Abstract 11552: Statin Therapy in Patients With Type 2 Diabetes: Development and Validation of a Model to Predict Individualized Treatment Effects Based on the ASCOT, ALLHAT and CARDS Trials

Abstract

Introduction: Clinicians need to translate results from randomized clinical trials to individual patients with type 2 diabetes mellitus (T2DM). In this study we aim to develop and validate a model for the prediction of absolute risk reductions (ARR) for major cardiovascular events by usual-dose statin therapy for individuals with T2DM. Methods: Data from the ASCOT-LLA trial (atorvastatin 10 mg versus placebo) of 2,725 T2DM patients were used to predict 10-year individualized ARR on major cardiovascular events (myocardial infarction, stroke or cardiovascular death). Predictions were based on a newly developed model of 8 predictors: age, sex, current smoking, systolic blood pressure, non-HDL cholesterol, fasting glucose, history of cardiovascular events and treatment allocation (statin or placebo). External validation was performed in ALLHAT-LLT (3,878 T2DM patients) and CARDS (2,838 T2DM patients). Possible effect modification of statin therapy was evaluated for baseline cardiovascular risk and single interactions between predictors and treatment allocation. Results: The predicted 10-year ARR for cardiovascular events was low (<2%) for 13% of the T2DM patients, translating in 10-year NNT >50 (Figure 1). About 30% of the T2DM patients had an ARR of >4% (10-year NNT <25). The addition of treatment interactions did not significantly improve model performance. The developed model showed adequate calibration and moderate discrimination in both external validation sets (concordance statistic 0.64 in ALLHAT and 0.68 in CARDS). Conclusion: Absolute risk reductions by statin therapy can be estimated for individual patients with T2DM using a model based on trial data. There is a wide distribution in absolute risk reduction by statin therapy in T2DM patients due to a distribution in cardiovascular risk. Individualized treatment effects are more informative for patients and clinicians than relative average effects and can be used to guide clinical decision-making.