Evidence of survival benefit was often ambiguous in randomized trials of cancer treatments

Abstract

ObjectivesThe objective of the study is to estimate the proportion of statistically significant survival improvements reported in randomized trials of cancer treatments that are also compatible with a clinically negligible benefit. Study Design and SettingThis is a cross-sectional study of reports of randomized clinical trials of cancer treatments that reported a statistically significant increase in overall survival, published in leading journals between 2009 and 2019. The main outcome variable was the hazard ratio (HR) for overall survival and its upper 95% confidence limit. An HR of 0.95 implies an absolute survival gain ≤1.9%, and an HR of 0.90 implies an absolute survival gain ≤3.8%; we reasoned that such survival gains can be considered clinically negligible, given the potential toxicity of oncologic treatments. ResultsAmong 234 trial results, the mean point estimate of the HR was 0.664, and all HRs were below 0.90. The mean upper 95% confidence limit for the HR was 0.897, but 37.6% of the values were ≥0.95, and 59.0% were ≥0.90. These proportions were lower when overall survival was the primary outcome of the trial (29.9% ≥ 0.95 and 51.3% ≥ 0.90). ConclusionsConsidering only point estimates of HRs, all trials reported clinically meaningful improvements in overall survival. However, the upper confidence limits of a large proportion of HRs were also compatible with clinically negligible survival gains. Acknowledging the uncertainty regarding treatment benefits presents a challenge for the reporting of trial results and for clinical decision-making.