Structural cumulative survival models for estimation of treatment effects accounting for treatment switching in randomized experiments.

This article focuses on addressing treatment switching in randomized controlled trials (RCTs) through a novel instrumental variable (IV) approach under a structural cumulative survival model (SCSM). Treatment switching occurs when patients change assigned treatments during follow-up, potentially biasing causal effect estimates, especially in time-to-event outcomes. The authors develop a recursive estimator leveraging randomization as an IV to consistently estimate the causal effect of nucleoside reverse transcriptase inhibitors (NRTIs) on safety outcomes in the OPTIONS HIV trial, where treatment switching was present. Simulation studies demonstrate the estimator's good finite-sample performance, and application to the OPTIONS trial reveals that adding NRTIs increases the risk of severe or worse signs or symptoms, a safety signal not detected by standard intent-to-treat analyses. The method, implemented in the R package "ivsacim," accommodates unmeasured confounding without requiring artificial censoring and can be extended to observational studies under certain assumptions.