Effect size estimation in cooperative group- and industry-sponsored phase 3 oncology trials.

Abstract

Overly optimistic estimations of effect sizes may lead to underpowered studies and risk of erroneously dismissing effective treatments. To understand the prevalence and factors contributing to estimation of effect sizes, we evaluated 385 superiority-design phase 3 oncology randomized controlled trials with prespecified and observed hazard ratios in published manuscripts. Of these, 88% were sponsored by industry and 22.6% were sponsored by cooperative groups. Few (10%) studies provided justification of their chosen sample size in the manuscript or available protocol. Overly optimistic estimations of effect sizes were common in cooperative group studies and, by contrast, uncommon in industry-sponsored studies. Moreover, industry-sponsored trials meeting the primary endpoint frequently achieved lower hazard ratios than expected. Together, these data suggest a correlation between study sponsorship, clinical trial power calculations, and subsequent trial outcomes.