Peritonitis-free survival in peritoneal dialysis: an update taking competing risks into account

Abstract

Peritonitis-free survival is commonly reported in the peritoneal dialysis (PD) literature. The Kaplan-Meier method appears to be the only technique used to date, although it has known limitations for cohorts with multiple outcomes, as in PD. In the presence of these 'competing risks' outcomes, the Kaplan-Meier estimate is interpretable only under restrictive assumptions. In contrast, methods which take competing risks into account provide unbiased estimates of probabilities of outcomes as actually experienced by patients. We analysed peritonitis-free survival in a cohort of 8711 incident patients from the 'Registre de Dialyse Péritonéale de Langue Française' between 1 January 2000 and 31 December 2007 by calculating the cumulative incidence (CI) of the first episode of peritonitis using the Kaplan-Meier method and a method accounting for competing risks. We compared the CI in different patient groups by the log-rank test and a test developed for competing risk data, Gray's test. After 5 years of PD, the CI of at least one peritonitis episode was 0.4, and the probability of any outcome was 0.96. The Kaplan-Meier method overestimated the CI by a large amount. Compared with the log-rank test, Gray's test led to different conclusions in three out of seven comparisons. The competing risk approach shows that the CI of at least one peritonitis episode was lower than reported by the Kaplan-Meier method but that survival peritonitis-free and still on PD was overall low. The competing risk approach provides estimates which have a clearer interpretation than Kaplan-Meier methods and could be more widely used in PD research.