In the past quarter of a century, statins—inhibitors of 3-hydroxy-3-methylglutaryl coenzyme A reductase—have become one of the most studied and prescribed classes of drugs in modern history. The efficacy of these drugs in reducing cardiovascular morbidity and mortality across a wide spectrum of risk is supported by an extensive dataset of randomised controlled trials,1 which have been largely reassuring about the safety of currently available types of statin. However, safety data from trials are inherently incomplete, given the relatively short follow-up periods of clinical trials and their limited external validity. In the linked paper (doi:10.1136/bmj.c2197), Hippisley-Cox and Coupland used routinely collected data on more than two million men and women from 368 general practices in England and Wales to estimate the association of type, dose, and duration of statin use with the occurrence of several end points.2 By using a large database they overcome the problem of the low incidence of statin associated adverse events, and they provide information on adverse effects that clinical trials are not adequately …