Effect of right censoring bias on survival analysis

TL;DR

This paper investigates how right-censoring bias affects Kaplan-Meier survival analysis in oncology, identifying mechanisms that influence survival estimates and proposing indexes to detect bias in datasets.

Contribution

It introduces two bias indexes to detect right-censoring related overestimation of survival and analyzes mechanisms impacting Kaplan-Meier estimates.

Findings

Right-censoring due to loss to follow-up can improve survival estimates.

Bias indexes effectively detect datasets with overestimated survival.

One mechanism of censoring significantly impacts hazard ratio estimates.

Abstract

Kaplan-Meier survival analysis represents the most objective measure of treatment efficacy in oncology, though subjected to potential bias, which is worrisome in an era of precision medicine. Independent of the bias inherent to the design and execution of clinical trials, bias may be the result of patient censoring, or incomplete observation. Unlike disease/progression free survival, overall survival is based on a well defined time point and thus avoids interval censoring, but right-censoring, due to incomplete follow-up, may still be a source of bias. We study three mechanisms of right-censoring and find that one of them, surrogate of patient lost to follow-up, is able to impact Kaplan-Meier survival, improving significantly the estimation of survival in comparison with complete follow-up datasets, as measured by the hazard ratio. We also present two bias indexes able to signal datasets with right-censoring associated overestimation of survival. These bias indexes can detect bias in public available datasets