Postprogression therapy and confounding for the estimated treatment effect on overall survival in phase III oncology trials

Abstract

Estimations of the treatment effect on overall survival (OS) may be influenced by post-progression therapies (PPTs). It is unclear how often OS analyses account for PPT effects. The purpose of this cross-sectional analysis was to determine the prevalence of OS analyses accounting for PPT effects in phase III oncology trials.We screened two-arm, superiority design, phase III, randomised, oncology trials reporting OS from ClinicalTrials.gov. The primary outcome was the frequency of OS analyses adjusting for PPT confounding. Logistic regressions computed ORs for the association between trial-level covariates and the outcome.A total of 334 phase III trials enrolling 265 310 patients were included, with publications between 2004 and 2020. PPTs were reported in 47% of trials (157 of 334), and an analysis accounting for PPTs was performed in only 12% of trials (N=41). PPT adjustments were often prespecified (N=23, 56%), and appeared to be more likely in cross-over studies (OR 5.04, 95% CI 2.42 to 10.38) and studies with discordant surrogate-OS findings (OR 2.26, 95% CI 1.16 to 4.38). In key subgroup analyses, PPT analyses were infrequent, including 8% of trials among those studying locoregional/first-line therapy and 11% of trials among those powered for OS.Although time on PPTs is an important component of OS, PPTs are rarely considered in OS analyses, which may introduce confounding on estimates of the treatment effect on OS. PPTs and methods to account for their effects on OS estimates should be considered at the time of trial design and reporting.