A. Hauck WW. Oakes Dl Vandaele W. Weisberg HI. Statistical methods for comparative studies. Techniques for bias reduction. John Wley & Sons. New-York: 1980. Anonymous. Quality of life in clinical trials (Editorial). Lancet l995;364: 1-2. Armitage P. Berry G. Statistical methods in medical research. 2nd edition. Blackwell Scientific Publications. Oxford : 1987. Amoldi CC. The aetiology of primary varicose veins. Dan Med Bull 1957;4:102-107. Augustin M. Dieterle W. Zschocke Il Brill ClTrefzer D, Peschen M. Schopf E, Vanscheidt W. Developement and validation of a disease-specifc questionnaire on the qualrty of life of patients with chronic venous insufficiency. Vasa l997;26:29l-3Ol. Baccaglini U. Consensus conference on sclerotherapy. Int Angiol 1995; 14:239-240. Baker DM. Tumbull NB. Peanon JCG. Makin GS. How successful is Variwse Vein Surgery ? A Patient Outcome Study Following Varicose Vein Surgery using the SF-36 Heaith Assessrnent Questionnaire. Eur J Vasc Endovasc Surg 1995;9:299-304. Banjo AO. Comparative study of the distribution of venous valves in the lower extremities of black Afncans and caucasians: pathogenic correlates of the prevalenœ of pnmary variwse veins in the two races. Anat Rec 1987;2l7:407-412.
Beaglehole R. Epidemiology of varicose veins. World J Surg 1986;10: 898-902. Beaglehole R. Pnor IAM. Salmond CE. Davidson F. Varicose veins in the South Pacific. Int J Epidemiol 1975,4:295-299. Beaglehole RI Salmond Ce, Pnor IAM. Varicose veins in New Zealand: prevalence and severity. N Z Med J 1W 6 ; W396-399. Balcaro G. Christopoulos D, Vasdekis S. Treatment of supeficial venous incornpetence with the SAVAS technique. A 4-year randomised, cantrolled trial comparing venous hemodynamic and costing after SAVAS. sclerotherapy and the dentist's technique. J Mal Vasc 1991;16:23-27.
Beresford S M , Chant ADB. Jones HO, Pichaud D et al. Varicose veins: a cornparison of surgery and injection1compression sclerotherapy. Five-year follow-up. Lanœt 1978;April29:921-924. Bergqvist 0,Jarozewski H. Deep vein thrombosis in patients with superficial thrombophlebitis of the kg. BMJ 1986; 292:658-9.
BESPE (Belgian Society for Phannaco-Epidemiology). A proposal for methodological guidelines for economic evaluation of pharmaceuticals in Belgium. Phanaco-Economic Cornmittee. Brussels, 1995. Blankfield RP. Finkelhor RS. Alexander JJ. Flocke SA,Maiocco J, Goodwin M. Zyzanski SJ. Etiology and diagnosis of bilateral leg edema in primary cara. Am J Med 1998;15(3):192-7. Bobek K. Cajzl L. Cepelak V. Slaisova VI Opatmy K, Barcal R. Etude de la Fréquence des maladies Phlébologiques et de l'Influence de Quelques Facteurs Etiologiques. Phlébologie 1966; l9(3):218-230. Boccalon H, Janbon C,Saumet JL. Tafani A, Roux T, Vilain C. Characteristics of chronic venous insufficiency in 895 patients followed in general practice. Int Angiol 1997;16:226-34. Bradbury A, Evans C. Allan P. Lee A, Ruckley CV, Fowkes FGR. M a t are the symptoms of varicose veins ? Edinburgh Vein Study cross sectional population survey. BMJ 1999; 318:353-6. Brand ÇN, Dannenberg AL. Abbott RD, Kannel WB. The epidemiology of varicose veins: the Framingham study. Am J Prev Med 1988;4(2):96-lOI. Burkitt DP. Haemorrhoids. vancose veins and deep venous thrombosis: epidemiologic features and suggested causative factors. Can J Surg 1975;18:48-88. Burkitt DP. Vancose veins, deep vein thrombosis, and haemorrhoids: epidemiology and suggested aetiology. BMJ 1972;2:556-561. Burkitt DP. Varicose veins. Facts and fantasy. Arch Surg 1976; 111:1327-1332. Burnand K. Clemenson G.Morland M. Jarret-PEM et al. Venous lipodematosclerosis: treatment by fibrinolytic enhancement and elastic compression. BMJ 198O;January 5:7-11.
a
Callarn MJ. Epidemiology of varicose veins. Br J Surg 1994; 81:167-173. Camilli S, Guamera G. External banding valvuloplasty of the supeficial femoral vein the treatment of primary deep valvular incompetence. InternationalAngiology 1994;13(3):218-222. Campbell B. Thrornbosis, phlebitis and varicose veins. BMJ lQ96;312:198. Canonico S. Gallo C.Paolisso G. Pacifico F. Signoriello G, Sciaudone G, Fenara N, Piegan V. Vamcchio M. Rengo F. Prevalence of Varicose Veins in an ltalian Elderly Population. Angiology l998;2(49):129-35. CBlP (Centre Belge d'Information Pharmacotherapeutique). Répertoire Commenté des Médicaments. Gand: 1996. Charles H. The impact of leg ulcers on patients' quality of Ife. Prof Nurse 1995; lO:Sil-574. Christopoulos O. Nicolaides AN. Belcaro G. Kalodiki E. Venous hypertensive micrangiopaty in relation to clinical severity and effect of elastic compression. J Dermatol Surg Oncol 1991; 17:809813.
-
Cteave TL. Varicose veins nature's error or man's? Lancet 1959;ii:172-175. Cohen G. Forbes J. Garraway M. lnterpreting self reported limiting long tenn illness. BMJ 1995; 31 1:722-4. Cook TA. Michaels JA, Galland RB. Vancose vein clinicsmodelling the effects of seasonal variation in refenals. JR Coll Surg Edinb 1997:42:400-402. Coon WW, Nilis PW, Keller JB. Venous thromboembolism and other venous disease in the Tecumseh community health study. Circulation 1973;48:839-846. Cornu-Thenard A. Maraval M. Boivin P, Parpex P. Prédominanœ gauche des varices:mythe ou réalité ? Phlébologie 1986: 39(2):405-473. Cornu-Thenard A. Dab W. De Vincenzi 1. Valty J. Relationship between blood groups (ABO) and varicose veins of the lower limbs. A case-control study. Phlebology 1989:4:37-40. Cornu-Thénard A. De Vincenzi 1. Evaluation of different systems for dinical quantification of varicose veins. J Dermatol Surg Oncol 1991:l7:345-348. Cornu-Thénard A, Boivin P. Baud JM. De Vincenzi I et al. Importance of the familial factor in varicose disease. Clinical study of 134 families. J Dennatol Surg Oncol 1994;20:31&326. Cornu-Thénard A. De Vincenzi 1. Evaluation of different systems for dinical quantification of varicose veins. J Dermatol Surg Oncol 1991;17:345-348.
Da Silva A, Widmer LK. Martin H, Mal1 TH, Glaus L, Schneider M. Varicase veins and chronic venous insuffîciency - prevaience and risk factors in 4376 subjects in the Basie Study II. VASA 1974:3:118-125.
Daynes G. Breighton P. Prevalence of varicore veins in Afncans. BMJ 1973;3:354.
De Backer G. Epidemiology of Chronic Venous Insufficiency. Angiology 1997; 48(7):569-576. De Jongste AB. Jonker JJC, Huisman MV, Ten Cate JW et al. A double blind three center clinical trial on the short-term eficacy of O-(b-hydroxyethy1)-nitosides in patients with post-thrombotic syndrome. Thromb Haemost 1989:62:826-829. Delecluse M. Ducros JJ. Egal G. Hamel H. Essai clinique pragmatique de Dioveno? 300 mg versus mélange de flavonoides B 90 % de diosmine dans le traitement es manifestations d'insuffisance veineuse chronique chez la femme active jeune. Artéres et Veines 1991;X(7):498503. Dodd H. Cockett FB. The pathology and surgery of the veins of the lower limb. Edinburgh: Livingstone. 1956:3. Drummond MF, Stoddart CL. Torrance GW. Methods for the Economic Evaluation of Health Care Programmes. Oxford University Press. Oxford:1989. Dnimmond M. O'Brien B. Clinical importance. statistical significance and the assessment of economic and quality-of-Iife outcornes. Heatth Economics 1993; 2:2O5-212. Ducirnetiere P. Richard JL, Pequignot G. Wamet JM. Varicose veins: a Risk factor for Atherosclerotic Disease in Middle-aged Men ? Int J Epidemiol 1981;10:329-335. Eisner MD. Katz PP. Yelin EH. Henke J, Smith S. Blanc PD. Assessrnent of asthma treated by family practitioners. allergists and pulrnonologists. Med care 1998; 36(11):1567-77. Ernst E, Saradeth Tl Resch KL. Cornplementary treatment of varicose veins placebo-controlled.double-blind trial. Phlebology 199035:157-163.
- A randomized.
Evans CJ. Fowkes FGR, Leng GL. Stonebridge P. Allan PL. Donnan PT, Ruckley CV. Role of duplex scanning in epidemiology studies of venous disease. J Epidemiol Comm Hlth 1993; 47: S418.
Evans CJ. Fowkes FGR. Hajivassiliou CA. Harper DR. Ruckley CV. Epidemiology of varicose veins. A review. Int Angiol 1994;l3:263-ZiO. Evans CJ, Leng GC.Stonebridge P. Lee AJ, Allan PL. Fowkes FOR. Reproducibility of Duplex ultrasound in the measurement of venous reflux. Phlebology 1995; 1O:l49-154.
Evans CJ. Allan PL. Lee AJ Bradbury AW. Ruckley CV Fowkes FGR. Prevalence of venous refiux in the general population on duplex scanning: The Edinburgh Vein Study. J Vasc Surg 1998;28:767-76.
0
Fanfera FJ. Palmer LH. Mawr B. Pregnancy and varicose veins. Arch Surg 1968;96:33-35.
Çitzpatrick R. Fletcher A. Gore S. Jones D. Spiegelhalter O and Cox D. Quality of life measures in health c m . I: Applications and uses in assessment. BMJ ,992: 305: 1074-77. Franks PJ, Wright DDI. Flecher AE. Moffat CJ, Stirling J et al. Prevalence of venous disease: a community study in West London. Eur J Surg l992a; 158:143-7. Franks PJ. Moffat CJ. Wright DDI. Flecher AE. Moffat CJ, Stirling J et al. A questionnaire to assess risk factors. quality of life. and use of health resources in patients with venous disease. Eur J Surg 1992b;158: 149-55. Franks PJ. Moffat CJ. Conolly M. Bosanquet N et al. Community leg ulcer clinics: effect on quality of life. Phlebology 1994;9:83-86. Franks PJ, Bosanquet N. Brown D. Harper DR and Ruckley CV. Perœived health in a randomised trial of single and multilayer bandaging for chronic venous ulceration. Phlebology 1995; Suppl. 1:17-19. Fraser IA. Perry EP. Hatton M. Watkin DFL. Prolonged bandaging is not required following sckrotherapy of varicose veins. Br J Surg 1985;72:488490. Garratt AM, MacDonald LM. Ruta DA, Russell IT et al. Towards measurement of outcome for patients with varicose veins. Qual Ass Health Care 1993a;2:5-10. Ganatt AM, Ruta DA. Abdalla MI. Buckingham JK et al. The SF 36 health survey questionnaire: an outcome measure suitable for routine use within the NHS? BMJ 1993b;306:1440-7444. Geelhoed GW. Burkitt DP. Varicose veins: a reapparaisal from a global perspective. Southem Medical Journal 1991: 84 (9): 1131-1134. Giblet ER. Blood groups and blood transfusion. In: Wilson et al., ed. Hanisson's Principles of Intemal Medicine. 12th ed. McGraw-Hill, New-York: 1991, Vol. 2, 1494-1500. Ginsberg JS. Brill-Edwards P. Burrows RF, Bona R, Prandoni P. Büller HR, Lensing A. Venous Thrombosis during Pregnancy: Leg and Trimester of Presentation. Thrombosis and Haemostasis 1992;67(5):5 19-520. Gooley NA. Sumner DS. Relationship of venous reflux to the site of venous valvular incornpetence: implications for venous reconstnidive surgery. J Vasc Surg l988;7(1):50-59. Greenland S. Response and follow-up bias in cohort studies. Am J Epidemiol 1977;106: 184-187. Guberan W. Widmer LK. Glaus L. Muller RI Rougemont A, Da Silva A, Gendre F. Causative factors of varicose veins: myths and facts. VASA 1973;2:115-120. Gundersen J. Hauge M. Hereditary factors in venous insuffïciency. Angio 1969;20:346-355. Gutman H, Zelikovski A. Haddad M. Reiss R. Clinical experienœ treating varicose veins in the aged. Am Surgeon 1989:55625628.
Hamer Cl Culum NA. Roe EH. Patients' perceptions of chronic leg ulcen. Journal of Wound Care 1994;3(2):358-36O. Hansen JM. Bytzer P. Schaffalitzky De Muckadeli OB. Management of dyspeptic patients in prirnary care. Value of the unaided clinical diagnosis and of dyspepsia subgrouping. Scand J Gastroenterol 1998;33(8):799-805. Hirai M. Kenichi N. Nakayama R. Prevalence and risk factors of varicose veins in Japanese women. Angiology 1990;41:228-232. Hobbs JT. Varicose veins. Br. Med. J. 1991; 303: 707-710. Hobson J. Venous insufficiency at work. Angiology 1997; 48(7):577-82. Hopkins NFG. Wolfe JHN. Deep venous insufficiency and occtusion. Br. Med. J. 1992; 304: 107-1I O . Hunt SM. McKenna SP. McEwen J. Backett EM. Williams J. Papp E. A quantitative approach to perceived health status: a validation study. J Epiderniol Community Health 1980;34:281-286. Hyland ME. Thomson B. Quality of life of leg ulcer patients: Questionnaire and preliminary findings. J Wound Care 19%;3:294-298. lafrati MD, O'Donneii TF. Kunkemueller A, Belkin MC et al. Clinical examination. duplex ultrasound and plethysmography for varicose veins. Phlebology 1994;9: 114-118. IBES (Institut Belge de I'Economie de la Santé). Aspects socio-économiques des soins de santé en Belgique. Bruxelles: février 1998.
IBES (Institut Belge de I'Economie de la Santé). Compendium de Statistiques de la Santé 1999. Bruxelles: 1999. INAMI (Institut National d'Assurance MaladitInvalidité). Nomenclature des prestations de santé et tarifs de remboursement. Ministère des Affaires Sociales. Bruxelles: 1998. Jakobsen BH. The value of different forrns of treatment for varicose veins. lWS66:182-184.
Br J Surg
James KV. Lohr JM. Deshmukh RM. Cranley JJ. Venous thrombotic complications of pregnancy. Cardiovascular Surgery 1996: 4(6):777-782. Johnston J. Econometric methods. 3d edition. McGraw-Hill International Book Company, Auckland: 1984. Jorgensen JO. Hanei KC, Morgan AM. Hunt JM. The incidence of deep vain thrombosis in patients with superficial thrombophlebitis of the lower limbs. J vasc Surg 1993; 18:70-3. Kakande 1. Varicose veins in Afncans as seen at Kenyatta National Hospital, Nairobi. East Afr Med J 1981;58:66%676.
@
Kleinbaum DG, Kupper LL. Morgenstern H Epidemiologic Research. Principles and quantitatives rnethods. Van Nostrand Reinhold, New-York: 1982. Klinkman MS, Coyne JC. Gallo S. Schwenk TL. False positives, false negatives. and the validity of major depression in primary care. Arch fam Med 1998;7(5):451-61. Labropoulos N, Touloupakis E, Giannoukas AD, Leon M, Katsamouris A, Nicolaides AN. Recurrent varicose veins: investigation of the pattern and extent of reflux with color fiow duplex imaging. Surgery 1996;119:406-9. Labropoulos N. Giannoukas AD. Delis K. Mansour A. Kang S. Where does venous reflux start ? J Vasc Surg 1997;26:73642. Lafurna A. Fagnani F. Peltier-Pujol F. Rauss A. La maladie veineuse en France: un problème de santé publique méconnu. J Mal Vasc 1994;19:185-189. Laing W. Chronic venous diseases of the leg. Studies of Current Health Problem, no 108. Office of Health Economics. London: December 1992. Lambelet F. Le traitement des troubles circulatoires veineux: une étude en double-insu avec le Sandovène". Praxis 1973;62:925-929. Lamping DL. Measuring Health-Related Quality of Life in Venous Disease:Practical and Scientific Considerations. Angiology 1997:48(1):51-57. Larnping DL and Schroter S. for the VEINES Group. Measuring QuaMy of Life and Symptoms in Chronic venous Disorders of the Leg: Developrnent and Psychometric Evaluation of the VEINESQOLNEINES-SYM Questionnaire. Study report. Health Services Research Unit. Department of Public Health and Policy. London School of Hygiene and Tropical Medicine, London. UK: Decemùer 1998. Languillat N. Vecchiali JF. Zuccarelli F. Bouxin A. Essai en double aveugle contre placebo des effets du VasobralB dans les troubles de la perméabilité capillaire liés a l'insuffisance veineuse par le test isotopique de Landis. Angiologie l986;39:1-4. Large J. Surgicai treatment of saphenous varices. with presewation of the main great saphenous tnink. J Vasc Surg 1985;2(6):886-891. Last JM. A Dictionary of Epidemiology. 3d ed. Oxford Univenit Press, New-York:1995. Latto C, Wilkinson RW, Gilmore OJA. 1973;l: 1089-1090.
Diverticular disease and varicose veins.
Lanœt
Launois R. Reboul-Marty J. Henry B. Construction et validation d'un indicateur spécifique de qualité de vie: Le cas de l'insuffisance veineuse chronique des membres inférieurs. Journal d8EconomieMédicale 1994;12:109-126.
Launois R, Reboul-Marty J. Henry B. Construction and validation of a quaMy of life questionnaire in Chronic Lower Limb Venous lnsufftciency (CIVIQ) Quality of Life Research 1996 ;5 539-554Laurent R, Gilly R. Frileux C. Clinical evaluation of a venotropic drug in man. Example of Daflon 500 mg. Int Angiol 1988;7(suppl. to no2):39-43. Laurikka J. Laara E. Sisto T. Tarkka M. Auvinen O, Hakama M. Misclassification in questionnaire survey of varicose veins. J Clin Epidemiol l9%;48 (9): 1175-1178. Laurikka J, Sisto T. Auvinen O. Tarkka M. Laara El Hakama M. Varicose veins in a Finnish population aged 40-60. J Epidemiol Community Health 1993; 47:355-357. Lees TA. Lamber D. Prevalence of lower limb ulceration in an urban health district. Br J Surg l992;7Q:1032-34. Leipnitz G. Kiesewetter H. Jung F. Witt RI Wenzel E. Prëvalenz von venerkrankungen in der Bevolkenrng. Phlebologie und proktologie 1989:1 8:2656. Lindhagen A. Bergqvist Dl Hallbook 1.Berndtsson 1. Venous function after fracture of the lower extremity. A 9-year folow-up of 150 cases. Acta Orthop Scand 1985; 56: 110-1 14. Lindholm Cl Bjellerup M. Christensen OB. Zederfeldt B. Quality of life in chronic leg ulcer patients. Acta Derm Venereol f993:73:440443. Lutter KS. Kerr TM. Roedersheimer LR. Lohr JM. Sarnpson MG. Cranley JJ. thrombophlebitis diagnosed by duplex scanning. Surgery 1991: 11O:42-6.
Superficial
Madar G , Widmer LK. Zemp E. Maggs M. Varicose veins and chronic venous insufficiency disorder or disease ? Vasa 1986:15 (2): 126-134. Maffei FHA. Magaidi C. Pinho SZ. Lastoria S et al. Vancose veins and chronic venous insufficiency in Brazil: prevalence among 1755 inhabitants of a country town. Int J Epidemiol l986;lS:2lO-2l7. Malhotra SL. An epidemiological study of varicose veins in lndian railroad worken from the south and north of India, with special reference to the causation and prevention of varicase veins. Int J Epidemiol 1972;1:1 17-183. Markel A. Manzo RA. Bergelin RO. Strandness E. Valvular reflux after deep vein thrombosis: incidence and time of occurrence. J Vasc Surg 1992;15(2):377-384. McAdam WAF, Horrocks JC. DeDombal FT. Assessrnent of the resub of surgery for varicose veins. Br J Surg l976:63:l37-l4O. Mekky S. Schilling RSF. Walford J. Varicose veins in women cotton workers. An epidemiological study in England and Egypt. BMJ l969;2:591-595. Miettinen OS. Theoretical Epidemiology. Principles of Occurrence Research in Medicine. John Wiley 8 Sons, New-York:1985.
Milne AA. Çtonebridge PA. Bradbury AW Ruckley CV. Venous function and clinical outcarne following deep vein thrombosis. Br J Surg l994;81(6):847-9. Moses LE. The series of consecutive cases as a device for assessing outcornes of interventions. In: Bailar JC. Mosteller F (ed). Medical Uses of Statistics. 2nd edition, NEJM Books, Boston : 1992, 125-140 Munn SR. Morton JB. MacBeth WA. Mcleish AR. To strip or not to strip the long saphenous vein? A varicose veins trial. Br J Surg 1981;68:426-428. Muschietti B. Comparaison clinique en double insu entre la diosmine naturelle et synthétique et le 1449-1452. tribénoside. Praxis 1978;67: Neglén P, Jonsson B. Einersson E. Eklof B. Socio-econornic benefits of ambulatory surgery and compression sclerotherapy for varicose veins. Phlebology 1986;1:225-230. Neglén P. Raju S. A rational approach to detection of significant reflux with duplex Doppler scanning and air plethysmography. J Vasc Surg 1993:17(3):590-595. Neumann HAM. Van Den Broek MJ. Evaluation of O-(b-hydroxyethyl)~sidesin chronic venous insufficiency by means of non-invasive techniques. Phlebology 1990;Suppl l(5): 13-20. Nicolaides AN et al. The Investigation of Chronic Venous Insufficiency. A Consensus Statement. Sixth and Final Post-Conference Draft. The Imperia1 College of Science, Technology and Medicine. London: January 1999. Submited to Circulation. NOM,G. Der varikose Symptomenkomplexes. Urban & Schwarzenberg. Wien: 1918. Nocker W. Diebschlag W. Lehmacher W. Clinical trials of the dose-related effets of O-(bhydroxyethy1)-rutosides in patients with chronic venous insufficiency. Phlebology 1990;Suppl 1(5):23-26. Novo SIAvellone G.Pinto A. Davi G.DiEredita FI Riolo F, Mandalà V. Abrignani MG. Strano A. Prevalence of primitive varicose veins of the lower limbs in a randomized population sample of 176-181. western Sicily . Int Angiol l988:7(2): Perrin M. Gobin JP. Grossetête C. Henry F et al. Valeur de l'association chirurgie itérativesclerotherapie après l'échec du traitement chirurgical des varices. J Mal Vasc 1993;18(4):314319. Phillips C. Burkitt OP. Varicose veins in developing countries. BMJ 1976;l: 1148. Phillips A, Davidson M. Greaves MW. Venous leg ulceration: evaluation of zinc treatment. serum zinc and rate of healing. Clin Exp Oenatol l977;2:395-399. Piachaud O. Weddell JM.
The economics of treating vaficose veins.
Int J Epidemiol
Pocter JM. Moneta GL and an International Consensus Cornmitte on Chronic Venous Disease. Reporthg standards in venous disease: An update J Vasc Surg 1995;21(4):63545. Porter JP. Rutherford RB. Clagett GP. Raju S et al. Reporting standards in venous disease. J Vasc Surg 1988:8(2):172-179. Prerovsky 1. Rottocil K. Hlavova A. Koleilat Z et al. The affect of hydroxyethylnitosides after acute and chronic oral administration in patients with venous diseases. A double-blind study. Angiologica 1972:9:408-414.
Pnce A. Makin GS The cost of day-case surgery for varicose veins. Phlebology 1991;6:223-225. Recoules-Arché J. Importance du sedentarisrne debout dans l'évolution et les complications des varices. Etude statistique. Angeiologie 1965;17:17-20. Renton S. Leon M. Belcaro G. Nicolaides AN. The effect of hydroxyethylnitosides on capillary filtration in moderate venous hypertension. Int Angiol 1994:13(3):259-262. Richardson JB, Dixon M. Varicose veins in tropical Africa. Lancet 1977;1:791-792. Rivlin S. Varicose veins in tropical Afrka. Lancet l974;l:1054. Robbins M. Frankel S. Nanchahal K. Coast J. Williams M. DHA Project: Research Programme. Epidemiologically Based Needs Assessment. Report 9: Varicose vein Treatments. Provisio~al Version. Commissioned by the NHS Management Executive, September 1992. Roeckaerts F. Vanden Bussche G. Double-blind placebo-controlled studies with fiunarizine in venous insufficiency. Angiology 1987:833-845.
Rosfors S. Bygdeman S. Nordstrom E. Assessment of deep venous incornpetence: a prospective study comparing duplex scanning with descending phlebography. Angiology 1990:June(41):463468. Rothman KJ. Greenland S. Modem Epidemiology. 2nd ed. Philadelphia: 1998.
Lippincott-Raven Publishen.
Rougemont A. Varicose veins in the tropics. BMJ 1973;2:547. Royle JP. Recurrent varicose veins. World J Surg 1996; 10:944-956. Rudofsky G. Chirurgie des Krampfadern. 20. Epidemiologie und Pathophysiologie des primare Varikosis. Langenbecks Arch Chir Suppl 11 1988; 139-144. Rutgers PA, Kistlaar PI. Emers EG. Photoplethysmography in the diagnosis of superficial venous valvular incornpetence. Br J Surg 1993; 80% 1-353.
a
Saarinen J, Sisto T, Laurikka J. Salenius JP, Tarkka M. Late sequelae of acute deep venous thrombosis: evaluation five and ten yean after. Phlebology 1995;10:106-109.
Sackett DL, Haynes RB, Tugwell P. Clinical epidemiology. A basic science for clinical practice. Little. Brown and Company. Boston:1985. Sadick NS. Predisposing faton of varicose veins and telangiedatic leg veins. J Dermatol Surg Oncol 1992:18:883-886. Sarin S. Scun JH, Coleridge-Smith PD. Can venous reflux be controlled by extemal compression? J Dermatol surg Oncol l992a;18:144. Sann S. Scun JH, CoMdge-Smith PD. Assessment of stripping the long saphenous vein in the treatment of primary varicose veins. Br J Surg f992b:79:889-93. Schellevis FG. van de Lisdonk E, van der Veiden J, van Eijk JT, van Weel C. Validity of diagnoses of chronic diseases in general practice. The application of diagnostic criteria. J Clin Epidemiol 1993; 46(5):461-8. Schultz-Ehrenburg Ut Weindorf N, Von Uslar d, Hirche H. Prospective epidemiological investigations on early and preclinical stages of varicosis. In: Davy A, Stemmer R. Eds. Phlébologie '89. John Libbey Ltd 1989:163-165. Scott TE, LaMorte WW, Gorin DR, Menzoian JO. Risk factors for venous insufficiency. A dual case-control study. J Vasc Surg 1995;22:622-8. Scurr JH. Coleridge-Smith P. Cutting P. Varicose veins: optimum compression following sclerotherapy. Ann R Coll Surg Engl l985;67: 109-111. Seidell JC, B a b KC, Deurenberg P. Van Den Hoogen HJM. Hautvast JGA. Stijnen T. Overweight and chronic illness -a retrospedive cohort study, with a follow-up of 6-17 yean. in men and wornen of initially 20-50 yean of age. J Chron Dis 1986;39(8):585-593. Sethia KK. Darke SG. Long saphenous incornpetence as a cause of venous ulceration. Br J Surg 1984171:754-755. Sisto T, Reunanen A, Laurikka J, Impivaara 0. Heliovaara, Knekt P,Aroma A. Prevalence and Risk Factors of Varicose Veins in Lower Extremities:Mini-Finland Health Survey. Eur J Surg 1995;161:4OWl4. Skillman JJ, Kent KC, Porter DH. Kim 0. Simultaneous occurrence of superficial and deep thrombophlebitis in the lower extremity. J vasc Surg 1990; 1l:818-24. Skudder PA. Farrington DT, Weld E, Putman C. Venous dysfunction of late pregnancy penists after delivery. J Cardiovasc Surg 1990; 31:748-752. Smith JM. Dore CJ. Charlett A, Lewis JD. A randomized trial of biofiim dressing for venous leg ulcers. Phlebology l992;7: 108-113. Stanhope JM. VaBcose veins in a population of New Guinea. Int J Epid 1975;4:221-225.
Stewart AM. Webb JW, Hewitt O. Social medicine studies based on Civilian medical Board records. II. Physical and occupational characteristics of men with varicose veins conditions. Brit J Prev Soc med 1955: 9:26-32. Strandness DE, Langlois Y, Cramer M. Randlett A, Thiele BL. Long-term Sequelae of Acute Venous Thrombosis. JAMA 1983; 250 (10): 1289-1292. Struckmann JI Mathiesen F. A noninvasive plethysmographic method for evaluation of the musculovenous pump in the lower extremities. Acta Chir Scand 1985;151:235-240. Struckmann J. Low compression. high gradient stockings in patients wih venous insufficiency: effed on the musculo-venous pump. evaluated by strain gauge plethysmography. Phlebology 1986a;1:189-196. Struckmann J. Compression stockings and their effed on the venous pump-a comparative study. Phlebology l986b; 1:37-45. Struckmann J, Christensen SJ, Lendorf A, Mathiesen F. Venous muscle pump improvement by low compression elastic stockings. Phlebology 1986~; 1:gï'-lO3. Stvrtinova V, Kolesar J, Wimmer G. Prevalence of varicose veins of the lower limbs in the women working at a department store. Int Angiol 1991;1O(1)%S. Thibault P. Bray A. Wlodar Czyk. Lewis W. Cosmetic leg veins: evaluation using duplex venous imaging. J dennatol Surg Oncol 1990;16:612-618. Toglia MR, Weg JG. 1996;355(2):108-114.
Venous thromboembolism during pregnancy.
N Engl J Med
Van Den Oever RI Hepp B. Debbaut B. Simon 1. Socio-economic impact of chronic venous insufficiency. An underestimated health problem. Int Angeiol 1998;17:161-7. Vasdekis S. Clarcke H, Nicolaides A. Quantification of venous reflx by means of duplex scanning. J Vasc Surg 1989; 1O:6f0-7. Vin F, Chabanel A, Taccoen A. Ducros J et al. Action de la troxénitine sur les paramètres cliniques, pléthysmographiques et rheologiques de l'insuffisance veineuse des membres inférieurs. Etude contrôlée contre placebo. Artères et Veines 1992;Xl volume spécial:333-342. Vin F. Chabanel A. Taccoen A, Ducros J et al. Double-Mind trial of the efficacy of troxenitin in chronic venous insufficiency. Phlebology 1994;9:71-76. Wacholder S. McLaughlin JK, Silvennan DT and Mandel JS. Selection of controls in casecontrol studies. 1. Principles. Am J Epidemiol 1992a; 135:1019-28. Wacholder S. Silverman DT, McLaughlin JK and Mandel JS. Selection of controls in casecontrol studies. II. Types of controls. Am J Epidemiol l992b; 135:1029-41. Walker AM. Observation and inference Epiderniolopy ~
M U ~ CInc. ~ S Boston:
1991.
Ware JE Jr. SF-36 Health Survey: Manual and lnterpretation Guide. The Health lnrtitute. New England Medical Center, Boston: 1993. Ware JE Jr., Kosinski MA, Keller SD. SF-36 Physicai and Mental Component Summary Measures: A User's Manual. The Health Institute, New England Medical Center, Boston: 1994. Ware JE, Keller SD, Gandek B, Brazier JE, Sullivan M and the IQOLA Project Group. Evaluating translations of health status questionnaires. Methods from the IQOLA Project. International Journal of Technology Assessrnent in Healh Care 1995;11 (3):525-551. Weddell JM. Varicose veins pilot study, 1966. Br J Prev Soc Med 1969;23:l79-186. Weinmann E. Sabman €W. DeepVein Thrombosis. N Engl J Med 1994; 331(24):1630-1641. WHO Collaborating Centre for Drug Statistics Methodology. Anatomical Therapeutic Chemical (ATC) classification index, including Defined Daily Doses (DDDs) for plain substances. Oslo. January 1996.
-
Widrner LK, ed. Peripheral venous disorders prevalence and socio-medical importance. Bern; Hans Gruber, 1978:1-90. Wdmer L. Biland L, Barras JP. DoxiumQ 500 in chronic venous insuficiency: a double-blind placebo. Controlled multicentre study. Int Angiol 1990;9:1OW 10. Widmer LK, Holz D, Morselli B, Zbinden O et al. Progression of varicose veins in 11 years. Observations on 1441 working penons of the Basle Study. Unpublished paper. Basle: Angiology Division of University Dept of Medicine, 1992. Wenert V. Epidémiologie et socio-économie des maladies veineuses en Allemagne. Phlébologie 1993;46:225-234. Willén J, Bergqvist O. Hallbook T. Venous insufficiency as a late complication after tibia1 fracture. Acta Orthop Scand 1982; 53: 149-153. Zuccarelli F. Efficacité clinique et tolérance de la coumarine rutine. Etude contrôlée en double aveugle versus placebo. La Gazette Médicale l987;94(32): 1-7. Zuccarelli F, Benedetti F. Houdot-Roque H, Laffont JG et al. Activité de veinobiaseQ sur les troubles fonctionnels de l'insuffisance veineuse dus a I'orthostatisme professionnel. Angiologie 1988;84(5):3-7. Zuccarelli F, Vincent 6. Enquête épidémiologique sur les 15 premières nouvelles patients consultant dans 200 cabinets de phlébologie pour des douleurs liées a une insuffisance veineuse des membres inférieurs. Artères et Veines, 1990; IX(8): 724-727. Zukowski AJ, Nicolaides AN, Szendro G. Haemodynamic significance of incompetent calf perforatingveins. Br J Surg 1991; 78:625-629.
