Derivation of Human Leukocyte Antigen Molecular Mismatch Risk Thresholds in a Diverse Cohort of Kidney Transplant Recipients

Background: Development of donor specific antibodies (DSA) is a known risk factor for rejection and poor allograft outcomes. HLA molecular mismatch (molMM) algorithms are increasingly being studied to risk stratify kidney transplant recipients (KTR). Studies associating incremental molMM and adverse outcomes generally consider molMM as a continuous variable. In clinical practice, use of a binary threshold is a more practical approach to assess those at high risk for de novo DSA or rejection. We derived such a cutoff and applied to our KTR cohort to compare common molMM algorithms, hypothesizing equivocal prognostic value to agMM using this approach. Methods: KTR with high resolution recipient and donor HLA typing were identified. We fit Cox proportional hazards models using predictors of agMM, HLA-Matchmaker (ver 3), HLA-EMMA, and PIRCHE-II (ver 4) as continuous and binary (90% sensitivity of those with +DSA) variables. Results: Of 431 KTR, 70 developed de novo DSA (16%). Correlation of molMM with agMM is seen in Fig 1. As continuous variables, all were significantly associated with dnDSA. This was consistent when using the binary cutoff, where total agMM had similar hazard to molMM for dnDSA. However, loci-specific moIMM had higher hazard for DSA than agMM (Table 1). Conclusion: In theory, molMM provides a more granular assessment of alloimmune risk. However, this has not been consistently shown in large, diverse cohorts using a clinically relevant threshold. Our results indicate application of moIMM at a sensitive cutoff has similar prognostic value to agMM for development of dnDSA when considering total mismatch, but molMM may be superior at HLA class II loci. This merits further investigation and prior to routine clinical use, we advocate evaluation of these algorithms on a biologic level.