MO994: Impact of Hla-A, B DR and DQ on renal transplant outcomes in the Tacrolimus/Mmf Era: An Artificial Intelligence Approach

Abstract

Abstract BACKGROUND AND AIMS One of the biological barriers that can increase risk of acute rejection in kidney transplantation is HLA mismatch. Several immunotherapy protocols have been implemented to reduce the effect of HLA-mismatch and improve outcomes. The aim of our study is to assess the effect of HLA mismatches on acute rejection rates in the Tacrolimus era. METHOD All kidney transplant patients registered in UNOS database between 1 January 2005 and 1 December 2019 were retrospectively reviewed. Inclusion criteria: deceased donor transplants that were discharged on Tacrolimus/Mycophenolate Mofetil. Exclusion criteria: multiple organ transplants, previous kidney transplants, recipient age <18 years old, living donor transplants, patients not discharged on Tacrolimus/Mycophenolate Mofetil immunotherapy, missing HLA mismatch or ABO incompatible transplant. We used double-selection lasso (‘least absolute shrinkage and selection operator) logistic regression model to assess for the effect of HLA-A, B, DR and DQ on acute rejection rates at one-year post-transplant. We used square-root lassos for the variables of interest. Variables of interest were HLA-A, B, DR and DQ mismatch. Variables Lasso selected from were: recipient characteristics (age, sex, BMI, ethnicity, diabetes, recipient/donor CMV status, time on dialysis), donor characteristics (KDPI score) and transplant characteristics (type of induction therapy, steroid therapy at time of discharge, cold ischemia time, delayed graft function, PRA). For survival analysis, we fit a penalized Cox model to the entire set of data and obtained the estimated set of alphas. The variables included in the penalized cox model donor, recipient and transplant factors in addition to the HLA mismatches. Then we determined the set of alpha for evaluation using optimized cross-validated grid-search. Moreover, we visualized how the coefficients changed for varying α using ridge regression model. RESULTS About 73 910 were included in our study. Worse acute rejection rates at one-year post-transplant were noted with incremental increase in HLA-DQ (Two HLA-DQ: OR = 1.56, P = 0.005, 95%CI: 1.04–1.27; One HLA DQ: OR = 1.15, P = 0.01, 95%CI: 1.03–1.24), HLA-DR mismatches (two HLA-DR: HR = 1.46, P < 0.01, 95%CI: 1.31–1.62; One HLA-DR: OR = 1.29, P < 0.01, 95%CI: 1.16–1.43), and Two HLA-A (OR = 1.12, P = 0.049, 95%CI: 1.0006–1.26). However, the effect of different HLA-mismatches on death-censored graft survival were minimal (HLA-DQ coefficient: 0.003, HLA-DR coefficient: 0.05, HLA-A coefficient: 0.008 and HLA-B coefficient: 0.007). Mean follow-up time was 3.79 years. Figure 1 shows the visualisation of coefficients among different alphas for the ridge penalised cox regression model. CONCLUSION HLA-DQ, DR and A mismatches still play a vital role in the occurrence of acute rejection in the Tacrolimus/MMF era. However, the effect of HLA mismatches on the graft survival is minimal.