Circ Cardiovasc Qual Outcomes 2016 05 11;9(3):213-21. Epub 2016 May 11.
From the Department of Vascular Medicine (L.K., F.L.J.V.) and Julius Centre for Health Sciences and Primary Care (Y.v.d.G.), University Medical Centre Utrecht, Utrecht, The Netherlands; ICCH, Imperial College London, London, United Kingdom (N.R.P., P.S.S.); Biomedical Research Institute, University of Dundee, Dundee, Scotland, United Kingdom (H.M.C., S.J.L.); Department of Cardiology, Academic Medical Centre, Amsterdam, The Netherlands (S.M.B.); and University of Texas Health Science Center, Houston (S.L.P., B.R.D.).
Background: In this study, we aimed to translate the average relative effect of statin therapy from trial data to the individual patient with type 2 diabetes mellitus by developing and validating a model to predict individualized absolute risk reductions (ARR) of cardiovascular events.
Methods And Results: Data of 2725 patients with type 2 diabetes mellitus from the Lipid Lowering Arm of the Anglo Scandinavian Cardiac Outcomes Trial (ASCOT-LLA) study (atorvastatin 10 mg versus placebo) were used for model derivation. The model was based on 8 clinical predictors including treatment allocation (statin/placebo). Ten-year individualized ARR on major cardiovascular events by statin therapy were calculated for each patient by subtracting the estimated on-treatment risk from the estimated off-treatment risk. Predicted 10-year ARR by statin therapy was <2% for 13% of the patients. About 30% had an ARR of >4% (median ARR, 3.2%; interquartile range, 2.5%-4.3%; 95% confidence interval for 3.2% ARR, -1.4% to 6.8%). Addition of treatment interactions did not improve model performance. Therefore, the wide distribution in ARR was a consequence of the underlying distribution in cardiovascular risk enrolled in these trials. External validation of the model was performed in data from the Antihypertensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial (ALLHAT-LLT; pravastatin 40 mg versus usual care) and Collaborative Atorvastatin Diabetes Study (CARDS; atorvastatin 10 mg versus placebo) of 3878 and 2838 patients with type 2 diabetes mellitus, respectively. Model calibration was adequate in both external data sets, discrimination was moderate (ALLHAT-LLT: c-statistics, 0.64 [95% confidence interval, 0.61-0.67] and CARDS: 0.68 [95% confidence interval, 0.64-0.72]).
Conclusions: ARRs of major cardiovascular events by statin therapy can be accurately estimated for individual patients with type 2 diabetes mellitus using a model based on routinely available patient characteristics. There is a wide distribution in ARR that may complement informed decision making.
Clinical Trial Registration: URL: http://www.clinicaltrials.gov. Unique identifier: NCT00327418 (CARDS) and NCT00000542 (ALLHAT).