HLA disease association studies aim to find protective/predisposing alleles by testing differences in haplotype frequencies across a population of patients and controls. Payami’s relative predispositional effects (RPE) is popular for analyzing disease associations because it corrects for multiple testing. The RPE method is difficult to apply to HLA donor registries because it cannot handle mixed-resolution data and does not adjust for demographic differences between patients and controls. We aimed to integrate both Rubin’s multiple imputation methods and a multivariate regression framework into the RPE method to create a robust framework for the analysis of ambiguous typing data. As an alternative to the X[2] test, we implemented an F test that is based on the difference between observed patient allele frequencies across m imputations and the expected patient allele frequencies after adjustment for demographics. We into account both within and between variance when determining allele significance. We sequentially remove the most significant allele until the F-score is insignificant. We applied our modified RPE method to an HLA association study at A, C, B, DRB1 loci of 2,887 patients with Multiple Myeloma and 50,000 controls. The summary results of RPE on the DRB1 locus are here:[table1]The method completed 3 iterations before the F test was no longer significant. The DRB1 alleles listed have significant differences in proportions between patients and controls. Our method for integration of multiple imputation with RPE can be applied in association studies where HLA typing ambiguity is present. Development of this method is needed to fully utilize all available information from registry databases to uncover HLA associations with diseases. We hope our associations can refine laboratory work investigating HLA protein function and cancer.