Progression-free survival estimates are shaped by specific censoring rules: Implications for PFS as an endpoint in cancer randomized trials

Progression free survival is a frequently used primary endpoint of randomized clinical trials in cancer. Kaplan Meier method is broadly used to analyze progression free survival data. In Kaplan Meier method only patients with the event of interest contribute to the analysis. Patients that do not contribute to dataset are usually censored. Kaplan-Meier method is based on the assumption that the censored patients are not different than those who contribute to the dataset. Breaking this central assumption can lead to major imbalance in survival analysis. Sympatico trial compared venetoclax combined with ibrutinib against placebo and ibrutinib in patients with relapsed/refractory mantle cell lymphoma and showed an increased progression free survival in the venetoclax and ibrutinib arm. Applying another set of censoring rules improved the primary endpoint only in the experimental arm. Here, we present based on these results, how imbalances in censoring rates can lead to major overestimation of progression free survival results favouring the experimental arm. Furthermore, we draw relevant conclusions for investigators regarding the volatility of progression free survival with respect to censoring rules. Finally, we propose several solutions to make the progression free survival a more reliable endpoint in randomized clinical trial.