Trajectory-based weight modeling distinguishes pathologic weight loss and predicts adverse events in nonhuman primate allogeneic kidney transplantation

Abstract

Weight change predicts mortality and graft loss in human kidney transplant recipients, yet posttransplant weight dynamics in nonhuman primate transplant models remain poorly characterized despite their importance in preclinical immunosuppression and tolerance studies. By failing to account for demographic characteristic-specific growth patterns, static weight thresholds reduce interpretability and mask early biological signals of posttransplant morbidity. We analyzed 879 weights from 40 nonhuman primates undergoing allogeneic kidney transplantation, with follow-up censored at 1-year posttransplant. Expected posttransplant trajectories were estimated using demographic-stratified linear mixed-effects models with natural splines. Deviations from predicted trajectories defined pathologic weight loss. Associations with adverse events (AEs) were evaluated using time-dependent cause-specific Cox regression. Expected trajectories differed significantly by age and species, with greater anticipated weight gain among younger animals. Each 1-standard deviation decrease below predicted weight was associated with a 2-fold increased hazard of AEs. The model predicted pathologic deviations before symptom onset and humane endpoints. Longitudinal weight trajectory modeling represents a computational modeling-based new approach methodology that improves interpretability of preclinical weight endpoints. Defining risk relative to expected trends enables earlier and more precise prediction of AEs, supports adaptive intervention, including immunosuppression de-escalation and infection-directed management.

Weight change predicts mortality and graft loss in human kidney transplant recipients, yet posttransplant weight dynamics in nonhuman primate transplant models remain poorly characterized despite their importance in preclinical immunosuppression and tolerance studies. By failing to account for demographic characteristic-specific growth patterns, static weight thresholds reduce interpretability and mask early biological signals of posttransplant morbidity. We analyzed 879 weights from 40 nonhuman primates undergoing allogeneic kidney transplantation, with follow-up censored at 1-year posttransplant. Expected posttransplant trajectories were estimated using demographic-stratified linear mixed-effects models with natural splines. Deviations from predicted trajectories defined pathologic weight loss. Associations with adverse events (AEs) were evaluated using time-dependent cause-specific Cox regression. Expected trajectories differed significantly by age and species, with greater anticipated weight gain among younger animals. Each 1-standard deviation decrease below predicted weight was associated with a 2-fold increased hazard of AEs. The model predicted pathologic deviations before symptom onset and humane endpoints. Longitudinal weight trajectory modeling represents a computational modeling-based new approach methodology that improves interpretability of preclinical weight endpoints. Defining risk relative to expected trends enables earlier and more precise prediction of AEs, supports adaptive intervention, including immunosuppression de-escalation and infection-directed management.