Competing Risk Models Should Be Considered When Estimating Kidney Allograft Failure

Journal of the American Society of Nephrology, vol. 36, pp. 747–748

Abstract

We read with interest the article “Competing and Noncompeting Risk Models for Predicting Kidney Allograft Failure” published in JASN.1 Kaplan–Meier (KM) analysis is commonly applied in time-to-event analyses. Because competing events are censored in KM models, conclusions drawn on the basis of these models can be biased in the presence of a competing event.2–4 In the context of kidney transplantation, death of a recipient with a functioning graft is a common competing event (risk) for graft loss, as a consequence KM-based evaluations may overestimate the actual incidence of graft loss. In their recent paper, Truchot et al. claimed to “carefully assess the performance of competing risk and noncompeting risk models in the context of kidney transplantation” and concluded that “competing and noncompeting risk models performed similarly in predicting long-term kidney graft failure.”1 While the involvement of 27 coauthors from 21 institutions and the reliance of multiple data resources1 creates an impression of authority, we would like to point out that the paper's firm conclusions are based on a follow-up that was confined to <8 years. As a consequence, conclusions by Truchot et al. are based on a dataset with approximately 9% competing events,1 and the paper's conclusions may not extend to conditions with longer follow-ups, higher-risk populations, or when also including events occurring in the first months after transplantation (excluded from the evaluation1). To illustrate the effect of progressive proportions of competing risk on KM-based graft loss estimates, we estimated the incidences of graft loss for different follow-up times (i.e., at increasing degrees of competing risks) for all primary kidney transplantations performed in The Netherlands between 2000 and 2022 through both KM and Aalen–Johansen (competing risk) analyses (Table 1). The data illustrate the progressive effect of competing risks on graft loss estimates and show that the advice to apply competing risk models also applies to the field of kidney transplantation. Failure to do so may lead to inappropriate conclusions when evaluating longer follow-up times, as well as for higher-risk programs in, for example, older recipients, programs with grafts from extended criteria donors (that are preferably transplanted in older recipients), or for recipients with a higher mortality risk (e.g., recipients with type 2 diabetes). Table 1 - Cumulative incidences of graft loss at different time points after transplantation estimated through Kaplan–Meier and Aalen–Johansen (competing risk) models for all primary living donor and deceased donor procedures performed in The Netherlands between 2000 and 2022 (n=14,529) Donor Type Follow-Up Period 5 yr 10 yr 15 yr 20 yr Living % Competing deaths 6.6 12.7 15.6 16.3 KM event probability 7.3 13.0 19.6 25.8 A-J incidence (%) 7.1 12.0 16.9 20.9 Absolute overestimation (%) 0.2 1.0 2.7 4.9 Relative overestimation (%) 2.8 8.3 16.0 23.4 Deceased % Competing deaths 9.7 17.9 21.9 23.0 KM event probability 7.1 14.0 20.7 26.2 A-J incidence (%) 6.7 12.4 16.8 19.8 Absolute overestimation (%) 0.4 1.6 3.9 6.4 Relative overestimation (%) 6.0 12.9 23.2 32.3 A-J, Aalen–Johansen; KM, Kaplan–Meier.

Authors 4

1. Jan H.N. Lindeman corresponding Aachen LUMC Transplant Center Department of Surgery, Leiden University Medical Center

   Leiden University Medical Center

   Affiliation as printed

   LUMC Transplant Center, Department of Surgery, Leiden University Medical Center, Leiden, The Netherlands

2. Hein Putter Aachen Section Medical Statistics Department of Biomedical Data Sciences

   Leiden University Medical Center

   Affiliation as printed

   Section Medical Statistics, Department of Biomedical Data Sciences, Leiden University Medical Center, Leiden, The Netherlands

3. Ian P.J. Alwayn Aachen LUMC Transplant Center Department of Surgery, Leiden University Medical Center

   Leiden University Medical Center

   Affiliation as printed

   LUMC Transplant Center, Department of Surgery, Leiden University Medical Center, Leiden, The Netherlands

4. Esther Bastiaannet

   University of Zurich

   Affiliation as printed

   Epidemiology, Biostatistics and Prevention Institute, University of Zurich, Zurich, Switzerland