Predicting Real-World Effectiveness of Cancer Therapies Using Overall Survival and Progression-Free Survival from Clinical Trials: Empirical Evidence for the ASCO Value Framework

Objectives To measure the relationship between randomized controlled trial (RCT) efficacy and real-world effectiveness for oncology treatments as well as how this relationship varies depending on an RCT's use of surrogate versus overall survival (OS) endpoints. Methods We abstracted treatment efficacy measures from 21 phase III RCTs reporting OS and either progression-free survival or time to progression endpoints in breast, colorectal, lung, ovarian, and pancreatic cancers. For these treatments, we estimated real-world OS as the mortality hazard ratio (RW MHR) among patients meeting RCT inclusion criteria in Surveillance and Epidemiology End Results-Medicare data. The primary outcome variable was real-world OS observed in the Surveillance and Epidemiology End Results-Medicare data. We used a Cox proportional hazard regression model to calibrate the differences between RW MHR and the hazard ratios on the basis of RCTs using either OS (RCT MHR) or progression-free survival/time to progression surrogate (RCT surrogate hazard ratio [SHR]) endpoints. Results Treatment arm therapies reduced mortality in RCTs relative to controls (average RCT MHR = 0.85; range 0.56–1.10) and lowered progression (average RCT SHR = 0.73; range 0.43–1.03). Among real-world patients who used either the treatment or the control arm regimens evaluated in the relevant RCT, RW MHRs were 0.6% (95% confidence interval −3.5% to 4.8%) higher than RCT MHRs, and RW MHRs were 15.7% (95% confidence interval 11.0% to 20.5%) higher than RCT SHRs. Conclusions Real-world OS treatment benefits were similar to those observed in RCTs based on OS endpoints, but were 16% less than RCT efficacy estimates based on surrogate endpoints. These results, however, varied by tumor and line of therapy.