Analysis of HLA-A, -B and -DR Alleles as Risk Factors for One-Year Mortality in Heart Transplants Using Artificial Neural Networks

Abstract: The beneficial effects of HLA matching at the HLA-A, -B and -DR loci on recipient short- and long-term survival in heart transplants have been established in a number of studies. The objective of the present study was to evaluate the importance of individual HLA alleles at the HLA-A, -B and -DR loci as risk factors for recipient death within one year after transplant by using artificial neural networks. Records of 24,838 heart transplants from the UNOS databased consisting of 19 risk variables which include the two recipient and donor HLA alleles at the study loci were analyzed by an ensemble of five multilayer perceptrons in a binary classification task. The importance of the classification variables as predictors was measured in a sensitivity analysis and odds ratios were calculated to identify risk factors for early death. The risk factor calculations for two alleles were validated in a Kaplan-Meier survival analysis. The influence of HLA matching on recipient one-year survival was analyzed in a sensitivity analysis and by risk factor calculations. The donors HLA-A 23 and HLA-B 53 alleles were identified as risk factors for early death with odds ratios that were comparable to an increased ischemic time of one hour. The HLA-B 27 allele could be associated with decreased risk of early death. These alleles are important predictors to the network ensemble. In the survival analysis, the difference in the survival probabilities of two groups containing 100 transplants with donor HLA-A 23 and HLA-B 27 alleles each was significant ($P=0.0078$, Log-rank test). At all loci, one mismatch and at the HLA-A and HLA-DR loci two mismatches were associated with a decreased risk of early death relative to no mismatches. At the HLA-B locus, two mismatches were associated with an increased risk of early death The network ensemble was able to identify alleles as risk factors and the result was validated. The analysis on the influence of HLA matching gave results that were in conflict with previous studies.