Predicting individual risk of chemotherapy-induced severe adverse reaction is a critical issue when selecting the best treatment for cancer patients. SNPs have been identified in genes involved in the pharmacodynamics of fluoropyrimidines, and guidelines even recommend genotyping some DPYD variants in order to estimate the risk of toxicity. However, the predictive value of this approach remains insufficient, thus limiting its clinical implementation. The aim of the present study was to identify new genetic variants by selecting a group of tag SNPs in genes associated with the pharmacodynamics of fluoropyrimidines (CDA, DPYD, ENOSF1, CES1, TYMS, SLC22A7, TYMP, and UMPS). For this purpose, 23 selected SNPs were genotyped on an OpenArray™ platform in a cohort of 301 colorectal cancer patients receiving capecitabine-based chemotherapy. Univariate and multivariate statistical analysis by logistic regression revealed 10 SNPs associated with severe adverse reactions to capecitabine (P < 0.05): rs1048977, rs12726436, and rs2072671 in CDA; rs12119882 in DPYD; rs2853741 in TYMS; rs699517 in TYMS/ENOSF1; rs2270860 and rs4149178 in SLC22A7; and rs2279199 and rs4678145 in UMPS. Except for rs2072671, no association had previously been reported between these SNPs and the risk of capecitabine-induced toxicity. The use of tag SNPs to find new polymorphisms related to adverse reactions to capecitabine was successful. These new variants could increase the predictive power of currently available tests and thus prevent severe adverse reactions to capecitabine.