Predictive Models for Recurrent Membranous Nephropathy After Kidney Transplantation

Area of Science:

• Nephrology
• Transplantation Immunology
• Biostatistics

Background:

• Recurrent membranous nephropathy (MN) affects a significant portion of kidney transplant recipients (KTRs), leading to substantial allograft loss.
• Human Leukocyte Antigen (HLA) risk alleles are potential predictors, but a comprehensive understanding of predictive factors is lacking.

Purpose of the Study:

• To develop and evaluate predictive models for recurrent MN in KTRs.
• To identify key donor-recipient HLA characteristics associated with MN recurrence.

Main Methods:

• Three prediction models (Group LASSO, penalized Cox regression, random forest) were developed using data from the Australian and New Zealand Dialysis and Transplant registry.
• Models were tuned via tenfold cross-validation in a derivation cohort and validated in a separate cohort.
• Model performance was assessed using the area under the receiver operating characteristic curve (AUC-ROC).

Main Results:

• A penalized Cox regression model demonstrated a high AUC-ROC of 0.91 in the derivation cohort and 0.73 in the validation cohort.
• Recipient HLA-A2, donor HLA-DR12, donor-recipient HLA-B65, and HLA-DR12 match were identified as important predictors across models.
• Random forest model showed lower predictive performance (AUC-ROC 0.62 derivation, not validated).

Conclusions:

• Penalized Cox regression offers a robust approach for predicting recurrent MN post-kidney transplantation.
• Donor-recipient HLA matching is a critical factor in the recurrence of MN.
• Further validation in larger datasets is recommended to confirm these findings.