Background: Delayed graft function (DGF) in pediatric kidney transplantation is a serious complication with a negative impact on patient and graft survival. It is primarily caused by ischemia-reperfusion injury in the immediate post-transplant period. Currently, there are no reliable methods available to assess the risk of DGF in children.
Methods: We performed a retrospective analysis of data from the Cooperative European Paediatric Renal Transplant Initiative (CERTAIN) registry to develop a specific DGF risk assessment model for pediatric kidney transplantation, based on recipient clinical and biochemical parameters associated with ischemia-reperfusion injury and available within the first 24 hours post-transplant. Patients were randomly assigned in a 2:1 ratio to a training and a test cohort.

The model was developed using multiple logistic regression with forward variable selection, supported by multiple imputation for missing data. Internal validation included both complete and imputed datasets. To control for interdependence of model parameters and to provide transparency, adjusted odds ratios were also calculated for each selected parameter in the model. In an extended analysis, we additionally explored whether selected pre-transplant parameters (pre-transplant dialysis mode, donor type, donor sex, recipient age, cold ischemia time, donor hemodynamic instability and number of HLA-DR mismatches) could enhance the model’s predictive performance.
Results: This study included 694 pediatric kidney transplant recipients. The overall DGF rate was 8.5%. The following four key post-transplant parameters were selected for the DGF risk assessment model: (i) occurrence of surgical complications requiring reoperation, (ii) immediate graft urine production, (iii) rate of change in recipient’s serum creatinine, and (iv) initial calcineurin inhibitor therapy. The model showed robust performance, with a ROC-AUC of 0.9043 in the training cohort and 0.878 in the validation cohort without imputed data. The model showed a relatively good performance in terms of calibration. 

In the extended analysis, the inclusion of selected pre-transplant factors did not improve the model’s predictive accuracy. Only the number of HLA-DR mismatches was retained as a parameter in the combined model. However, this parameters contribution to the combined pre- and post-transplant model was marginal and it did not enhance the ROC-AUC (0.878) beyond the performance of the post-transplant model alone.
Conclusions: This multivariable model based on early post-transplant parameters can predict the occurrence of DGF in pediatric kidney transplant recipients with high accuracy. The reliable identification of patients at high risk for DGF may support individualized post-transplant care by aiding clinical decision-making in the immediate postoperative period and it may facilitate future interventional trials targeting ischemia-reperfusion injury in children.