Fig. 6: Parameter values for Weibull fits to overall survival (OS) data scored by trial outcome.

a Differences in Weibull α and β parameters for experimental and control arm OS data (121 comparisons from 116 trial figures). For α, the value for the control arm was subtracted from the value for the experimental arm. Differences in β were computed by determining the percent change in β in the experimental arm with respect to the control arm (positive values indicate larger β in the experimental arm). Asterisks denote trials that tested immune checkpoint inhibitors (ICIs). Success in all cases was judged based on the original report and most often corresponded to a hazard ratio less than one at 95% confidence based on Cox proportional hazards regression. Published hazard ratios and associated confidence intervals (red lines) and mean Weibull ratios of cumulative hazards (blue lines) with corresponding 95% confidence intervals, along with the reconstructed survival curves, of four clinical trials in the data set: b NCT01377376 (tivantinib plus erlotinib vs. erlotinib), c NCT01607957 (TAS-102 vs. placebo), d NCT01673867 (nivolumab vs. docetaxel), e NCT00861614 (ipilimumab vs. placebo after radiotherapy).