Correcting treatment effect for treatment switching in randomized oncology trials with a modified iterative parametric estimation method

In randomized oncology trials, patients in the control arm are sometimes permitted to switch to receive experimental drug after disease progression. This is mainly due to ethical reasons or to reduce the patient dropout rate. While progression-free survival is not usually impacted by crossover, the treatment effect on overall survival can be highly confounded. The rank-preserving structural failure time (RPSFT) model and iterative parametric estimation (IPE) are the main randomization-based methods used to adjust for confounding in the analysis of overall survival. While the RPSFT has been extensively studied, the properties of the IPE have not been thoroughly examined and its application is not common. In this manuscript, we clarify the re-censoring algorithm needed for IPE estimation and incorporate it into a method we propose as modified IPE (MIPE). We compared the MIPE and RPSFT via extensive simulations and then walked through the analysis using the modified IPE in a real clinical trial. We provided practical guidance on bootstrap by examining the performance in estimating the variance and confidence interval for the MIPE. Our results indicate that the MIPE method with the proposed re-censoring rule is an attractive alternative to the RPSFT method. Copyright (c) 2016 John Wiley & Sons, Ltd.