Abstract
BACKGROUND: Kidney marginality criteria consider the impact of donor characteristics on graft failure risk. To evaluate graft quality according to both the donor and recipient, we identified recipient characteristics that may modify the predictive capacity of a donor marginality score. METHODS: From the French DIVAT (Données Informatisées et VAlidées en Transplantation) cohort, we included 8299 patients who received a single deceased donor kidney graft between 2000 and 2022. From the learning sample (n = 5533), we constructed a Cox model with death or return to dialysis as the outcome. The model was both internally (n = 2766) and externally (n = 3178) validated. A recipient-relative marginality score was defined by screening the interactions between recipient characteristics and a donor-only graft marginality score. The scores' performances were compared using time-dependent receiver operating characteristic curves. RESULTS: Five donor characteristics (age, cerebrovascular death, cytomegalovirus serology, and histories of diabetes and hypertension) defined the donor marginality score. This score had a nonlinear interaction with recipient age. Thus, the level of risk related to marginal kidneys depended on recipient's age. Marginal grafts induced a higher graft failure risk for old recipients (hazard ratio [HR], 1.79 [for a 0.5-unit increase in marginality for a 70-y-old recipient]; 95% confidence interval [CI], 1.40-2.15) compared with middle-age recipients (HR, 1.47 [for a 55-y-old recipient]; 95% CI, 1.34-1.70; HR, 1.36 [for a 40-y-old recipient]; 95% CI, 1.14-1.64). The donor score provided good discrimination on the whole sample but not after stratifying by recipient age. However, the recipient-relative score retained acceptable discrimination in recipients older than 60 y. CONCLUSIONS: Our results suggest that the consideration of candidate profiles may improve graft quality assessments.