Revised HLA-DP TCE-Core Permissiveness Model Better Defines Relapse Risk and Survival Following Haploidentical Transplant

The presence of an HLA-DPB1 nonpermissive mismatch (NPMM) by the TCE-3 model has been associated with improved survival following haploidentical donor transplantation (HIDT) using post-transplantation cyclophosphamide (PTCy). With the development of a revised model (TCE-Core) that further separates TCE-3 "group 3" alleles into "core" (C) and "noncore" (NC) alleles, a formerly permissive mismatch (PMM) resulting from group 3 alleles in both donor and recipient is now considered a C-NPMM if 1 or more of those alleles is NC. We aimed to study the additional effect of HLA-DPB1 C-NPMM according to the TCE-Core algorithm, as well as the directional vector of the mismatch, on outcomes following HIDT. To this end, we analyzed 242 consecutive HIDT recipients with acute leukemia or myelodysplastic syndrome who underwent transplantation between 2005 and 2021 (median age, 51 years; range, 19 to 80 years). The median follow-up was 62 months (range, 23 to 199 months). Of the 136 HIDTs classified as PMM by TCE-3, 73 were reclassified as a C-NPMM by the TCE-Core algorithm, of which 36 were in the graft-versus host (GVH) vector (37 were host-versus-graft [HVG] only). Given comparable survival between conventional NPMM and C-NPMM, GVH/bidirectional were analyzed together (nonpermissive). HVG-only C-NPMM were combined with HLA-DPB1-matched and PMM (permissive) because of similar outcomes. The presence of a TCE-Core-defined nonpermissive HLA-DP mismatch resulted in superior 5-year overall survival (OS) (66% versus 47%) and disease-free survival (DFS) (60% versus 43%). Compared to the conventional TCE-3 algorithm, TCE-Core identified a higher percentage of nonpermissive transplants (38% versus 23%) and better discriminated outcomes between nonpermissive and permissive status, with a larger difference in survival outcomes using TCE-Core compared to TCE-3 (OS D, 18.3% versus 12.7%; DFS D, 16.5% versus 8.5%). In multivariable analysis (MVA), a nonpermissive TCE-Core mismatch led to improved OS (hazard ratio [HR], .54; P = .003) and DFS (HR, .62; P = .013), largely due to decreased