Impact of IgG monitoring and machine learning based prediction on outcomes of ABO incompatible kidney transplantation in blood type O recipients

Abstract

ABO-incompatible kidney transplantation (ABO-i KT) facilitates transplantation across blood types; however, antibody-mediated rejection (ABMR) remains a major concern, particularly in blood type O recipients. This retrospective study evaluated the effect of immunoglobulin G (IgG) monitoring and machine learning (ML)-based IgG prediction on post-transplant outcomes in 408 ABO-i KT recipients treated between 2014 and 2020. In blood type O recipients, the introduction of IgG monitoring (Era 2) was associated with a significantly lower incidence of ABMR (P = 0.041) and acute rejection (P = 0.037) compared with Immunoglobulin M (IgM)-only monitoring (Era 1). A higher initial IgM titer was identified as a risk factor for ABMR. To address the absence of IgG data in the IgM-only cohort, an ML model was developed using 610 cases to predict pre-transplant IgG titers based on IgM levels, number of plasmapheresis sessions, and ABO blood type. The model demonstrated good predictive performance (mean absolute error [MAE] = 0.593, R2 = 0.721) and indicated that 12.2% of type O recipients in the IgM-only era were estimated to have high IgG titers (>= 1:64). These findings support the clinical utility of IgG monitoring and ML-based estimation to enhance immunologic risk stratification and optimize preconditioning strategies in ABO-i KT.