Modelling competing risks in nephrology research: an example in peritoneal dialysis

• Autor(a) principal: Teixeira, L
• Data de Publicação: 2013
• Outros Autores: Rodrigues, A, Carvalho, MJ, Cabrita, A, Mendonca, D
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10216/114822
• topic: Cause-specific hazard model; Competing risks; Cumulative incidence function; Peritoneal dialysis

Background: Modelling competing risks is an essential issue in Nephrology Research. In peritoneal dialysis studies, sometimes inappropriate methods (i.e. Kaplan-Meier method) have been used to estimate probabilities for an event of interest in the presence of competing risks. In this situation a competing risk analysis should be preferable. The objectives of this study are to describe the bias resulting from the application of standard survival analysis to estimate peritonitis-free patient survival and to provide alternative statistical approaches taking competing risks into account. Methods: The sample comprises patients included in a university hospital peritoneal dialysis program between October 1985 and June 2011 (n = 449). Cumulative incidence function and competing risk regression models based on cause-specific and subdistribution hazards were discussed. Results: The probability of occurrence of the first peritonitis is wrongly overestimated using Kaplan-Meier method. The cause-specific hazard model showed that factors associated with shorter time to first peritonitis were age (≥55 years) and previous treatment (haemodialysis). Taking competing risks into account in the subdistribution hazard model, age remained significant while gender (female) but not previous treatment was identified as a factor associated with a higher probability of first peritonitis event. Conclusions: In the presence of competing risks outcomes, Kaplan-Meier estimates are biased as they overestimated the probability of the occurrence of an event of interest. Methods which take competing risks into account provide unbiased estimates of cumulative incidence for each specific outcome experienced by patients. Multivariable regression models such as those based on cause-specific hazard and on subdistribution hazard should be used in this competing risk setting.