Identifying these three HLA subgroup clusters we sought to estimate the survival for hypothetical patients with these three match grades after adjusting for the important and statistically significant clinical risk factors including patient age, race, Karnofsky performance status, year of transplant, conditioning intensity, and GVHD prophylaxis. In the figures below ( and ), we illustrate one year () and five year () survival for a first transplant, CMV seronegative patient with a 25-year-old marrow donor of match grade shown. The patient examples are assumed to be Caucasian with a Karnofsky performance status of 90-100%, transplanted in 2003-2006 using tacrolimus plus short course methotrexate as GVHD prophylaxis. The example transplant recipients all received myeloablative conditioning intensity except the 55-year-old MDS patients who are assumed to have received a reduced intensity conditioning regimen.
One Year Survival Estimates for Model Patients in Different HLA Matching Groups
Five Year Survival Estimates for Model Patients in Different HLA Matching Groups
As shown, for all six example patients (Panels A-F, and ), the defined matched grade groupings identify clearly distinct survival at one year and five years for all patient cohorts including those with either early or late disease stage, who were either older or younger. HLA-matching groups resulted in approximately 10-11% decrements in 5 year survival between the well-matched, partially matched and mismatched groups for each modeled patient cohort.
Greater differences were recognized within the first year post-transplant suggesting that the impact of HLA-matching has a more profound effect on early post-transplant mortality presumptively due to graft failure, GVHD and infection, than on later post-transplant (5 year outcomes) which may be dominated by both relapse and chronic GVHD and thus possibly less affected by HLA disparity.
Examining the HLA-matching groups in detail () identifies best outcomes associated with no recognized mismatch and informative, though not necessarily high-resolution data at all four loci (HLA-A, B, C and DRB1) or with allele matching at HLA-A, B and DRB1. In group 4, high resolution typing at HLA-A, B and DRB1 overcomes the potentially unrecognized mismatches at the untested HLA-C locus because of strong linkage disequilibrium with HLA-C in high resolution HLA-B matched pairs. Though outcomes overlap with group 5, the lack of any defined mismatch suggests this subgroup should be in the well-matched group. Well-matched transplants therefore, are characterized by informative data at the HLA-A, B and DRB1 loci and no defined mismatch, even if high resolution typing is not available at Class I loci (groups 1-4).
In contrast, the partially matched group (, groups 5-11) includes defined, single locus mismatches at any of the four loci using either high resolution or low/intermediate resolution (allele or antigen) typing. Most surprisingly, the group traditionally called “matched” and referred to in the casual parlance as “MUD” (matched unrelated donor) is group 9. These are matched at low/intermediate resolution for HLA-A and B and high resolution for DRB1 with unknown data at HLA-C. This largest (in the partially matched) subgroup (1,742 patients, 12% of the whole population) has projected one year survival, 9% inferior to the well-matched group. This clearly indicates that the widely used term “6 of 6 matched URD” is both insufficiently precise and a definitive misnomer. In this group, many unrecognized mismatches exist which contradict that this MUD group is “matched” and explaining the poorer outcomes in this group. Similarly, 8 of 8 low resolution matching (group 10), while uncommon, also leads to worsened survival at one year. This emphasizes the importance of examining all four loci and the level of resolution to understand the impact of matching on outcome.
The third, mismatched cohort (, groups 12-21) has greater heterogeneity and includes either mismatch or missing information at two or more loci in all groups. Additionally, either single low resolution mismatch (group 14) or high resolution mismatch (groups 17 and 21) with unknown C typing yields poorer survival as does 6 of 6 low resolution matching with unknown data at HLA-C (group 13).