342 Background: Cisplatin-based NAC in UC confers a survival benefit, but prediction of which patients (pts) will benefit is not possible. We attempted to validate pharmacogenomic (PGx) markers of cisplatin sensitivity derived from a cell-based model and previously associated with response in a population of pts receiving NAC. Methods: Three germline single nucleotide polymorphisms (SNPs) were previously associated with pathologic response at cystectomy in a single-institution discovery set of 59 cT2 UC pts. These SNPs were tested for association with NAC response in a prospectively-identified, multi-institution, independent cohort. The primary analysis tested for association between rs244898 and rs7937567 and pathologic complete response (pT0). A replication set of 134 pts would provide 80% power to detect effects of both SNPs at p=0.05 independently using a multivariate logistic model. Results: N=146 pts with ≥cT2 N0 disease were identified from three institutions. All pts received ≥3 cycles of cisplatin-based NAC and underwent definitive surgery. Rates of pT0 and <pT2 were 26% and 50%, respectively, comparable to the discovery population. Though each SNP had a >5 odds ratio of effect on pT0 in the discovery set, neither was associated with achievement of pT0 in the replication set. The third SNP (rs10964552) which was associated with pathologic downstaging in the discovery set, also failed to replicate. All three SNPs were in Hardy-Weinberg equilibrium, and minor allele frequencies were similar between discovery and validation. Conclusions: Three germline SNPs previously associated with platinum sensitivity in pre-clinical and clinical models were not associated with NAC response in a large replication cohort of UC pts. Reasons for failure may include unmeasured clinical differences between the pt cohorts, a higher proportion of MVAC use in the replication set compared to gemcitabine/cisplatin in discovery pts, or simply spurious association in the discovery cohort. These results emphasize the need for replication when evaluating PGx markers, and demonstrate that multi-institutional efforts are feasible and will likely be necessary to achieve advances in UC PGx.