ClinicalScience (1981) 61,403s-411s

403s STATE OF THE A R T REVIEW

The problem of treatment in mild hypertension

W. S. P E A R T Medical Unit, St Mary’s Hospital, London

In a Conference devoted to the study of hypertension it may seem strange that it is necessary to define the term. The usual definitions are shown in Table 1 [ 11 and it is a particular pleasure for me at the present time to refer to the first definition proposed by George Pickering, who said that you never define but describe the quantity. Among his many contributions one of the most important was the description of blood pressure as a continuous variable without any sharp cut-off point at which the term hypertension could properly be used. The most commonly used definition is no. 4 from the Table, which I call the working or insurance definition, and this is that which most clinicians will use. It gives a probability to the expectation of morbidity or mortality for the individual within a time such as 5 to 20 years hence. This enables me to stress the difference between clinical epidemiology and clinical practice; the former depends upon large numbers so that it is, for example, possible to give a probability of a particular morbid event occurring based on one single reading of systolic blood pressure [2]. This contrasts sharply with ordinary clinical practice where due weight has to be given to the emotional state of the patient and even of the observer. Only at extremes of blood pressure is it possible to give high risk factors to individuals and even there prognosis may be extremely variable. I want to emphasize this by reference to the condition known variously as labile or borderline hypertension, represented in Table 2 [ 11. It can be

seen that I believe this may be a mistaken classification and that the term should be labile blood pressure, since this is an accurate description of the clinical findings. There have been many studies in which longitudinal measurements of blood pressure in cohorts of young individuals classified as having labile blood pressure indicate that as many as three-quarters in some studies do not achieve levels of pressure requiring treatment many years later [31. As indicated in Table 2, I feel confident that selecting out at one time time groups of subjects with labile blood pressure and then attributing characteristics which are given the title of labile hypertension may be a circular argument and that most of these characteristics are related to a physiological defence reaction [4]. The rise of blood pressure with age in most of our western societies has to be considered a pathological phenomenon, since this is not so for all individuals and certainly not for all societies in the world [5]. Furthermore the higher ranges of blood pressure are associated with the highest number of cardiovascular morbid events. The major difficulty confronting the clinician is shown in Fig. 1, which illustrates the process of ‘tracking’. This implies that the individual will follow a track of either a stable blood pressure with rising age or a steady increase of blood pressure over the same time span. It is known from population and cohort studies that the divergence of blood pressures, leading to different mean blood pressures for different groups within populations, occurs between the ages of 20 and

TABLE1. Definitions of hypertension

TABLE2. Labile hypertension

I. Never define, describe quantity (Pickering). 2. Statistical deviation from young adult blood pressure (say 120/80 mmHg). 3. Deviation from mean of whole population at given age (SDunits). 4. Working or insurance definition: level which is associated with increased morbidity and mortality at some future time (5-20 years) compared with whole population (excess mortality).

I. Misnomer: should be labile blood pressure. Cohort longitudinal studies, 2. Investigated: selection creates group characteristics. 3. Defence reaction: increased heart rate, cardiac output, muscle flow decreased skin and kidney flow anxiety state?

W. S. Peart

404s zoo 1

0

/-

I 10

30 40 Age (years)

20

50

70

60

FIG. 1. Diagrammatic representation of blood pressure 'tracking' which leads to the familiar rise of mean blood pressure for the whole cohort or population. Initial pressure differences are small up to age 20 years.

I:;

; 5

,

,

,

,

,

,

,

,

,

,

,

*, *,

,

.

,

-30 100

140

120

160

180

200

240

220

Mean blood pressure (mmHg)

FIG. 2. Mean change of systolic pressure (ordinate) in 10 years according to mean systolic pressure (abscissa); Rhondda Fach females aged 15-74 years at first survey. (From Miall & Lovell 171). +40

$

A +30.

-

A

A

A

'IA

-101

1 1

A

A

'

2

A

A

A

'

I 1 ' 5 6 7 8 D e c k mean systolic pressure (Age 24-54 years) I

3

I

4

I

9

I

10

FIG. 3. Mean change of systolic pressure (ordinate) by level of mean systolic pressure from mean age 24 to 54 years (abscissa). (From Harlan el al. [81).

30 years [6]. It is very difficult to differentiate at much younger ages. This, of course, is why so much effort is going into identifying those characteristics in young people which mark them out for a rise of blood pressure later in life. From epidemiological cohort studies there is evidence to support the tracking suggestion. The data in Fig. 2, from the study by Miall & Lovell [71 in South Wales, show that, when the blood pressure of an individual is measured again some years later, the second pressure is related more to the initial reading, the initial systolic pressure predicts the final systolic pressure, and Miall & Lovell pointed out that age was not significant. This was also supported by one of the best cohort studies I know, that of Harlan et al. [81, involving the study of men entering the American Naval Air Service during World War 2. Several of their physical characteristics were measured including blood pressure and when they were followed up over the course of 30 years the relationship shown in Fig. 3 emerged. In short, the higher the starting pressure the higher the pressure 30 years later. However, from Fig. 3 it can be seen that it is only at the higher initial pressures (over 160 mmHg systolic pressure) that the rise of pressure later in life becomes very marked. It would clearly be very difficult to pick out individuals at the lower levels and give them any sort of expectation for later life. I therefore want to put the problem of diagnosis and treatment of mild hypertension firmly into a clinical context before seeing whether the results of therapeutic trials, which of course are based on epidemiological criteria, offer any help to the clinician. The clinical progress of the young woman shown in Table 3 was chosen for its particular difficulties and I am quite sure that many physicians would have treated her, and yet at present I am not doing so. It can be seen that she started under observation when her blood pressure rose when receiving the contraceptive pill. This was stopped and when she married in 1976 her blood pressure again was observed to be elevated and the physician at that time started treatment. However, she was not happy with this course and treatment was stopped. In 1977 I first saw her and was impressed by her anxious state and her labile blood pressure. This was marked as so often by a fast pulse rate, and despite the very high figures gained on measurement in the clinic I decided against treatment; but her initial home blood pressure measurements taken by her husband convinced me that her blood pressure was likely to be at reasonable levels, as shown by the figures in January 1978 where at the end of a

Problem of treatment in mild hypertension

405s

TABLE3. Clinical course of a female patient with labile blood pressure, followed through two pregnancies, contrasting the. hospital and home blood pressure readings Patient (26 years)

1973 1976

Blood pressure (mmHg) Contraceptive pill 1 year Married

Pulse (min-')

0 Hygroton Trasicor

BP rose BP rose

1977 1978 1978 1978 1979 1979 1979 1979 1979 1980

St Mary's Hospital January Hospital January-February Home BP, 2 weeks twice daily June-March Pregnancy (Hospital) March April May 0ctober January

Delivery: normal baby Home BP, 2 weeks twice daily Hospital Hospital Hospital

160/90 (start), 140/80 (end) 17011 10 I80A 15 180/120

144 1 I6 80

1980

February

Home BP, 2 weeks twice daily

Range 120/70-1 l0/60

60

1980

June

Hospital

190/115

80

1980

August-February

Pregnancy (Hospital)

1980

OctoberNovember DecemberFebruary February

Home BP, 2 weeks twice daily

1981 1981

150/100-170/110 190/120 170/110 (start), 120/80,(end) 190/120(start), I20/80 (middle), 160/110(end)

190/120(start), 150/110 (middle), l80/l 10 (end) Range 120/70-130/75

Treatment

0 150

0 0 0

0 0

0 Propranolol (40 mg b.d.) Propranolol (40 mg b.d.) Propranolol (40mg b.d.) 0

0 Home BP every 2-3 days

IS5/85 (start), 140/85 (middle), 135/85 (end)

0

Delivery: normal baby (birth weight 3.5 kg)

2 week period of twice-daily blood pressure measurements figures of 120/80 mmHg were obtained. In June 1978 she became pregnant and the obstetrician concerned was doubtful whether the pregnancy should be continued with pressures such as 190/120 mmHg. However, I said that I thought her pressure was not really at that level most of the time, and further attendance at the Ante-Natal Clinic was associated with a fall of pressures to a range which, while higher than the obstetrician liked, did not seem to be harming the development of the foetus and she was subsequently delivered of a healthy baby. There are incidentally many women who do drop their blood pressure during pregnancy; some of them may, like this patient, have a blood pressure which is elevated only by measurement in the clinic. I repeated the series of home blood pressure readings for 2 weeks in April 1979 after the pregnancy and again the figures obtained did not suggest to me that I should start treatment. Her subsequent clinic visits were similar and usually marked by a tachycardia. I then decided to see if slowing her pulse with a P-blocking compound would lower her blood pressure but the figures obtained from January 1980, when she received propranolol (40 mg twice daily), show

that she was capable of slowing her pulse but that her blood pressure remained elevated. This is a not uncommon finding in many patients. However, close to that time, when the home blood pressure readings were repeated, as can be seen from the Table her pulse rate was even slower and her blood pressure was lower and at levels which would have been considered very acceptable by most physicians. Then in August 1980 she became pregnant again and once more the obstetrician was concerned about her very high ante-natal visit pressure. I once more made the same, I hope, reassuring remarks and she continued during her pregnancy with clinic blood pressures at high levels but, in order to reassure both myself and the obstetrician, she carried out home blood pressure measurements from October to November and the range was, of course, as usual much lower. For double assurance she repeated the measurements during December to February about every 3 days, and during the last 2 months of her pregnancy her blood pressure was reasonably stable around a figure of 135/85 mmHg. In February she was delivered of a healthy baby weighing 3.5 kg. I stress the normal birth weight since the placenta is meant to bear the brunt of hypertensive vascular changes and therefore to

406 s

W. S. Pearl

lead to a small birth weight. Clearly this was not so in her case, lending support to the general view that her pressure was not usually at levels which cause damage to her blood vessels, and the question now before me is whether I should start treatment on the basis of her clinic blood pressures which are sure to return to the previous levels. I can therefore turn to see if I can seek help from any of the large-scale trials of trestment in mild hypertension which have been concluded or which are still in progress [9-141.

Initial question There is a variety of Murphy’s law which particularly applies to clinical trials. This states: ‘If it can go wrong it will and if not at the start then in the middle or at the end’; but in the beginning it is the initial question which is asked of a trial which governs everything else and it is vital that this question is clear, unambiguous and capable of an answer by the subsequent design of the trial. In looking at the current trials it is clear that different questions have been asked. For example, in the Australian [ l 11 and MRC [ 141 trials the basic question is whether control of mild hypertension leads to the prevention of development of morbid events in the cardiovascular system, and this is in contrast to the question asked in the American HDFP trial [121, where the question was whether improved treatment in special hospital clinics was better than the ordinary care available outside hospital. This was the difference between a ‘stepped care’ programme and a ‘referred care’ programme. Since the initial questions were different it is no use expecting the eventual answers to be the same, and attempts to press the data to give answers to prove a point would be unwise.

Numbers Once the initial question has been settled then the size of the trial needs consideration and, in the prevention‘ of cardiovascular morbidity and mortality, knowledge of the vital statistics of the population under study is required. This is usually very difficult to obtain and in the MRC trial it was necessary to extrapolate from data obtained on smaller populations within Britain, as well as the Registrar General’s mortality data, and fortunately these estimates have turned out to be about right. In this trial it was estimated that 18 000 participants would be required, which meant that the screening process had to be applied to hundreds of thousands of individuals. In the HDFP trial it was necessary to admit

10 940 subjects and in the Australian trial 3640. It is of interest that the numbers even so are so different and it does suggest that there are differences hidden within these populations which may ultimately make comparison between them difficult, quite apart from the statistical end point chosen for significance. These considerations lead me to the conclusion that where low levels of blood pressure are being studied the trials numbered only in hundreds will never be able to give a conclusive answer. This applies to the small completed USPHS I101 and Oslo [131 trials, the answers to which were inconclusive for these reasons.

Prevention of development of morbid events When this is the main question, as in the Australian and MRC trials, then it is essential to exclude from the trial those subjects who have target organ disease or diseases which may increase the number of morbid events, so that those with stroke, myocardial infarction, angina, renal disease, diabetes and gout, for example, would be excluded. If they are not, then you can be sure that a sub-section of Murphy’s law will apply, which states ‘Where a characteristic can be distributed on a non-random basis between a control and a treated group it will be so distributed’. This is known as ‘the bias by chance’ phenomenon. The best example of this I know occurred in the United States clinical trial of oral hypoglycaemic agents (University Group Diabetes Programme [151) and the ensuing disputes lasted for years [161. This is important in the context of trials of the treatment of hypertension and if bias is introduced at the outset then the outcome will be biased. In the Australian trial even though the entry limits for diastolic pressure were set between 95 and 109 mmHg 16% of the entrants had a final diastolic pressure greater than 110 mmHg, since this was not made an excluding point; clearly more severe cases could dilute the supposed mild entry population. In the HDFP trial, since the question being asked was not the same, the protocol allowed 25% of those in the lower stratum of blood pressure to enter the trial while they were receiving treatment for hypertension. Furthermore target organ disease could be included in the entry characteristics so that it would be expected that even in the lower blood pressure range more severely affected subjects would be included, and especially if the myocardium was involved the entry blood pressure might not reflect past higher blood pressure. This is clearly brought out in Table 4, which shows the

407s

Problem of treatment in mild hypertension TABLE 4. Control group mortality Trial

Sex

Age range (years)

Diastolic BP range

HDFP' (Stratum I) I121 ANBPS I 1 11 Oslo I131 MRC I141

M&F

30-69 30-69 40-49 35-64

90-104 95-109 1107 90-109

M&F M M&F

Mortality All causes

CHD$

Stroke

1.54

0.54 0.10 0.10 0.15

0.19 0.05 0.10 0.03

0.28 0.43 0.39

Referred care group.

t Systolic BP 150-179

mmHg.

$ CHD, Coronary heart disease.

la1 women Age 3 5 - 4 4

Ibl Men Age 4 5 - 5 4

Age 3 5 - 4 4

Age 55-64

Age 45-54

Age 5 5 - 6 4

1 7 0 1 systolic

16-04

.~

8

0 J m 0 1 2 3 4 5

rn m

0 1 2 3 4 5

0 1 2 3 4 5

Years in trial

0 1 2 3 4 5

0 1 2 3 4 5

0 1 2 3 4 5

Years in trial

FIG. 4. Blood pressure change in treated (0, bendrofluazide or propranol) and placebo or observation (0)groups by decades and sex over 5 years, (a) for women and (6) for men. Note marked and maintained fall in placebo group. (Data from MRC Trial 1141.)

morbid events occurring in the control groups for the major trials or in the HDFP trial the morbidity in the referred care group. I should also emphasize the point that the original Veterans Administration trial [9] was drawn from a population with much target organ damage at the outset (more than 50%), and the same applies to the HDFP trial [12] but to a lesser extent. Both the Australian [ l l ] and the MRC 1141 trials are very comparable and have lower morbidity in their control populations.

Placebo The best type of trial will always be a doubleblind trial, followed by single blind and then by, worst of all, an open trial. On a large scale this is the most difficult organizational problem to solve but there is little doubt that it is one of the most important decisions that is ever taken at the outset of any trial. Bias over the selection of subjects can come into a trial with the greatest of ease and it may not be perceived until the trial has 14

been run for some time. Therefore the introduction of a placebo is imperative. This can readily be shown from the results of the MRC trial, as in Fig. 4(a, b), which shows that the blood pressure both in men and women drops considerably during the course of the trial in those receiving the placebo and it is only between 4 and 5 years within the trial that the control group mean pressure starts to rise. We know that nearly 40% of those receiving a placebo reduce their blood pressure to the level aimed at in the treatment group so many are being treated for the benefit of few, and this must be one of the most important conclusions of any large-scale trial. The same is also true of the Australian trial. This has an important bearing on any initial prejudice in the types of patients admitted to treatment or placebo groups, and on looking at the Australian trial I was interested that there were nearly twice as many subjects withdrawn from the placebo group by their own doctors as from the treated group. This was in contrast to the hospital clinic experience and does, of course, suggest an

408s

W.S . Peart

unconscious prejudice. Since the final analysis must depend upon the hardest data then any prejudice in favour of those with the slightest sign of target organ disease will inevitably bias the final answer. The placebo effect in the MRC trial was shown not to depend upon the administration of a tablet since in a sub-set of the control population only repeated measurements of blood pressure were made without tablets, and the same fall of pressure was observed with a slow rise in the mean pressure over the next 5 years, which was not different from the main placebo group. It might be thought that the correct treatment for hypertension was merely to measure the blood pressure. Compliance and quality control

In any large-scale trial these are two of the most important aspects. The answer lies in the enthusiasm of the participants, whether they are doctors, nurses or other personnel, and, of course, most of all the subjects. If there is a big fall-out from the trial and particularly if it affects one group more than another then the whole answer to the trial may be at peril. If the reasons for the withdrawal from treatment differ among different groups within a trial then again the final analysis will be affected. It is interesting in the Australian trial that 30% of the participants left mostly within the first year and they included more smokers and more women. There is a definite but lesser problem in the MRC trial. The importance of quality control so that the data are uniformly good from all centres and that the end points are being achieved, as well as the compliance (measured by pill counts or biochemical tests), requires the same enthusiastic participants. In the MRC trial, although not perfect, the end point was being achieved in a very high percentage of participants (nearly 70%). In the HDFP trial enormous efforts were made to see that compliance and quality control were at the highest level in the ‘stepped care’ group. This involved providing free transport and free drugs as well as the services of a physician at all times. Special personnel were used so that waiting was at a minimum, and it could be said to be the Cadillac compared with the Ford. It reminds me that in 1948 the Socialist Minister of Health in Britain, Aneurin Bevan, in introducing the National Health Service believed that the provision of better health care would reduce the demands on the Health Service due to a healthier population. Unfortunately events showed this view to be a little over optimistic. In congratulating American colleagues in demonstrating for the first time on

TABLE 5 . Treatment regimen MRC Trial Bendrofluazide 5 mg b.d. fixed. Propranolol up to 320 mg daily individually adjusted. Fixed placebo tablets for each main drug. Methyldopa: Supplementary drug to either regimen if diastolic pressure not below 90 mmHg.

such a scale that all causes of mortality can be reduced by careful medical surveillance [ 121, and perhaps especially in poor social conditions, I have to state my belief that even the United States cannot afford this approach and furthermore that this does not answer the question being asked either in the Australian or in the MRC trial, which is whether morbidity due to mild hypertension can be improved by treatment, particularly in countries with different types of health care. Drugs

Where treatment will usually have to be for life it becomes increasingly important, particularly in young people with low levels of blood pressure, to use drugs with low toxicity. This particularly applies to subjects in trials [ l l . 121 who enter without symptoms or signs. Ideally in a trial in mild hypertension only one drug should be used for each group under supervision. In various trials this has not been possible and this lessens the ability to assign any noxious effects to any particular drug. In the MRC trial an attempt has been made to keep the two major drugs, bendrofluazide and propranolol, apart with as little supplementation by other drugs as possible. This has not been an easy task but it is believed that some of the characteristics of these two drugs will be brought out on subsequent analysis and already interesting information has become available. The therapeutic regimen used is shown in Table 5. Most of the other trials have involved mixtures of drugs, the main aim being of course to achieve a particular level of pressure in the treated group. I am sure, however, that we have to become more aware of morbidity due to drugs if we are going to commit large numbers of the population to treatment for 20 years or more. Analysis of results

Perhaps the most important point here is that analysis must be carried out only on the initially randomized groups so that the entry population in the two groups is compared, like with like. Any deviation from this principle leads to statistical

Problem of treatment in mild hypertension

409s

TABLE 6.ANPS Trial I l l 1 Significance: * P

< 0.05; ** P < 0.25; *** P < 0.01.

Rate per 1000 person-years exposure to risk Trial end points

10

On treatment

Intention to treat Active

Placebo

Active

n = 1721

n = 1706

n = 1721

No.

Rate

No.

Fatal Cardiovascular Non-cardiovascular Total Non-fatal

8 17 25 113

1.1 2.4 3.6 16.2

18 17 35 133

All trial end points

138

19.7

14.1

Rate

Placebo ? n = 1706

No.

Rate

No.

2.6.. 2.5 5.1 19.4

4 5 9 82

0.8 0.9 1.7 15.5

13 19 108

2.5” 1.2 3.7’ 20.8”

168

24.5.

91

17.2

127

24.5”’

%

5-

8-

6

Rate

7%

(bl

4-

.O%

7%

h

E x .-

h

6-

3-

s

2-

-m .-

L

g

Gx z

42-

.9%

1-

0-

Stratum DBP (V)

0-

I 90-104

I1 105-1 14

I11 115+

CVS Non-CVS Stratum I ! 90-104

CVS Non-CVS Strata I1 and I11 105+

FIG.5. (a) All-causes mortality (5 .years)showing a greater reduction in stratum I, the group with the lowest pressure. (b) Comparison of cardiovascular (CVS) and non-cardiovascular(Non-CVS) mortality (5 years) in the different strata, with the greatest effect shown in stratum I, cardiovascular mortality. 0,Referred care; El, stepped care. (From [121.)

imprecision and a weakening of the strength of the trial. From the results in the Australian trial shown in Table 6 it can be seen that the numbers under discussion for fatal end points were very small. Since two-thirds of all the end points were due to myocardial causes, mostly non-fatal, any bias in this analysis could easily alter the significance of the results. The Australian results were analysed in two ways. The first was on the basis of ‘intention to treat’ in which the results are analysed on the originally randomized groups no matter what their subsequent fate. The other was on an ‘on treatment’ basis in which only those events which occurred while the subjects were in that category were analysed. This method could lead to bias since the groups might diverge subsequently.

Analysis of myocardial infarction using the two methods of analysis shows that fatal myocardial infarction occurred in five compared with 11 control subjects on the ‘intention to treat’ method and in two compared with eight of those in the ‘on treatment’ method. These results are not significant at the 5% level. Interestingly, of the non-fatal myocardial infarcts 28 occurred in the treated group, compared with 22 in the control on the ‘intention to treat’ basis and 18 as to 17 in the ‘on treatment’ analysis. As an example of the effect of analysing the data in two different ways it can be stated that there were 60 fatal events of which 26 were cardiovascular in the ‘intention to treat’ group but only 28 of which 17 were cardiovascular in the ‘on treatment’ group. There were 246 non-fatal end points in the ‘intention to treat’

W. S. Peart

4 10s

should have been allowed to continue to a point where all doubt as to significance was removed, particularly over the form of most rigid analysis of the data. It is significant in this respect that the Australian trial was stopped after 14000 patient-years and that the M R C trial is still continuing with over 30 000 patient-years recorded. The HDFP trial cannot answer the question about the treatment of mild hypertension, except in a very restricted sense, since that was not the original question, quite apart from the points which have already been made about the form of the trial. If the subject is under 50 years, especially female, if the diastolic pressure is less than 100 mmHg and if there is no cardiac or vascular sign of damage at entry then none of the trials, starting with the Veterans Administration trial 191, provides any clear evidence that treatment is of benefit.

analysis and 190 in the 'on treatment' analysis. The 'on treatment' analysis therefore excluded 53% of the fatal events but only 23% of the non-fatal end points thus biasing the analysis towards softer data. The analysis of the HDFP data is brought out in Fig. 5(a, b), which shows that the 'all causes' mortality is lowered more in the group with the lowest blood pressure (stratum I) when the 'stepped care' group is compared with the 'referred care' group. Turning to the cardiovascular mortality and morbidity, it is in the lowest blood pressure range that the maximum benefit appears to have occurred. This is in contrast to the previous findings of the Veterans Administration trial [91. It is in this group that 25% of the entrants were already on antihypertensive treatment. It is nevertheless surprising that those with the lowest blood pressure received the greatest benefit from treatment when stepped and referred care are compared.

Noxious drug effects Conclusions from results of the trials

As referred to above, this is a most important question and Table 7 shows the rate of withdrawal from treatment in the M R C trial, relating to the two major drugs and the placebos. The breakdown of the reasons for withdrawal as well as for the adverse reactions noted in general direct attention in both sexes to the high incidence of diabetes on the thiazide and especially to impotence and gout in the male. Raynaud's

Do these results give a clear indication to the clinician for treatment of someone he has decided has a level of pressure which, though not very high, nevertheless may cause trouble in the future? My conclusion has to be 'no'. The major completed trial, the Australian one, was concluded with at least my feeling (and there is nothing easier to have than clear hindsight) that it

TABLE I . MRC Trial 1141:principal reasonsf o r withdrawal of randomized treatment A small number of patients appear more than once in this Table. Significance: * P < 0.05; ** P < 0.01; *** P < 0.001. Numbers of patients are shown in parentheses. Placebo groups are larger than active drug groups because they include control patients for both the active drugs, and because patients were not included if they had been withdrawn from randomized treatment or were taking supplementary drugs. ~~

Rates per I000 patient-years (number of reports) Adverse reactions

Men Bendrofluazide (2452 patientyears)

Impaired glucose tolerance Gout (symptoms and serum uric acid > 50 I rnmol/l) Impotence Raynaud's phenomenon Skin disorder Dyspnoea Constipation Lethargy Nausea Dizziness Headache

Women

Propranolol

Placebos

(2738 patient- (6776 patient-

years)

years)

Bendrofluazide (2827 patientyears) 6.01*** I .06

Propranolol years)

(17)1.12 (3) -

9.38"' 12.23***

(23) 3.65 (30) 2.56

(10) (7)

2.51

19.58***

(15) (15)

0.89

_

_

_

0.15

0.35 0.35

(I)

5.62". 1.87' 6.37.'.

1.63.6.93'''

(48) 5.48*** _ 5.48." ( I ) 2.19** - 8.04*** (4) 0.37 (17) 10.23***

8.56"'

(21)

_

0.41

-

1.03

(6) 0.15 (22) (I) (28)

0.15

-

-

2.12.. 2.12'

6.57*** (18)

1.18

0.30

16.27***

(I)

-

Placebos

(2667 patient- (6 122 patient-

years) (3) -

0.82

-

(5)

_

(15) (5)

_

_

_ _ 0.33 (2) _ _

(17) (6) (6) 15.00*** (40) 0.33

(2)

2.12

(13)

(46) 13~87"'

(37)

Problem of treatment in mild hypertension phenomenon and dyspnoea were common to both sexes on propranolol and, of course, impotence in the male. What is not known are other effects which may be of great importance in the long-term results of the use of these drugs.

Conclusion It is not a great consolation to be told that the biggest benefits in hypertension at any stage of treatment are to be found in those subjects aged over 50 years with diastolic pressures over 100 mmHg, who already have signs of target organ disease, in the heart, brain or kidney; so that the clinician seeking guidance from the current trials as to what should be done in the case of a young symptomless subject must still be regrettably in a dilemma, particularly if the short-term side-effects are considered, quite apart from any long-term ones. That difficult to measure ‘quality of life’ has to come into the reckoning. However, I have words of comfort since I think that time is always on the side of the physician and that, in the individual case, to follow up the blood pressure level is usually all that is necessary, aided where appropriate by home blood pressure readings [171, especially if you believe with me that most morbidity is associated with a rise of blood pressure over time. I believe that the actual morbidity at the lower pressures is so slight that it is quite unjustified to place large numbers of symptomless people on long-term treatment without a long period of observation. I have not had the experience that this causes undue anxiety if it is handled in the proper way, and that is part of the physician’s role, and therapeutically inclined epidemiologists should be resisted. Acknowledgments I am particularly indebted to Bill Miall for analyses of data as well as to my other colleagues on the MRC Working Party for their collective wisdom over the years. However, I am personally responsible .far all expressions of opinion, which are my considered view. References 111 PEART,W.S. (1980) Concepts in hypertension. The Croonian Lecture 1979. Journal of the Royal College of Physicians of London, 14,141-152.

411s

121 KANNELL,W.B. (1979) The epidemiologic significance of

familial and genetic factors in hypertension. In: Hypertension-Determinants, Complications and Intervention, pp. 49-57. Ed. Onesti, G. & Klimt, C.R. Grune and Stratton, New York. 131 JULIUS,S. (1977) Borderline hypertension: epidemiologic and clinical implications. In: Hypertension, p. 630. Ed. Genest, J., Koiw, E. & Kuchel, 0. McGraw-Hill, New York. 141 BROD,J., FENCL, V., HEJL, Z. & JIRKA,J. (1959) Circulatory changes underlying blood pressure elevation during acute emotional stress (mental arithmetic) in normotensive and hypertensive subjects. Clinical Science, 18,269-279. 151 EPSTEIN,F.H. & ECKHOFF,R.D. (1967) In: The Epidemiology of Hypertension: Proceedings of an International Symposium, p. 155. Ed. Stamler, J., Stamler, R. & Pullman, T.N. Grune and Stratton, New York. 161 LONDE,S., GOLDRING, D., GOLLUB,S.W. & HERNANDEZ, A. (1977) In: The Epidemiology of Hypertension: Proceedings of an International Symposium, p. 13. Ed. Stamler, J., Stamler, R. & Pullman, T.N. Grune and Stratton, New York. 171 MIALL, W.E. & LOVELL, H.G. (1967) Relation between change of blood pressure and age. British Medical Journal, ii, 660. 181 HARLAN, W.R., OBERMAN, A., MITCHELL, R.E. & GRAYBIEL, A. (1973) A 30-year study of blood pressure in a white male cohort. In: Hypertension: Mechanisms and Management, p. 85. Ed. Onesti, G., Kim, K.E. & Moyer, J.H. Grune and Stratton, New York. 191 VETERANS ADMINISTRATION CO-OPERATIVE STUDYGROUP (1970) Effects of treatment on morbidity in hypertension. Results in patients with diastolic blood pressures averaging 90 through 114 mmHg. Journal of the American Medical Associotion, 213, 1143-1 152. 1101 MCFATESMITH,W., EDLAVITCH, S.A. & KRUSHAT,W.M. (1979) U.S. Public Health Service Hospitals Intervention Trial in Mild Hypertension. In: Hypertension-Determinants, Complications and Intervention, pp. 38 1-399. Ed. Onesti, G. & Klimt, C.R. Grune and Stratton, New York. I 1 1 I AUSTRALIANBLOOD PRESSURE STUDY MANAGEMENT (1980) The Australian Therapeutic Trial in Mild COMMITTEE Hypertension. Lancet, i, 1261-1267. I121 HYPERTENSION DETECTIONAND FOLLOW-UP PROGRAM COOPERATIVE GROUP(1979) Five-year findings of the Hypertension Detection and Follow-up Program. (1) Reduction in mortality of persons with high blood pressure, including mild hypertension. (2) Mortality by race, sex and age. Journal of the American Medical Association. 242.2562-2577. A. (1980) Treatment of mild hypertension: a five I131 HELGELAND, year controlled drug trial. The Oslo Study. American Journal of Medicine, 69,725-732. I141 MRC WORKING PARTYON MILD TO MODERATEHYPERTENSION (1977) Randomised controlled trial of treatment for mild hypertension: design and pilot trial. British Medical Journal, i, 1437-1440. GROUPDIABETESPROGRAM (1970) A study of 1151 UNIVERSITY the effects of hypoglycemic agents on vascular complications in patients with adult-onset diabetes. Diabetes, 19 (Suppl. 2), 747. I161 SHIAO-WEISHEN & BRESSLER,R. (1977) Clinical pharmacology of oral anti-diabetic agents. New England Journal of Medicine, 296,787-793. 1171 JULIUS, S., ELLIS, C.N., PASCUAL A.V., MATICE. M., HANSSON, L., HUNYOR, S.N. & SANDLER, L.N. (1974) Home blood pressure determination. Value in borderline (‘labile’) hypertension. Journal of the American Medical Association, 229,663-666.