European Journal of Cardiovascular Prevention & Rehabilitation http://cpr.sagepub.

com/

Contemporary model for cardiovascular risk prediction in people with type 2 diabetes
Andre Pascal Kengne, Anushka Patel, Michel Marre, Florence Travert, Michel Lievre, Sophia Zoungas, John Chalmers, Stephen Colagiuri, Diederick E Grobbee, Pavel Hamet, Simon Heller, Bruce Neal and Mark Woodward European Journal of Cardiovascular Prevention & Rehabilitation 2011 18: 393 originally published online 28 February 2011 DOI: 10.1177/1741826710394270 The online version of this article can be found at: http://cpr.sagepub.com/content/18/3/393

Published by:
http://www.sagepublications.com

On behalf of:
European Society of Cardiology

European Association for Cardiovascular Prevention and Rehabilitation

Additional services and information for European Journal of Cardiovascular Prevention & Rehabilitation can be found at: Email Alerts: http://cpr.sagepub.com/cgi/alerts Subscriptions: http://cpr.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav

>> Version of Record - May 18, 2011 OnlineFirst Version of Record - Feb 28, 2011 What is This?
Downloaded from cpr.sagepub.com by guest on September 21, 2013

Original scientific paper

EURO PEAN SO CIETY O F CARDIOLOGY

Contemporary model for cardiovascular risk prediction in people with type 2 diabetes
Andre Pascal Kengne1, Anushka Patel1, Michel Marre2, Florence Travert2, Michel Lievre3, Sophia Zoungas1, John Chalmers1, Stephen Colagiuri4, Diederick E Grobbee5, Pavel Hamet6, Simon Heller7, Bruce Neal1, Mark Woodward1 (on behalf of the ADVANCE Collaborative Group)

European Journal of Cardiovascular Prevention & Rehabilitation 18(3) 393398 ! The European Society of Cardiology 2011 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/1741826710394270 ejcpr.sagepub.com

Abstract Background: Existing cardiovascular risk prediction equations perform non-optimally in different populations with diabetes. Thus, there is a continuing need to develop new equations that will reliably estimate cardiovascular disease (CVD) risk and offer flexibility for adaptation in various settings. This report presents a contemporary model for predicting cardiovascular risk in people with type 2 diabetes mellitus. Design and methods: A 4.5-year follow-up of the Action in Diabetes and Vascular disease: preterax and diamicron-MR controlled evaluation (ADVANCE) cohort was used to estimate coefficients for significant predictors of CVD using Cox models. Similar Cox models were used to fit the 4-year risk of CVD in 7168 participants without previous CVD. The models applicability was tested on the same sample and another dataset. Results: A total of 473 major cardiovascular events were recorded during follow-up. Age at diagnosis, known duration of diabetes, sex, pulse pressure, treated hypertension, atrial fibrillation, retinopathy, HbA1c, urinary albumin/creatinine ratio and non-HDL cholesterol at baseline were significant predictors of cardiovascular events. The model developed using these predictors displayed an acceptable discrimination (c-statistic: 0.70) and good calibration during internal validation. The external applicability of the model was tested on an independent cohort of individuals with type 2 diabetes, where similar discrimination was demonstrated (c-statistic: 0.69). Conclusions: Major cardiovascular events in contemporary populations with type 2 diabetes can be predicted on the basis of routinely measured clinical and biological variables. The model presented here can be used to quantify risk and guide the intensity of treatment in people with diabetes.

Keywords Diabetes mellitus, cardiovascular diseases, prognosis, prevention, risk score

Received 10 June 2010; accepted 19 November 2010

Introduction
Preventive cardiovascular medicine increasingly relies on modelling techniques to estimate the individuals absolute risk of a future cardiovascular event in
1 The George Institute for International Health; The University of Sydney, Australia. 2 pital Bichat-Claude Bernard and Universite Paris 7, Paris, France. Ho 3 Claude Bernard, Lyon, Service de Pharmacologie Clinique, Universite France.

4 Institute of Obesity, Nutrition & Exercise, Faculty of Medicine, The University of Sydney, Australia. 5 Julius Centre for Health Sciences and Primary Care, University Medical Centre Utrecht, Utrecht, The Netherlands. 6 de Montre al; Universite de Montre al, Centre Hospitalier de lUniversite al, Canada. Montre 7 University of Sheffield and Sheffield Teaching Hospitals national Service (NHS) Foundation Trust, Sheffield, United Kingdom.

Corresponding author: John Chalmers, The George Institute for International Health, University of Sydney, PO Box M201, Level 10 KGV Building, RPAH, Missenden Road, Camperdown, Sydney NSW 2050, Australia Email: chalmers@georgeinstitute.org.au

394

European Journal of Cardiovascular Prevention & Rehabilitation 18(3) Outcomes were coded according to the 10th revision of the International Classication of Diseases (ICD-10) and centrally validated by an independent endpoint committee.

order to inform decisions about therapeutic approaches.1 Risk models have been developed for the general population and, separately, for people with diabetes.2,3 However, these risk models often do not perform optimally when applied to populations other than the one from which the tool was derived.35 Thus, there is a continuing need to develop risk prediction tools that will reliably estimate cardiovascular disease (CVD) risk and oer exibility for adaptation in various economic and geographical settings. In particular, the models should allow for evolution in treatment practices and take account of the totality of glucose exposure. Data from the multi-centre, worldwide, clinical trial (Action in Diabetes and Vascular disease: preterax and diamicron-MR controlled evaluation) (ADVANCE)6,7 permit the derivation of new equations for cardiovascular risk prediction in people with diabetes, whilst addressing some of these issues.

Validation cohort
DIABHYCAR (The non-insulin-dependent diabetes, hypertension, microalbuminuria or proteinuria, cardiovascular events, and ramipril study) was a clinical trial of ramipril among individuals with type 2 diabetes conducted in 16 countries between 1995 and 2001.9 Of the 4912 randomized participants, 3711 had no history of CVD at baseline. Among these, 1836 had complete data, suitable for use in validation. Denitions of CVD in DIABHYCAR were similar to those in ADVANCE, and all outcomes were centrally validated by an independent endpoint committee. During 4 years of follow-up, DIABHYCAR recorded 183 CVD events.

Research design and methods

ADVANCE was a factorial randomized controlled trial of blood pressure (perindopril-indapamide versus placebo) and glucose control (glicazide MR-based intensive intervention versus standard care) on the incidence of microvascular and macrovascular events among 11,140 high risk individuals with type 2 diabetes, recruited from 215 centres across 20 countries in Asia, Australasia, Europe and Canada.8 Participants had to be at least 55 years of age at entry, diagnosed with diabetes at the age of 30 years or older, and be at increased risk of vascular events. CVD risk derivation was restricted to the 7168 participants (53.6% men) free of known CVD at baseline, with data available on predictors of interest. Approval for the conduct of the ADVANCE study was obtained from the Institutional Ethics Committee of each participating centre and all participants provided written informed consent. Baseline levels of predictors were used in this analysis. Cigarette smoking was categorized as current or not current. Pulse pressure (PP) (mmHg) was the dierence between systolic (SBP) and diastolic (DBP) blood pressure and mean arterial blood pressure (MAP) was DBP PP/3. Non-HDL cholesterol was total (TC) minus HDL cholesterol, and the lipid ratio was TC/ HDL-cholesterol. Retinopathy was dened by a history/presence of typical changes due to diabetes, regardless of the level of severity, or previous laser treatment, and atrial brillation by known history or reported presence of atrial brillation on ECG at baseline. Use of medications was self-reported. CVD was dened as fatal or non-fatal myocardial infarction or stroke or cardiovascular death.68

Statistical analysis
Cox regression models10 were used to relate predictors to incident CVD during the rst 4.5 years (75th percentile) of follow-up in ADVANCE (blood pressure arm). This time frame was chosen to avoid the likely disproportionate inuence on parameter estimates of longer survival among a few participants. Predictors considered were age at diagnosis of diabetes, duration of diagnosed diabetes, sex, SBP, DBP, MAP, PP, TC, HDL and non-HDL cholesterol, lipid ratio and triglycerides, body mass index (BMI), waist circumference, waist-to-hip ratio, bloodpressure-lowering medication (i.e. treated hypertension), statin use, current smoking, retinopathy, atrial brillation (past or present), logarithmically transformed urinary albumin/creatinine ratio (ACR) and serum creatinine (Scr ), haemoglobin A1c (HbA1c), fasting blood glucose and randomized treatments (blood pressure lowering and glucose control regimens). Variables were selected for inclusion in the nal model following a stepwise approach11 and standard model checking was undertaken.12 For blood pressure variables, when more than one index was signicant in univariate models, the one with the great^ was rst entered into the est eect on the -2log L multivariable model. The variable included in the nal model was the one associated with the smallest Akaikes information criterion. A similar approach was used to choose between athropometric variables, as well as baseline HbA1c and fasting blood glucose. Sex interactions, and interactions between

Kengne et al. blood-pressure-lowering treatment and blood-pressure variables were also tested.

395

Results
Table 1 shows the characteristics of the study participants who were free of CVD at baseline. During the rst 4.5 years of follow-up, 473 CVD events were recorded. In univariate Cox models, age at diagnosis, duration of diabetes, sex, SBP, PP and MAP, waist-to-hip ratio, treated hypertension, use of aspirin, retinopathy, HbA1c, fasting glucose, total and non-HDL cholesterol, ratio total/HDL cholesterol, log of ACR and log of Scr were all signicant predictors. Table 2 shows the nal multivariable model in which all predictors were independently signicant (p 0.022).

Model validation. Calibration was assessed via a

Hosmer and Lemeshow (HL) test.13 Discrimination was assessed by the area under the receiver-operating characteristic curve (AUC). Internal validation of the models used 1000 bootstrap re-samples. The model was also assessed, and compared with currently available risk models,14,15 on DIABHYCAR. A 4-year risk score table was developed from the nal Cox model by assigning points to each predictor.16 All analyses were performed using SAS/STAT software v.9.1 (SAS Institute Inc., Cary, NC, USA).

Table 1. Baseline clinical and biochemical characteristics of 7168 ADVANCE and 1836 DIABHYCAR participants free of any history of CVD at baseline ADVANCE (n 7168) 3327 (46.4) 65.8 (6.3) 57.9 (8.8) 7.9 (6.3) 5.27 (1.18) 1.28 (0.36) 4.37 (1.38) 4.0 (1.13) 1.95 (1.32) 145.3 (21.3) 80.7 (10.9) 64.6 (16.8) 102.2 (13.0) 28.2 (5.3) 98.0 (13.2) 0.93 (0.08) 165.3 (9.5) 8.55 (2.78) 7.54 (1.57) 2.83 (1.37) 4.40 (0.25) 1068 (14.9) 1711 (23.9) 4614 (64.4) 1517 (21.2) 2297 (32.1) 388 (5.4) DIABHYCAR included (n 1836) 554 (30.2) 64.4 (8.1) 54.8 (9.5) 9.6 (7.4) 5.78 (1.05) 1.32 (0.35) 4.65 (1.33) 4.47 (1.04) 2.19 (1.38) 144.8 (14.0) 82.2 (8.3) 62.6 (11.6) 103.1 (9.0) 28.6 (4.6) 166.8 (8.3) 9.73 (3.03) 7.88 (1.78) 4.44 (0.21) 272 (17.2) 966 (52.6) 640 (34.9) 158 (8.6) DIABHYCAR Excluded (n 1875) 673 (35.9) 63.8 (8.4) 54.8 (9.4) 8.8 (6.9) 5.75 (1.14) 1.33 (0.36) 4.54 (1.31) 4.42 (1.10) 2.21 (1.44) 147.1 (18.8) 83.6 (10) 63.6 (14.9) 104.8 (11.6) 29.1 (4.6) 166.7 (9.1) 10.06 (3.46) 7.55 (1.60) 4.45 (0.22) 291 (19) 988 (52.7) 349 (18.7) 193 (10.3) p value for the difference 0.0002 0.015 0.88 0.01 0.56 0.27 0.10 0.32 0.84 <0.0001 <0.0001 0.03 <0.0001 0.0002 N/A N/A 0.56 0.002 <0.0001 N/A 0.17 0.18 N/A 0.96 <0.0001 0.07 N/A

Variables Women, n (%) Age (years), mean (SD) Age at diagnosis (years), mean (SD) Known duration of diabetes (years), mean (SD) Total cholesterol (mmol/l), mean (SD) HDL cholesterol (mmol/l), mean (SD) TC-to-HDL ratio, mean (SD) Non-HDL cholesterol (mmol/l), mean (SD) Triglyceride (mmol/l), mean (SD) Blood pressure (mmHg) Systolic blood pressure, mean (SD) Diastolic blood pressure, mean (SD) Pulse pressure, mean (SD) Mean arterial pressure, mean (SD) Body mass index (kg/m2), mean (SD) Waist circumference (cm), mean (SD) Waist-to-hip ratio, mean (SD) Height (cm), mean (SD) Fasting blood glucose (mmol/l), mean (SD) HbA1c (%), mean (SD) Log of ACR, mean (SD) Log of serum creatinine, mean (SD) Current smoker (%) Retinopathy, n (%) Treated hypertension, n (%) Use of statins, n (%) Use of aspirin, n (%) Atrial fibrillation, n (%)

ACR, urinary albumin/creatinine ratio; HbA1c, haemoglobin A1c; HDL, high density lipoprotein; TC, total cholesterol.

396

European Journal of Cardiovascular Prevention & Rehabilitation 18(3)

Table 2. Beta coefficients and standard error (SE) for predictors in the final model Final model [S0(4) 0.951044] Variable Age at diagnosis (per 1 year increase) Sex (women vs men) Known duration of diabetes (per 1 year increase) Pulse pressure (per 1 mmHg increase) Retinopathy (yes vs no) Atrial fibrillation (old/present vs absent) HbA1c (per 1% increase) Log of urinary albumin/creatinine ratio (per 1 log mg/g increase) Non-HDL cholesterol (per 1 mmol/l increase) Treated hypertension (yes vs no) Parameter estimates (SE) 0.06187 0.4736 0.08263 0.00665 0.38248 0.60106 0.09945 0.19341 (0.00778) (0.09805) (0.00995) (0.002707) (0.10066) (0.15363) (0.02707) (0.03267) Hazard ratio (95% CI) 1.06 0.62 1.09 1.007 1.47 1.82 1.11 1.21 (1.051.08) (0.510.76) (1.071.11) (1.0011.012) (1.201.79) (1.352.66) (1.051.16) (1.141.29) p value <0.0001 <0.0001 <0.0001 0.016 0.0001 <0.0001 0.0002 <0.0001 0.0002 0.022 p functional form 0.195 0.305 0.315 0.700 0.210 0.335 p for PH 0.795 0.300 0.280 0.345 0.650 0.370 0.645 0.440 0.890 0.115

0.12619 (0.03437) 0.24219 (0.10575)

1.13 (1.061.21) 1.27 (1.041.57)

S0(4), 4-year baseline survival; CI, confidence interval; HbA1c, haemoglobin A1c; HDL, high density lipoprotein; PH, proportional hazard; SE, standard error.

None of the interactions tested reached statistical signicance (p > 0.07). Figure 1 shows the points score table for this nal model.

Discussion
We present a new cardiovascular risk equation specic for a large contemporary population of people with type 2 diabetes, treated according to current guidelines. The population is drawn from the ADVANCE trial that included patients from 20 countries worldwide. We also present validation of this risk equation in an independent population of patients with type 2 diabetes, participating in DIABHYCAR.9 Several risk predictions tools developed from prospective observational studies and clinical trials exist for risk stratication and prioritization of individuals with diabetes for cardiovascular risk reduction interventions. Most such tools have been derived from a homogenous population and have focused on specic manifestations of CVD.3 Nevertheless, risk equations for predicting combined CVD in people with diabetes have been published recently,1820 each of which has its own distinct limitations. Tools for total cardiovascular risk prediction in the general population, which can also be used in people with diabetes, have been published.3 The limitations of these tools have been detailed elsewhere.15 Inclusion in these tools of classical indicators of glucose exposure such as HbA1c will probably improve their performance in people with diabetes, but such improvement would still be sub-optimal. Known duration of diabetes allows for the assessment of the exposure after clinical diagnosis, but not before. Exposure prior to clinical diagnosis varies substantially and because it contributes to the overall risk, it has to be approximated in some way. Microvascular complications of diabetes may be

Model performance
The 4-year AUC (95% condence interval) was 0.702 (0.6760.728) when the equation was tested on ADVANCE itself. The sensitivity and specicity of the top fths were, respectively, 45% and 82%. The equivalents after bootstrap validation were 0.699 (0.6950.702); 44% and 79%, respectively. The HL test p 0.76, with little change after bootstrapping (Supplementary Figures 1 and 2). When the model was applied to DIABHYCAR, equivalent results were 0.685 (0.6460.724); 37% and 82%, respectively. There was a modest risk underestimation: predicted/observed risk (95% condence interval) 0.82 (0.710.95), HL p 0.032 (Figure 1 and Supplementary Figure 1).17 For comparison, equivalent measures for Framingham CVD scores in DIABHYCAR were: Anderson14 1.65 (95% CI: 1.411.93), p < 0.0001), AUC 0.646 (0.6030.689) (p 0.02 compared with the ADVANCE model); DAgostino15 1.99 (1.702.32), p < 0.0001), 0.638 (0.5960.680), (p 0.005 compared with ADVANCE). Based on a cut-o for 4-year predicted risk of !8%, the ADVANCE model identied 22% of ADVANCE participants, in whom 48% of all CVD events occurred, and 39% of the DIABHYCAR participants in whom 66% of CVD events occurred (Supplementary Figure 3).

Kengne et al.

397

Step 1 Age at diagnosis (years) 2934 3539 4044 4550 5156 5762 6368 6974 7580 8186 Step 2 Known duration (years) 0 15 610 1115 1620 2125 2630 3135 36+ Step 3 Sex Men Women Step 4 Atrial fibrillation No Old or present

Points 0 1 2 3 4 5 6 7 8 9

Step 5 Retinopathy No Yes Step 6 Treated hypertension No Yes Step 7

Pulse pressure (mmHg)

Points 0 1

Step 11 Sum up points from steps 1 through 10 Look up predicted 4-year risk of major CVD in the table

Points 0 1

Points 0 1 2 3 4 5 6 7 8

<50 50110 111+

Points 0 1 2

Predicted 4-year risk of major CVD

Total points

Step 8 HbA1c (%) <6 6<9 9+ Step 9 Albuminuria Normoalbuminuria Microalbuminuria Macroalbuminuria Step 10 Non HDL-c (mmol/l) <3 3<6 6<9 9+

Points 0 1 2

Points 0 1

Points 0 2 3

Points 0 2

Points 0 1 2 5

5 or less 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

4-year risk (%) <0.5 0.5 0.7 1.0 1.4 2.1 3.0 4.3 6.2 8.9 12.6 17.8 24.7 33.7 41.9 57.8 71.4 Above 83

Figure 1. Schema for estimating the 4-year risk of CVD by the ADVANCE model equation.

among the most reliable indicators of total exposure to higher glucose. Importantly microvascular eye diseases and albumin excretion rate, both indicators of microvascular disease in diabetes, have been found to be associated with cardiovascular outcomes.21,22 The major strengths of the present analyses are the large sample, the inclusion of participants from many countries, inclusion of indicators of microvascular retinal complications as predictors and the central adjudication of outcomes. Bootstrap and external validation are other advantages. There are also limitations. These include the selective nature of trial participants and the potential inuence of randomized treatments, which may aect the generalizability of the model. Comparison with available cross-sectional data suggest that ADVANCE participants are likely to be typical of current populations of patients with type 2 diabetes being managed in the community across many global settings.5 Based on the ndings from this analysis, the current recommendation by clinical practice guidelines,23 of treating all individuals with diabetes as risk-equivalent

to CVD survivors, may not address the actual continuum of risk in this population. While individuals with diabetes as a group are at high risk of CVD and should receive due attention, this risk is not uniformly distributed. In the ADVANCE population, in which most patients received preventive medications, the residual risk of CVD could accurately be captured by the model developed here. The model described reliably identied the one-quarter of participants in whom nearly half of all major CVD events were recorded. Further intensifying treatment in such a group is likely to achieve substantial further gains in preventing major cardiovascular events in the population.

Funding
J Chalmers holds research grants from Servier, administered through the University of Sydney as Co-Principal Investigator for ADVANCE. A Patel, S Colagiuri, S Heller, P Hamet, M Marre, CYu Pan, S Zoungas, DE Grobbee, B Neal, J Chalmers and M Woodward have received lecturing fees from Servier.

398 References

European Journal of Cardiovascular Prevention & Rehabilitation 18(3)

12. Lin D, Wei L and Yin Z. Checking the Cox model with cumulative sums of Martingales-based residuals. Biometrika 2003; 80: 557572. 13. DAgostino RB and Nam BH. Evaluation of the performance of survival analysis models: discrimination and calibration measures. In: Balakrishnan N and Rao CR (eds) Handbook of statistics. Amsterdam: Elsevier, 2004, pp.125. 14. Anderson KM, Odell PM, Wilson PW and Kannel WB. Cardiovascular disease risk profiles. Am Heart J 1991; 121(1 Pt 2): 293298. 15. DAgostino Sr RB, Vasan RS, Pencina MJ, Wolf PA, Cobain M, Massaro JM, et al. General cardiovascular risk profile for use in primary care: the Framingham Heart Study. Circulation 2008; 117: 743753. 16. Sullivan LM, Massaro JM and DAgostino Sr RB. Presentation of multivariate data for clinical use: The Framingham Study risk score functions. Stat Med 2004; 23: 16311660. 17. Rockhill B, Spiegelman D, Byrne C, Hunter DJ and Colditz GA. Validation of the Gail MH, Brinton LA, Byar DP, Corle DK, Green SB, Schairer C, Mulvihill JJ model of breast cancer risk prediction and implications for chemoprevention. J Natl Cancer Inst 2001; 93: 358366. 18. Cederholm J, Eeg-Olofsson K, Eliasson B, Zethelius B, Nilsson PM and Gudbjornsdottir S. Risk prediction of cardiovascular disease in type 2 diabetes: A risk equation from the Swedish National Diabetes Register (NDR). Diabetes Care 2008; 31: 20382043. 19. Coleman RL, Stevens RJ and Holman RR. Estimating cardiovascular disease risk for individuals with type 2 diabetes: the new UKPDS risk engine. Diabet Med 2007; 24(Suppl.1): 56. 20. Davis WA, Knuiman MW and Davis TM. An Australian cardiovascular risk equation for type 2 diabetes: The Fremantle Diabetes Study. Intern Med J 2010; 40: 286292. 21. Yang X, So WY, Kong AP, Ho CS, Lam CW, Stevens RJ, et al. Development and validation of stroke risk equation for Hong Kong Chinese patients with type 2 diabetes: the Hong Kong Diabetes Registry. Diabetes Care 2007; 30: 6570. 22. Yang X, So WY, Kong AP, Ma RC, Ko GT, Ho CS, et al. Development and validation of a total coronary heart disease risk score in type 2 diabetes mellitus. Am J Cardiol 2008; 101: 596601. 23. Buse JB, Ginsberg HN, Bakris GL, Clark NG, Costa F, Eckel R, et al. Primary prevention of cardiovascular diseases in people with diabetes mellitus: a scientific statement from the American Heart Association and the American Diabetes Association. Diabetes Care 2007; 30: 162172.

1. Graham I, Atar D, Borch-Johnsen K, Boysen G, Burell G, Cifkova R, et al. European guidelines on cardiovascular disease prevention in clinical practice: full text. Fourth Joint Task Force of the European Society of Cardiology and other societies on cardiovascular disease prevention in clinical practice (constituted by representatives of nine societies and by invited experts). Eur J Cardiovasc Prev Rehabil 2007; 14: S1S113. 2. Asia Pacific Cohort Studies Collaboration. Coronary risk prediction for those with and without diabetes. Eur J Cardiovasc Prev Rehabil 2006; 13: 3036. 3. Chamnan P, Simmons RK, Sharp SJ, Griffin SJ and Wareham NJ. Cardiovascular risk assessment scores for people with diabetes: a systematic review. Diabetologia 2009; 52: 20012014. 4. Brindle P, Beswick A, Fahey T and Ebrahim S. Accuracy and impact of risk assessment in the primary prevention of cardiovascular disease: a systematic review. Heart 2006; 92: 17521759. 5. Kengne AP, Patel A, Colagiuri S, Heller S, Hamet P, Marre M, et al. The Framingham and UKPDS risk equations do not reliably estimate the probability of cardiovascular events in a large ethnically diverse sample of patients with diabetes: the Action in Diabetes and Vascular Disease: Preterax and Diamicron-MR Controlled Evaluation (ADVANCE) Study. Diabetologia 2010; 53: 821831. 6. Patel A, MacMahon S, Chalmers J, Neal B, Billot L, Woodward M, et al. Intensive blood glucose control and vascular outcomes in patients with type 2 diabetes. N Engl J Med 2008; 358: 25602572. 7. Patel A, MacMahon S, Chalmers J, Neal B, Woodward M, Billot L, et al. Effects of a fixed combination of perindopril and indapamide on macrovascular and microvascular outcomes in patients with type 2 diabetes mellitus (the ADVANCE trial): a randomised controlled trial. Lancet 2007; 370: 829840. 8. ADVANCE Collaborative Group. Rationale and design of the ADVANCE study: a randomised trial of blood pressure lowering and intensive glucose control in highrisk individuals with type 2 diabetes mellitus. Action in Diabetes and Vascular Disease: PreterAx and DiamicroN Modified-Release Controlled Evaluation. J Hypertens Suppl 2001; 19: S21S28. 9. Marre M, Lievre M, Chatellier G, Mann JF, Passa P and Menard J. Effects of low dose ramipril on cardiovascular and renal outcomes in patients with type 2 diabetes and raised excretion of urinary albumin: randomised, double blind, placebo controlled trial (the DIABHYCAR study). BMJ 2004; 328: 495. 10. Cox DR. Regression models and lifetables. J Royal Stat Soc (series B) 1972; 34: 184220. 11. Collett D. Modelling survival data in medical research, 2nd ed. New York: Chapman & Hall/CRC, 2003, pp.55109.