Introduction: We recently demonstrated that cognitive impairment (CI) is prevalent and persistent in HM patients treated with BMT (Sharafeldin et al., J Clin Oncol, 2018). While the impact of genetic factors on cognition in the non-oncology realm is well-recognized, limited information exists in HM patients treated with BMT. We tested the hypothesis that genetic variants in blood-brain barrier (BBB), drug transporter, telomere homeostasis, neurotransmission, neural repair and DNA repair are associated with CI in HM patients treated with BMT. We previously presented preliminary results of the Discovery cohort (Sharafeldin et al., Blood, 2016); here we present additional analyses (at the gene level) on the Discovery cohort, and replication of salient findings in an independent Validation cohort.Methods: The Discovery cohort consisted of prospective, longitudinally-assessed patients undergoing autologous or allogeneic BMT for HM at a single center. Cognitive function was assessed objectively using a battery of standardized tests prior to BMT, and at 6mo, 1y, 2y, and 3y post-BMT; germline DNA was collected pre-BMT. Global Deficit Score (GDS) was used as an indicator of CI (GDS ≥0.5 indicated CI). We employed a candidate gene approach, relying on biologically-plausible hypotheses for gene selection. Standard quality control procedures yielded 985 SNPs in 68 genes with a total genotyping call rate of 99% in 277 BMT recipients. Generalized estimating equation models were fitted to examine association of individual SNPs with CI. Linkage disequilibrium-based pruning yielded 326 independent tests, with an overall p-value threshold of 1.53x10-4 using Bonferroni correction for multiple testing. We also conducted gene-level analysis considering each gene as a SNP-set using the machine learning approach of logic regression, with cross-validation, to identify SNP-SNP interactions at the gene-level (Bonferroni p ≤ 7.94x10-4). An independent Validation case-control set from the BMT Survivor Study was used to replicate significant findings from the Discovery cohort, using conditional logistic regression models. Cases (n=141) were BMT survivors who reported: i) presence of memory/learning problems; and ii) severity of memory/learning problems. Controls (n=258) were BMT survivors without cognitive problems, matched (1-2/case) on race/ethnicity, HM, and time from BMT.Results:Discovery: Among the 277 patients in the Discovery cohort, mean age at BMT was 50y, 59% were males, 69% were non-Hispanic whites and 47% received an allogeneic BMT. Male sex (OR=3.2, 95%CI, 1.5-6.7, p=0.002) and lower pre-BMT cognitive reserve (OR=4.6, 95%CI, 2.2-9.6, p <0.0001) were associated with CI. Risk of CI was increased in the younger patients (<50y) if they received TBI (age*TBI interaction OR=3.86, 95%CI, 1.0-15.1, p=0.05). Final genetic models were adjusted for relevant variables with significant independent main effects including BMT type, age at BMT, sex, race/ethnicity, and cognitive reserve. We identified significant associations between CI and BBB, telomere homeostasis and DNA repair SNPs using the single-SNP (Fig 1) and gene-level analyses. Replication: The mean age at BMT was 44y, 54% were males, 88% were non-Hispanic whites and 5% received an allogeneic BMT. Replication models were adjusted for age at BMT, sex, race, and education and income levels. Of the 69 SNPs identified for replication, DNA repair SNPs from the single-SNP analysis rs13006837 on XRCC5 (OR=5.8, 95%CI, 2.7-12.3, p=2.0 x 10-5) and rs293796 on OGG1 gene (OR=11.4, 95%CI, 4.6-28.5, p-value=1.08 x 10-6) were successfully replicated. Five SNPs on DNA repair genes from the gene-level analysis showed individual statistically-significant associations in the Validation cohort: rs238417 on ERCC2 (OR=0.48, 95%CI, 0.25-0.91, p=0.03); rs2526698 on EXO1 (OR=0.57, 95%CI: 0.33-0.98, p=0·04); rs3793906 on MGMT (OR=1.96, 95%CI: 1.2-3.3, p=0.01); rs4725016 on RPA3 (OR=0.20, 95%CI, 0.05-0.92, p=0.04); and rs1479569 on XRCC4 (OR=2.3, 95%CI, 1.3-4.1, p=0.005).Conclusion: Our study provides genetic characterization of adult HM patients at increased risk of cognitive impairment following BMT; associations with DNA repair variants were independently replicated. These findings could have implications towards the personalized management of patients undergoing BMT. DisclosuresForman:Mustang Therapeutics: Other: Licensing Agreement, Patents & Royalties, Research Funding.