Impact of Informative Censoring on the Kaplan-Meier Estimate of Progression-Free Survival in Phase II Clinical Trials

Abstract

Informative censoring in a progression-free survival (PFS) analysis arises when patients are censored for initiation of an effective anticancer treatment before the protocol-defined progression, and these patients are at a different risk for treatment failure than those who continue on therapy. This may cause bias in the estimated PFS when using the Kaplan-Meier method for analysis. Although there are several articles that discuss this issue from a theoretical perspective or in randomized phase III studies, there are little data to demonstrate the magnitude of the bias on the estimated quantities from a phase II trial. This article describes the issues by using two oncology phase II trials as examples, evaluates the impact of the bias using simulations, and provides recommendations. The two trials were selected because they demonstrate two different reasons for censoring. Simulations show that the magnitude of the bias depends primarily on the proportion of patients who are informatively censored and secondarily on the hazard ratio between the group of patients who remain on study and the group of patients who are informatively censored. Recommendations include using an alternative end point, which includes inadequate response and initial signs of clinical progression as treatment failure, and a competing risk analysis for studies in which competing events preclude or modify the probability of observing the primary event of interest. If informative censoring cannot be avoided, then all patients should be observed until progression, and sensitivity analyses should be used as appropriate.

An evaluation of the impact of missing deaths on overall survival analyses of advanced non–small cell lung cancer patients conducted in an electronic health records database

“…Mortality serves as the primary outcome in many analyses across disease areas and particularly for oncology. However, it is often incomplete because of imperfect data collection systems, workflows not designed to capture mortality data, and patients lost to follow up . The purpose of this study was to examine the potential impact of missing death data in an EHR‐derived oncology data source, which is of critical importance to establishing a research‐grade EHR‐derived database and should provide guidance with respect to an acceptable level of completeness …”

Hard-Wired Bias

“…These patients are likely to be different from those who tolerate therapy well. In a recent study, Campigotto and Weller provide two examples in which patients who are censored are likely to have better or worse survival than those who remain on study 28 . The authors then provide a range of estimates for the outcome had these patients not been excluded, based upon simulation.…”

The role of censoring on progression free survival: Oncologist discretion advised

“…Time to treatment failure includes discontinuation for toxicity, and would minimise the role unequal discontinuation may play in PFS estimates. A second solution would be to use formal competing risk analysis, as has been attempted by other groups [8]. However, we believe a simple journal standard would go a long way to elucidating the extent of the problem.…”