OBJECTIVES : Overall survival (OS) is considered the gold-standard primary endpoint in oncology clinical trials. Protective OS treatment effects demonstrate clear patient benefits; however, large sample sizes and extensive follow-up are required, potentially delaying access to novel therapeutics. The use of intermediate endpoints such as objective response rate (ORR) as surrogates for OS offers a solution to these challenges but requires validation. A variety of techniques are available to conduct surrogate endpoint validation. We sought to compare the approaches used to assess the surrogacy of ORR for OS in oncology clinical trials.

METHODS : The PubMed and Cochrane Systematic Reviews databases were queried from 2010 to 2020 to identify studies that assessed ORR as a surrogate for OS. Analytical strategies were reviewed and extracted to compare OS endpoints (median, hazard ratio [HR]), ORR endpoints (percentage, odds ratio [OR], relative risk [RR]), correlation types, regression models, surrogate threshold effects (STE), and controlling for patient crossover.

RESULTS : The review identified 24 meta-analysis studies in 12 solid tumor and hematologic oncology populations. Median OS and OS HR were the primary endpoints in 9/24 (37.5%) and 17/24 (70.8%), respectively. Choice of ORR measure varied among studies; absolute ORR was used in 9/24 (37.5%), ORR OR in 15/24 (62.5%), and ORR RR in 6/24 (25.0%), with several studies employing multiple analyses. Furthermore, these measures were logarithmically transformed in 11/24 (45.8%) studies. Correlations were estimated in 15/24 (62.5%), and associations were modeled using regression techniques in 23/24 (95.8%). Few studies estimated the STE (20.8%) or attempted to control for patient crossover (16.7%).

CONCLUSIONS : Methodologies to validate ORR as a surrogate endpoint for OS varied between studies. Further research is required to understand the impact of analytical methodology on the validation of surrogate endpoints, which may inform best practices for future studies.