Abstract Background: Paclitaxel-induced neuropathy is a common, severe adverse event that seems to be related to cumulative drug exposure. Examining a neoadjuvantly treated cohort, we previously found increased neuropathy risk in patients carrying the CYP2C8*3 variant, which is associated with altered paclitaxel metabolism. To confirm this association, we examined the univariate association of CYP2C8*3 genotype and paclitaxel-induced neuropathy first in an independent cohort, then as a combined multivariable analysis. Methods: CYP2C8*3 is more common in Caucasians (Allele Frequency AF=0.14) than other racial groups (African-American AF=0.04). To avoid potential issues with population stratification, initial univariate analysis was performed on 209 self-reported Caucasian breast cancer patients from a prospective cohort study (LCCC 9830) who were treated with paclitaxel-based regimens and had not been previously analyzed. The CYP2C8*3 (K399R) variant was genotyped on the Affymetrix DMET™ Plus Chip (Affymetrix, Inc., Santa Clara, CA, USA) at Gentris Corp. (Gentris Corp. Morrisville, NC) from germline DNA collected at diagnosis. The primary endpoint was the dose-at-grade 2+ neuropathy as defined by NCI CTC criteria. Statistical analysis was carried out using the log-rank test across the three genotype groups (*1/*1, *1/*3, *3/*3). The Caucasian cohort was then combined with 78 Caucasian patients from our previous neoadjuvant study and 124 non-Caucasian patients to build a multivariate Cox proportional hazards model. We performed model selection using backward elimination with AIC on a main effects model that included potential covariates: race, age, diabetes, paclitaxel schedule, and supplemental neuropathy therapy. A standard alpha=0.05 was used as the significance threshold for the primary log-rank analysis. Results: The allele frequencies were similar to that expected and the distribution of alleles conformed to Hardy-Weinberg proportions for the Caucasian and non-Caucasian cohorts. 209 Caucasian breast cancer patients treated with paclitaxel were evaluated in the primary analysis, 35 (17%) of whom experienced grade 2+ peripheral neuropathy. The risk of neuropathy was significantly associated with CYP2C8*3 in the primary analysis (log-rank p = 0.006). A combined cohort of 411 patients were evaluable in the Cox model, 76 (18%) of whom experienced grade 2+ neuropathy during treatment. After backward elimination of covariates that did not contribute to the Cox model, increased age (HR = 1.02 95% CI: 1.00–1.04, p = 0.102), non-Caucasian race (HR = 1.76 1.05–2.93, p = 0.031), and CYP2C8*3 (Additive Model: HR=1.98 1.25–3.13, p = 0.004, no model assumed: p = 0.023) were associated with increased risk of paclitaxel-induced neuropathy. Conclusions: We have replicated in an independent population the finding that patients carrying CYP2C8*3 are at increased risk of paclitaxel-induced neuropathy, with risk approximately doubling for each *3 variant carried. After adjusting for CYP2C8 genotype we detected an increase in neuropathy risk for non-Caucasians which is consistent with a previous finding and supports the need to better understand the overall etiology of neuropathy risk. Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr PD10-07.