Statistical power for MACE and individual secondary endpoints in cardiovascular outcomes trials for type 2 diabetes: a systematic review

Abstract

Abstract Cardiovascular outcomes trials (CVOTs) with novel drugs to treat type 2 diabetes have uniformly chosen the composite “major adverse cardiovascular events (MACE)” as their primary endpoint, but they also report hazard ratios for individual cardiovascular outcomes (myocardial infarction, stroke, cardiovascular death, all-cause death, hospitalization for heart failure). We wanted to scrutinize the power to identify significant differences with respect to individual as compared to composite outcomes. We estimated post hoc the statistical power to detect significant differences of 10–25% for published studies, comparing the proportions of patients with an event (two-sided log-rank tests). For MACE, the power to detect a 15% difference ranged from 82.3 to 100.0% for larger trials, but was only 69.1 and 50.5 for smaller, preliminary trials (SUSTAIN-6 and PIONEER-6). For individual endpoints, the power, as a rule, was substantially lower. In conclusion, cardiovascular outcomes trials had appropriate power to detect significant reductions in hazard ratios with respect to the primary endpoint, but not for individual cardiovascular outcomes. This was particularly the case for small, preliminary studies. Our results call for caution when comparing results regarding individual endpoints between CVOTs, if the aim is to identify heterogeneity within or between medication classes.