Ó The Author 2008. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: [email protected].

doi:10.1093/fampra/cmn024

Family Practice Advance Access published on 30 May 2008

A simple risk score identifies individuals at high risk of developing Type 2 diabetes: a prospective cohort study Mushtaqur Rahmana, Rebecca K Simmonsb, Anne-Helen Hardingb, Nicholas J Warehamb and Simon J Griffinb Rahman M, Simmons RK, Harding A-H, Wareham NJ and Griffin SJ. A simple risk score identifies individuals at high risk of developing Type 2 diabetes: a prospective cohort study. Family Practice 2008; 25: 191–196. Background. Randomized trials have demonstrated that Type 2 diabetes is preventable among high-risk individuals. To date, such individuals have been identified through population screening using the oral glucose tolerance test. Objective. To assess whether a risk score comprising only routinely collected non-biochemical parameters was effective in identifying those at risk of developing Type 2 diabetes. Methods. Population-based prospective cohort (European Prospective Investigation of CancerNorfolk). Participants aged 40–79 recruited from UK general practices attended a health check between 1993 and 1998 (n = 25 639) and were followed for a mean of 5 years for diabetes incidence. The Cambridge Diabetes Risk Score was computed for 24 495 individuals with baseline data on age, sex, prescription of steroids and anti-hypertensive medication, family history of diabetes, body mass index and smoking status. We examined the incidence of diabetes across quintiles of the risk score and plotted a receiver operating characteristic (ROC) curve to assess discrimination. Results. There were 323 new cases of diabetes, a cumulative incidence of 2.76/1000 personyears. Those in the top quintile of risk were 22 times more likely to develop diabetes than those in the bottom quintile (odds ratio 22.3; 95% CI: 11.0–45.4). In all, 54% of all clinically incident cases occurred in individuals in the top quintile of risk (risk score > 0.37). The area under the ROC was 74.5%. Conclusion. The risk score is a simple, effective tool for the identification of those at risk of developing Type 2 diabetes. Such methods may be more feasible than mass population screening with biochemical tests in defining target populations for prevention programmes. Keywords. Diabetes, general practice, incidence, risk score, screening. acceptable and cost-effective way of identifying people who might benefit from health promotion interventions aimed at reducing the burden of disease associated with hyperglycaemia. Finding simpler, more pragmatic methods for identifying individuals at high risk of future progression to diabetes and who might benefit from targeted prevention is an important goal. Various strategies have been developed to identify those with prevalent but undiagnosed diabetes,7,8 and some of these have also been assessed for their ability to predict the future development of diabetes.9–11 All have required that people in the target population

Introduction Type 2 diabetes is a growing public health problem1 and gives rise to significant morbidity, mortality and long-term financial costs.2,3 There is now strong evidence that diabetes is preventable through changes in key behaviours such as diet and physical activity.4–6 However, to date, the major diabetes primary prevention trials have targeted people with impaired glucose tolerance (IGT), which can only be identified through an oral glucose tolerance test.5,6 Mass population screening by such means is unlikely to be an

a General Practice and Primary Care Research Unit, Department of Public Health and Primary Care, University of Cambridge, Cambridge and bMRC Epidemiology Unit, Institute of Metabolic Science, Box 285, Addenbrooke’s Hospital, Hills Road, Cambridge, CB2 0QQ, UK. Correspondence to Simon Griffin, MRC Epidemiology Unit, Institute of Metabolic Science, Box 285, Addenbrooke’s Hospital, Hills Road, Cambridge, CB2 0QQ, UK; Email: [email protected]

191

192

Family Practice—an international journal

complete a questionnaire or provide a blood sample. In contrast, the Cambridge Diabetes Risk Score only includes information on non-biochemical risk factors that are increasingly becoming routinely available to the family doctor. The risk score was independently developed and validated in population-based crosssectional studies and demonstrated a sensitivity of 77% and a specificity of 72% for the detection of prevalent undiagnosed diabetes compared to the oral glucose tolerance test in a primary care population.12 In this report, we aimed to evaluate how well the score would predict the future clinical incidence of diabetes in a large, population-based cohort from the European Prospective Investigation of Cancer (EPIC)-Norfolk study.

Methods Study design EPIC-Norfolk is a prospective cohort study in which men and women aged 40–79 were recruited from general practices in the Norfolk region. Full details of the population are reported elsewhere.13 In brief, between 1993 and 1998, 77 630 individuals were invited to take part and 25 639 (33%) attended a baseline health examination. This included anthropometric and blood pressure measurements and completion of a general health questionnaire, with questions on personal and family history of disease, medication and lifestyle factors, including smoking habits. Participants were also asked whether a doctor had ever told them that they had any of the conditions contained in a list that included diabetes, heart attack and stroke. Non-fasting blood samples were taken, and starting in 1995 when funding became available, glycosylated haemoglobin (HbA1c) was measured on fresh ethylenediaminetetraacetic acid blood samples using high-performance liquid chromatography in a single laboratory with a Diabetes Control and Complications Trial-defined normal range (Bio-Rad Diamat Automated Glycated Haemoglobin Analyser, Hemel Hempstead, UK). There was a second health check between 1998 and 2000 and 15 028 (58.6%) participants returned for this assessment. The Norfolk area is slightly healthier than the general UK population with a standardized mortality ratio of 94 (Office for National Statistics, UK). However, EPIC-Norfolk is similar to a nationally representative sample for anthropometric variables, blood pressure and serum lipids.13 Incident cases of diabetes were ascertained from multiple sources including self-report at follow-up health checks in EPIC, searches of hospital and general practice registers, death certificates and data on prescribing of hypoglycaemic medication.14 A capture–recapture analysis was conducted to assess case ascertainment rate. Medical records of cases identified

only through the general practice diabetes register and of cases identified only through self-report, together with matched controls, were reviewed using a standardized process. The reviewer was blinded to the case or control status of the participants. This process indicated that 99% case ascertainment was attained.14 Statistical analysis Baseline characteristics were summarized separately in men and women and compared using unpaired t-tests for continuous variables and chi-square tests for categorical variables. We calculated the Cambridge Diabetes Risk score for each EPIC-Norfolk participant using published regression coefficients12 for baseline age, sex, family history of diabetes, smoking status, prescription of steroid or anti-hypertensive medication and body mass index (BMI; kg/m2) (Fig. 1). Logistic regression was used to assess the association between each Cambridge Risk Score variable and clinically incident diabetes. We re-coded the risk score into quintiles and compared the risk of developing diabetes in each quintile with the lowest category of risk (reference group). To test the performance of the score in identifying individuals with incident diabetes, we calculated the sensitivity, specificity and positive likelihood ratio value at different risk thresholds and plotted a receiver operating characteristic (ROC) curve to examine the discrimination of the score. The EPIC-Norfolk cohort may have included some individuals with previously undiagnosed diabetes at baseline and excluded those with a biochemical, but not clinical, diagnosis of diabetes at follow-up. A sensitivity analysis of 11 140 participants with baseline and follow-up HbA1c measures was therefore undertaken. An HbA1c level of greater than 7% at baseline was used to exclude those likely to have prevalent but undiagnosed diabetes, while incident diabetes was defined as an HbA1c level of greater than 7% at followup, in addition to health records and self-report. A value of HbA1c of 7% has the maximum specificity and sensitivity for detecting diabetes15; this value has also been shown to predict nephropathy and retinopathy as well as fasting blood glucose and 2-hour postchallenge blood glucose.16 All analyses were completed using Stata Version 7 (STATA Corp., College Station, TX). The EPICNorfolk study was approved by the Norfolk Local Research Ethics Committee and participants gave written consent prior to the first health check.

Results Individuals with known diabetes at baseline (n = 845) and those with missing data for the variables in the risk score (n = 299) were excluded, leaving 24 495

Predicting incident diabetes using a simple risk score

participants for analysis. Baseline characteristics for the study population are shown in Table 1. Men and women were aged 58 years on average. Women had significantly lower baseline values for BMI and systolic and diastolic blood pressure than men. Women were also significantly more likely to have a family history of diabetes and to have never smoked. This study accumulated a total of 117 027 personyears of follow-up, with a mean duration of 4.8 ± 1.3 (standard deviation) years. During this time, 323 people were diagnosed with diabetes, a cumulative incidence rate of 2.76 per 1000 person-years. Table 2 shows the association between individual risk score

193

variables and the clinical incidence of diabetes. Age and BMI were significantly related to diabetes at follow-up. Men, those with a positive smoking history or family history of diabetes and those prescribed antihypertensive medication were also at increased risk of developing diabetes. In total, 23% of individuals had a risk score > 0.37 (top quintile). These individuals were 22 times more likely to develop diabetes than those in the bottom quintile [odds ratio (OR) 22.3; 95% CI: 11.0–45.4] (Fig. 2). Over half (54%) of those with incident diabetes had a risk score in the top quintile (>0.37). The cumulative incidence of diabetes for this group was 7.5 per 1000 person-years. A threshold of 0.38 produced a likelihood ratio of a positive test of 2.73 (95% CI: 2.46–3.03) (Table 3). The area under the ROC was 74.5% (Fig. 3). Sensitivity analyses in the subgroup of participants with HbA1c data demonstrated that the association between the top quintile of the Cambridge Risk Score and incident diabetes was attenuated but remained highly statistically significant (OR 10.4; 95% CI: 3.1–34.3).

Discussion

FIGURE 1 The Cambridge Diabetes Risk Score

TABLE 1

EPIC-Norfolk cohort baseline characteristics, by men and women (n = 24 495) Men

Sex % (n) 45 (11 030) Age in yearsa 58.9 (9.3) % With positive family history of diabetes (n)a None 88.5 (9764) Parent or sibling affected 10.7 (1175) Parent and sibling affected 0.8 (91) 26.5 (3.3) BMI in kg/m2 137 (18) Systolic blood pressure in mmHga 84 (11) Diastolic blood pressure in mmHga % Prescribed medication (n) Anti-dyslipidemia 1.4 (156) Anti-hypertensive 17.6 (1940) Steroid 3.0 (331) % Cigarette smoking (n)a Never 33.7 (3714) Ex-smoker 54.0 (5962) Current smoker 12.3 (1354) a

Sex difference: P < 0.001.

Women 55 (13 465) 58.2 (9.3) 87.2 (11 734) 12.0 (1618) 0.8 (113) 26.2 (4.3) 134 (19) 81 (11) 1.4 (183) 18.0 (2426) 3.4 (457) 56.5 (7613) 32.0 (4300) 11.5 (1552)

The Cambridge Diabetes Risk Score performs moderately well at predicting clinically incident Type 2 diabetes in the EPIC-Norfolk cohort, with an area under the ROC of 75%. Those in the top quintile of risk were 22 times more likely to develop diabetes than those in the bottom quintile. The Cambridge Risk Score appears to be a simple and effective tool for identification of those at risk of developing Type 2 diabetes using routinely collected non-biochemical information from patient records. The Cambridge Risk Score has been shown to predict undiagnosed prevalent diabetes with an area under the ROC of 80% in a cross-sectional study,12 as well as undiagnosed hyperglycaemia,17,18 metabolic syndrome18 and all-cause mortality.19 It also works well without information on family history and smoking,12 factors which used to be poorly recorded in general practice. In this study, we have shown that it can also identify individuals at risk of developing incident Type 2 diabetes in a large prospective cohort. Furthermore, we have demonstrated that the Cambridge Risk Score has good predictive value when tested in a population distinct from the one in which it was originally developed. Methods for identifying individuals at future risk of diabetes have included biochemical tests, questionnaires and surveys on dietary habits and physical activity.9,11,20,21 However, such methods either require laboratory testing or the collection of new information, which will incur cost and may be associated with

194

Family Practice—an international journal

FIGURE 2 Association of Cambridge Risk Score quintiles with clinically incident diabetes in the EPIC-Norfolk cohort

as self-report to identify individuals with diabetes at follow-up. Similarly, a German Diabetes Risk Score demonstrated an area under the ROC of 0.82–0.84 in different populations using information on age, waist circumference, height, history of hypertension, physical activity, smoking, and consumption of red meat, whole-grain bread, coffee, and alcohol, but this score required the collection of detailed information on physical activity and dietary behaviour.11 The Cambridge Diabetes Risk Score, which uses information routinely available in primary care records, could be used to help identify individuals or subgroups of the population who might benefit from further investigation, e.g. with an oral glucose tolerance test (OGTT) and more comprehensive risk assessment, or even direct preventive action. The key issues for identification of a high-risk group for targeted prevention are the ease with which the group can be identified, the total number of individuals identified as being at high risk and the level of absolute risk in that group, which when combined with the efficacy of the preventive intervention impacts on the number needed to treat to prevent one incident case. Defining a high-risk group on the basis of biochemical results such as IGT clearly identifies individuals at high absolute risk of progression to diabetes,23,24 but the major problem is one of feasibility. By contrast, the risk score that we have used in this study is easy to employ and does not require the collection of new information. Using the approach described here, 20% of individuals could be defined as being at high risk of progression to diabetes and 54% of incident cases would come from that group. Those with scores > 0.37 (top quintile) were five times more likely to be diagnosed with diabetes than those with scores < 0.37 (bottom four quintiles). If an intervention similar to that used in the Diabetes Prevention Program5 were targeted at this group, then 48 individuals with a high-risk score would require a lifestyle intervention to prevent one case of diabetes. The relative merits of risk prediction and direct preventive action compared to the approach of labelling individuals as having a biochemical disorder with subsequent behavioural and pharmacological therapy need to be assessed. Similarly, a preventive strategy involving a risk score as part of systematic stepwise risk assessment would need formal evaluation, e.g. in a randomized trial.

false reassurance or anxiety. In addition, the use of arbitrary thresholds in single tests may not be as beneficial as the use of multiple risk factors in risk assessment.22 A previous diabetes risk score has shown the effectiveness of simple clinical parameters in the identification of those at risk of diabetes, but only drug-treated diabetes was used as the outcome measure in this study.9 We used patient records as well

Strengths and limitations EPIC-Norfolk is a large population-based cohort, so the performance of the risk score was unlikely to be affected by spectrum bias, e.g. if an instrument is found to be effective in identifying a condition in one setting, it may not be as effective in another clinical situation where the spectrum of disease is different. Indeed, the Cambridge Risk Score has been found to be effective in predicting a range of related

TABLE 2 Crude association between Cambridge Risk Score variables and clinically incident diabetes in the EPIC-Norfolk cohort (n = 24495) Without diabetes

% (n) 98.7 (24 172) % Male (n) 44.9 (10 844) % Prescribed anti17.6 (4256) hypertensive medication (n) % Prescribed steroid 3.2 (775) medication (n) Age in years 58.5 (9.3) BMI in kg/m2–% (n) 30 14.5 (3499) Family history of diabetes–% (n) No family history 87.9 (21 235) Parent or sibling 11.3 (2741) with diabetes Parent and sibling 0.8 (196) with diabetes Cigarette smoking–% (n) Never smoked 46.4 (11 210) Former smoker 41.7 (10 091) Current smoker 11.9 (2871)

With incident diabetes

Crude OR (95% CI)

1.3 (323) 57.6 (186) 34.1 (110)

– 1.67 (1.34–2.08) 2.42 (1.92–3.05)

4.0 (13)

1.27 (0.72–2.22)

61.9 (8.2)

1.04 (1.03–1.05)

12.7 (41) 22.9 (74) 22.0 (71) 42.4 (137)

1.00 2.59 (1.77–3.80) 4.04 (2.75–5.95) 9.29 (6.54–13.2)

81.4 (263) 16.1 (52)

1.00 1.53 (1.13–2.07)

2.5 (8)

3.30 (1.61–6.75)

36.2 (117) 53.0 (171) 10.8 (35)

1.00 1.62 (1.28–2.06) 1.17 (0.80–1.71)

All values are means (standard deviation) unless specified otherwise

Risk Score 0.0 – 0.037

1.0

0.038 – 0.084

3.2 (1.4 to 7.0)

0.085 – 0.17

5.7 (2.7 to 12.1)

0.18 – 0.36

8.8 (4.2 to 18.4)

0.37 – 1.0

22.3 (11.0 to 45.4)

123 5 8 10

15

20

40

50

Odds Ratio (95% CI)

195

Predicting incident diabetes using a simple risk score

TABLE 3 Performance of the Cambridge Risk Score in identifying people with incident diabetes by quintiles of risk score threshold in the EPICNorfolk cohort Risk score threshold % >Threshold (n) Sensitivity % (95% CI) Specificity % (95% CI) Likelihood ratio positive (95% CI)

0.37 20.4 (5002) 54.5 (48.9–60.0) 80.0 (79.5–80.5) 2.73 (2.46–3.03)

0.18 40.2 (9848) 75.9 (70.8–80.4) 60.3 (59.7–0.61) 1.91 (1.79–2.03)

0.085 60.1 (14 717) 89.8 (86.0–92.9) 40.3 (39.7–40.9) 1.50 (1.45–1.56)

0.038 80.0 (19 592) 97.5 (95.2–98.9) 20.3 (19.7–20.8) 1.22 (1.20–1.25)

Risk score range: 0.21–0.99.

conditions in different populations.17,18,25 High ascertainment of diabetes incidence was achieved14 and overall incidence was similar to figures reported in other UK studies looking at comparable populations.26 EPIC-Norfolk is a predominantly Caucasian cohort and the sensitivity and specificity of the original Cambridge Risk Score have been shown to be reduced in predicting undiagnosed hyperglycaemia in Caribbean and South Asian populations living in the UK.25 Diabetes risk scores do not typically perform as well in populations in which they were not developed and the Cambridge Risk Score will need to be validated in other prospective cohorts. The accuracy of any risk score depends on the level of association of different risk factors with diabetes and the prevalence of risk factors, which will change over time and in different populations. As such, there is a need to reform, develop and update existing risk scores (as associations and beta coefficients change) to improve prediction of risk. Finally, the Cambridge Risk Score is based on information that might not be readily available from patient records in less-developed health care settings and may need to be modified accordingly. Similarly, the risk score might perform less well in routine practice due to less precise assessment of risk factors than occurred in this study. The length of follow-up in the EPIC-Norfolk cohort was relatively short. Thus, while the risk score may be good at identifying individuals rapidly progressing to diabetes, it may miss those with a slower onset. Alternatively, longer follow-up might increase the ORs for the association between the risk score and the incidence of diabetes if more of those in the highest risk score category go on to develop diabetes. In terms of our definition of incident diabetes, some selection bias may be present as people attending GP surgeries may have been more likely to be tested and, consequently, diagnosed with diabetes, e.g. if they were visibly overweight, or reported an unhealthy diet and/or low levels of physical activity. Thus, those individuals with diabetes who had low-risk scores may have been missed through a lack of testing. However, as we achieved 99% case ascertainment for all EPICNorfolk participants, whether or not they returned for the second health check, few clinically incident

FIGURE 3 Graph showing area under the ROC curve for the Cambridge Risk Score in assessing incident diabetes in the EPIC-Norfolk cohort

individuals with diabetes are likely to have been missed. In addition, incident diabetes was also defined biochemically using HbA1c > 7% in our sensitivity analyses, and while the association was attenuated due to the exclusion of those with prevalent undiagnosed diabetes at baseline, it remained significant, indicating that the effect of selection bias is unlikely to be large.

Conclusion The Cambridge Diabetes Risk Score is a simple, effective tool for the identification of those at increased risk of future incident diabetes, as well as prevalent undiagnosed diabetes12,25,17 and mortality.19 As it comprises information routinely available in primary care, it may have a role in defining individuals and populations for programmes of testing, treatment and prevention.

Acknowledgements We gratefully acknowledge the contribution of EPICNorfolk participants.

196

Family Practice—an international journal

Declaration Funding: British Heart Foundation; Cancer Research UK; Department of Health; Food Standards Agency; European Union (Europe Against Cancer Programme); Ministry of Agriculture, Fisheries and Food; Medical Research Council; Stroke Association; Wellcome Trust; World Health Organization; NHS Executive Eastern (MR). Ethical approval: The Norwich District Ethics Committee approved this study and participants gave written consent prior to the investigations. Conflicts of interest: None declared.

11

12

13

14

15

16

References 1

2

3

4

5

6

7

8

9

10

Amos A, McCarty D, Zimmet P. The rising global burden of diabetes and its complications: estimates and projections to the year 2010. Diabet Med 1997; 14: 1–85. Marks L. Counting the Cost: The Real Impact of Non-InsulinDependent Diabetes. London: British Diabetic Association, 1996. UKPDS Group. UK prospective diabetes study 6. Complications in newly diagnosed type 2 diabetic patients and their association with different clinical and biochemical risk factors. Diabetes Res 1990; 13: 1–11. Pan X-R, Li GW, Hu YH et al. Effects of diet and exercise in preventing NIDDM in people with impaired glucose tolerance. The Da Qing IGT and diabetes study. Diabetes Care 1997; 20: 537–544. Knowler W, Barrett-Connor E, Fowler SE et al. Reduction in the incidence of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med 2002; 346: 393–403. Tuomilehto J, Lindstrom J, Eriksson JG et al. Prevention of type 2 diabetes mellitus by changes in lifestyle among subjects with impaired glucose tolerance. N Engl J Med 2001; 344: 1343–1350. Tabaei B, Herman W. A multivariate logistic regression equation to screen for diabetes. Diabetes Care 2002; 25: 1999–2003. Glumer C, Carstensen B, Sandbaek A et al. A Danish diabetes risk score for targeted screening. Diabetes Care 2004; 27: 727–733. Lindstrom J, Tuomilehto J. The Diabetes Risk Score: a practical tool to predict type 2 diabetes risk. Diabetes Care 2003; 26: 725–731. Aekplakorn W, Bunnag P, Woodward M et al. A risk score for predicting incident diabetes in the Thai population. Diabetes Care 2006; 29: 1872–1877.

17

18

19

20

21

22

23

24

25

26

Schulze MB, Hoffmann K, Boeing H et al. An accurate risk score based on anthropometric, dietary, and lifestyle factors to predict the development of type 2 diabetes. Diabetes Care 2007; 30: 510–515. Griffin SJ, Little PS, Hales CN et al. Diabetes risk score: towards earlier detection of type 2 diabetes in general practice. Diabetes Metab Res Rev 2000; 16: 164–171. Day N, Oakes S, Luben R et al. EPIC-Norfolk: study design and characteristics of the cohort. European Prospective Investigation of Cancer. Br J Cancer 1999; 80 (suppl 1): 95–103. Harding AH, Day NE, Khaw KT et al. Dietary fat and the risk of clinical type 2 diabetes: the European prospective investigation of Cancer-Norfolk study. Am J Epidemiol 2004; 159: 73–82. Rowley K, Daniel M, O’Dea K. Screening for diabetes in indigenous populations using glycated haemoglobin: sensitivity, specificity, post-test likelihood and risk of disease. Diabet Med 2005; 22: 833–839. McCance DR, Hanson RL, Charles MA et al. Comparison of tests for glycated haemoglobin and fasting and two hour plasma glucose concentrations as diagnostic methods for diabetes. BMJ 1994; 308: 1323–1328. Park PJ, Griffen SJ, Sargeant L et al. The performance of a risk score in predicting undiagnosed hyperglycemia. Diabetes Care 2002; 25: 984–988. Heldgaard PE, Griffin SJ. Routinely collected general practice data aids identification of people with hyperglycaemia and metabolic syndrome. Diabet Med 2006; 23: 996–1002. Spijkerman A, Griffin SJ, Dekker J et al. What is the risk of mortality for people who are screen positive in a diabetes screening programme but who do not have diabetes on biochemical testing? Diabetes screening programmes from a public health perspective. J Med Screen 2002; 9: 187–190. Stern M, Williams K, Haffner S. Identification of persons at high risk for type 2 diabetes mellitus: do we need the oral glucose tolerance test? Ann Intern Med 2002; 136: 575–581. Schmidt M, Duncan BB, Bang H et al. Identifying individuals at high risk for diabetes. Diabetes Care 2005; 28: 2013–2018. Brindle P, Fahey T. Primary prevention of coronary heart disease. BMJ 2002; 325: 56–57. Williams DR, Wareham NJ, Brown DC et al. Undiagnosed glucose intolerance in the community: the Isle of Ely diabetes project. Diabet Med 1995; 12: 30–35. Wareham NJ, Bryne CD, Williams R et al. Fasting proinsulin concentrations predict the development of type 2 diabetes. Diabetes Care 1999; 22: 262–270. Spijkerman AM, Yuyun MF, Griffin SJ et al. The performance of a risk score as a screening test for undiagnosed hyperglycaemia in ethnic minority groups: data from the 1999 health survey for England. Diabetes Care 2004; 27: 116–122. Perry IJ, Wannamethee SJ, Walker MK et al. Prospective study of risk factors for development of non-insulin dependent diabetes in middle aged British men. BMJ 1995; 310: 560–564.