A Joint Model for Binary Longitudinal and Right-Censored Survival Data in the Context of Antibody Mediated Rejection After Kidney Transplantation

• Activity Number: 612 - Statistics for Clinical Trials and Medical Research
• Type: Contributed
• Date/Time: Thursday, August 1, 2019 : 8:30 AM to 10:20 AM
• Sponsor: ENAR
• Author(s): Maarten Coemans* and Aleksandar Senev and Marie-Paule Emonds and Maarten Naesens and Geert Verbeke
• Companies: KU Leuven and KU Leuven and KU Leuven and KU Leuven and Catholic University of Leuven
• Keywords: joint modeling; survival analysis; mixed models; kidney transplantation; biopsy data

Abstract

Antibody mediated rejection (ABMR) is the main cause of late kidney graft failure after transplantation and leads to histological damage (ABMRh). We analyzed the effect of having pre-transplant donor-specific antibodies (DSA) on the post-transplant occurrence of ABMRh and evaluated mediation, via ABMRh, of its effect on graft survival. We performed a single-center cohort study, including 985 kidney transplantations (2004-2013) and 4317 biopsies, obtained up to 5 years post-transplant. The effect of pre-transplant DSA on the occurrence of ABMRh and on graft survival was studied using mixed and survival models respectively. Mediation was analyzed using a joint model where a survival model is combined with a logistic mixed model. The mixed model includes a cubic B-spline for post-transplant time, random intercepts and random slopes. The survival model assumes proportional hazards and a flexible baseline hazard. We concluded that, within the first 5 years after transplantation, patients with pre-transplant DSA are more prone to develop ABMRh in the transplanted kidney. In fact, the negative effect of pre-transplant DSA on graft survival is fully mediated by the occurrence of ABMRh.