ACCORD(ing) to a Trialist

Abstract

“ Medicine is a science of uncertainty and an art of probability .”1 —William Osler (1849 to 1919) The practice of medicine will always remain an art in which important individual patient decisions are at best guided by incomplete evidence. Randomized controlled clinical trials (RCTs), although imperfect, are the premier tools to generate reliable data to assist in decision making.2 In cardiovascular medicine, we have been fortunate to have a rich legacy of definitive RCTs demonstrating survival benefits of therapies that have deservedly become cornerstone components of clinical practice. However, RCTs with mortality as the primary objective are practical only for evaluating populations with exceedingly high anticipated mortality rates or would require prohibitively long follow-up to obtain a reliable answer. Construction of composite outcomes combining nonfatal clinically important events with deaths into a single end point offers an effective statistical compromise whereby the potential benefits and risks of an intervention can be assessed with relatively smaller sample sizes than in mortality RCTs.3,4 Using composite outcomes that have clinically indisputable and important components extends the application of clinical outcome trials to more prevalent populations with less severely impaired prognosis. However, the compromise is that use of composite end points introduces uncertainty into the interpretation from possible inconsistencies among the components on effect size and at times even direction.5 An even greater end-point compromise is more frequently made in RCTs by targeting a primary objective of determining whether a readily quantifiable laboratory measure considered a surrogate marker of disease progression can be favorably altered by the proposed intervention. Because all patients can generally provide a measure of the surrogate or biomarker, these mechanistically targeted trials require considerably fewer patients and shorter durations.6 A strong link between the surrogates with observational associations of …