Analysis of the probability and risk of cause-specific failure.

Abstract

Kaplan-Meier curves are frequently misused in the analysis of nonsurvival endpoints, such as time to local failure or time to late complications. More appropriate analyses are available and described. Cumulative incidence is an unbiased estimate of probability of cause-specific failure. Cumulative conditional probability of cause-specific failure reflects risk to patients remaining at risk. Hazard rates also measure risk. Kaplan-Meier curves overestimate the probability of late complications when there is a high mortality rate. Cumulative incidence and cumulative conditional probability accurately give the probability and risk of cause-specific failure. Kaplan-Meier analysis of cause-specific failure should be avoided because of its misinterpretation as an estimate of probability, in favor of appropriate methods.