Comparison of Search Prognosis Algorithms Demonstrates the Value of HapLogicTM-Based Assessment in Predicting the Likelihood of Identifying 8/8 HLA-Allele Matched Unrelated Donors (URD) & Receiving an 8/8 URD Transplant

Abstract

<h3>Background</h3> Using preliminary search National Marrow Donor Program (NMDP) HapLogic™ predictions, we have developed a search prognosis algorithm that accurately predicts the likelihood of patients having an 8/8 HLA-matched URD (<i>Davis et al, BBMT 2018</i>). An alternative approach is to categorize search prognosis by patient HLA genotype frequency & ancestry (<i>Wadsworth et al, BMT 2016</i>). We compared these search prognosis assessment methods. <h3>Methods</h3> We performed a preliminary URD search on 830 potential adult allograft recipients & categorized the patients into 3 prognosis groups (Good, Fair, Poor) using HapLogic™-based & recipient genotype-based definitions. The HapLogic™-based prognosis groups of Very Good/ Good were classified as "Good", the Fair/ Poor as "Fair", & the Very Poor /Futile as "Poor". Search outcomes (8/8 URD(s) identified & of those transplanted the percentage receiving an 8/8 URD) by prognosis category for each method were then compared. <h3>Results</h3> The 2 methods resulted in different distributions of patients in the Good, Fair & Poor categories (p < 0.001, Figure 1). A similar percentage of patients were classified as having Good searches (HapLogic™ 321/830, 39% vs recipient genotype 357/830, 43%). However, the HapLogic™ method classified a higher number of patients as having Poor searches (HapLogic™ 298/830, 36% vs genotype 185/830, 22%) resulting in a lower percentage of Fair patients according to the HapLogic™ method. When the percentages of 8/8 URDs identified for the Good prognosis patients by each method were compared a very high percentage by each method had 1 or more 8/8 URDs identified. Similarly, a very low percentage of Poor patients by either method had 8/8 URD(s) identified. However, the success of 8/8 URD identification was lower in the Fair search prognosis by the recipient genotype method (Figure 2). This was attributable to patients with Very Poor/ Futile searches according to HapLogic™ being classified as Fair by the genotype classification, and thus the genotype-based prognosis over-estimating the chance of search success. Of the 445 patients transplanted in the study period with either a 7-8/8 URD or other alternative donor, the distribution of the patients into Good/ Fair/ Poor groups was different with more patients being classified as Fair by the recipient genotype method. Moreover, the percentage receiving an 8/8 URD was different with a lower number of Fair patients by recipient genotype receiving an 8/8 URD (p < 0.001, Figure 3). <h3>Conclusions</h3> Search prognosis using HapLogic™ predictions obtained from the preliminary search allows better identification of patients who are very unlikely to have an 8/8 URD. Search prognosis should incorporate preliminary search HapLogic^TM predictions to improve search accuracy & efficiency. This is especially important for minorities who are less likely to have 8/8 URDs.