PRM233 - THE USE OF THE TWO-STAGE METHOD IN ADJUSTING FOR TREATMENT CROSSOVER. A SIMULATION STUDY.

Abstract

A subsequent therapy that is applied in a clinical trial, but not registered in clinical practice, might hamper how reflective the reported overall survival (OS) is of clinical practice. For this, methods like two-stage approach and rank preserving structural failure time models (RPSFTM) have been developed. However, there might also be situations that the treatment switch is reflective of clinical practice but that the percentage of patients receiving a subsequent therapy differs. The objective is to present a methodology using the two-stage method that adjusts survival for differences in frequency of subsequent therapy between clinical trials and clinical practice. PFS and OS data of a trial with two treatment arms was simulated (an active and a placebo arm). It was assumed that in the trial 45% of patients in the placebo arm switched treatment. However, theoretically in clinical practice 80% of patients would receive active therapy after progression on placebo. We applied the two-stage method, where first the counterfactual survival times were estimated to reflect a scenario where nobody switches to an active subsequent therapy. Then we estimated counterfactual survival times assuming that everybody switched. Subsequently, Weibull models, which had best fit, were fitted on both datasets and a weighted average Weibull, was estimated, reflecting the 80% treatment switch pattern in clinical practice. The unadjusted average OS was 25.5 months. Average OS of patients on placebo, assuming nobody switched was 24 months and assuming everybody switched was 28.4 months. The weighted OS based on the theoretical clinical practice percentage of switchers was 27.5 months. This method can be applied to adjust the OS outcomes of trial such that they reflect treatment switching patterns in real clinical practice. The methodology can be further tested on methods like the RPSFTM and other datasets.