Leapfrogging Data

Abstract

Properly designed and conducted randomized controlled clinical trials (RCTs) are the premier tool for both testing mechanistic hypotheses and critically ascertaining the risks and benefits of a therapy or strategy for clinical care. The sample size of a trial is mainly a function of the rates of its primary objectives and the presumed influence of the intervention. Trials focusing on a primary outcome variable that can be readily quantified in each subject, such as blood pressure or plasma cholesterol levels, require substantially fewer participants and shorter durations to determine whether their predefined measurement is altered compared with a morbidity and mortality trial. Trials designed to determine whether clinical prognosis is altered by an intervention depend on the proportion of patients experiencing the predefined adverse clinical event(s) and often require 100s-fold–greater patient-time exposures to test their primary hypothesis and provide even modest information about safety. These resource-intense morbidity and mortality trials are generally only performed when information from observational studies as well as smaller mechanistic and surrogate- outcomes RCTs are so highly supportive of a favorable outcome that they justify the effort. Despite this understandable stacking of the cards with the best available information, many of the morbidity and mortality trials conducted to test for a potential favorable impact of an intervention conclude by not supporting the prestudy hypothesis-generating data.1 The lessons in humility offered by these neutral or negative outcomes trials underscore the importance of obtaining crucial risk–benefit data before widespread adoption of even an apparently favorable therapy.2 Articles pp 2506 and 2515 For rational therapeutic decision making, we would ideally like to have both a framework of reliable mechanistic information and robust clinical outcomes and safety data. Sometimes major clinical outcomes trials are designed with a complement of embedded ancillary trials to generate a more complete picture …