Background: We evaluated the association of germ-line variation in 10 candidate genes important in prostate cancer biology (NKX3-1, JAK2, SLCO2B1, STAT3, CYP19A1, HSD17B4, TRMT11, PRMT3, HSD17B12, NCOA4) with failure of AA in advanced prostate cancer patients. Methods: Patients (N=619) were enrolled for genotyping at the time of AA failure. Candidate genes were selected from the literature and previous pilot studies in smaller cohorts which had identified variation in TRMT11 and HSD17B12 to be associated with AA failure. Genes were tagged using single nucleotide polymorphisms (SNPs) from HapMap with minor allele frequency of >5% and r²≥0.8. DNA was extracted from peripheral blood and genotyped using Illumina Veracode platform. The primary endpoint was time to failure on AA, defined as time from initiating continuous AA to two serial PSA increases over the nadir or clinical or imaging based progression, whichever came first. Principal component analysis was used for gene-levels tests. Association with the primary endpoint was assessed using proportional hazards regression models at the gene-level and at the SNP level. For SNP level results we estimated per allele (ordinal model) hazard ratios (HR) and 95% confidence intervals (CI) using Cox regression, adjusted for Gleason score (GS). Results: We successfully genotyped 60 SNPs (from 10 genes) in 519 subjects. The median age of the cohort was 71 years (range 43, 92). The GS distribution was 51% subjects with GS≥8; 33% with GS=7 and 17% with GS<7. Median time to AA failure for the cohort was 2.3 years (IQ range: 1.0-4.5). In gene level analyses adjusting for GS, TRMT11 (p=0.004), HSD17B12 (p=0.016), and JAK2 (p=0.044) genes were associated with time to AA failure. Four of the 18 genotyped SNPs in JAK2 were associated with AA failure after adjustment for GS (Table). Conclusions: Variation in the JAK2 gene is significantly associated with time to AA failure. Validation is needed to develop these as predictive biomarkers for AA in advanced prostate cancer.