Statin Use and Incident Diabetes Explained by Bias Rather Than Biology

Abstract

The evidence supporting a link between statin use and incident diabetes is seemingly robust; it has been observed in multiple prospective randomized trials and confirmed by meta-analyses. However, differences in survival among statin vs placebo-treated patients in randomized trials might have caused bias with respect to diabetes surveillance. Bias might have been further exaggerated from the strong association between diabetes and cardiovascular events, which were the primary end points in major statin trials. Meta-analyses of randomized trials have demonstrated a 9% increase in the odds of incident diabetes among patients who receive statins compared with placebo (odds ratio [OR], 1.09; 95% confidence interval [CI], 1.02-1.17) and a 12% increase of high-dose statin therapy vs low doses (OR, 1.12; 95% CI, 1.04-1.22). To simulate the possible effect of bias in these meta-analyses, each pooled trial was modified by assuming that 10% of the patients who experienced primary end points subsequently developed diabetes undetected. Meta-analyses of these simulated trials attenuated the association to a nonsignificant level among placebo-controlled trials (simulated OR, 1.04; 95% CI, 0.98-1.10), and high- vs low-dose trials (simulated OR, 1.07; 95% CI, 1.00-1.15). Our results demonstrate how a small influence of bias in each randomized trial could have contributed substantially to the observed association between statin use and diabetes.