Retrospective studies have suggested that UDP-glucuronosyltransferase (UGT)1A1, UGT1A7, and UGT1A9 predict severe toxicity and efficacy of irinotecan-containing regimens. We prospectively evaluated the impact of UGT1A genotypes and haplotypes on severe toxicity and efficacy in patients treated with fluorouracil, leucovorin, and irinotecan combination chemotherapy (FOLFIRI) for metastatic colorectal cancer (mCRC) from the two prospective multicenter phase II studies in Japan. The FLIGHT1 study was a first-line FOLFIRI trial, and FLIGHT2 was a FOLFOX-refractory, second-line FOLFIRI trial. A total of 73 patients agreed to additional analysis, and were genotyped for UGT1A polymorphisms, UGT1A1*28 (TA6>TA7), UGT1A1*6 (211G>A), UGT1A1*27 (686C>A), UGT1A1*60 (−3279T>G), UGT1A1*93 (−3156G>A), UGT1A7 (−57T>G), UGT1A7*3 (387T>G, 622T>C), and UGT1A9*22 (T9>T10). Of 73 patients, 34 developed G3/4 severe hematological toxicities. The toxicities were significantly more frequent in patients with UGT1A1*6 (211A), UGT1A7 (387G), and UGT1A9*22 reference alleles (T9). Haplotype I, which consists of all favorable alleles, was associated with a significant reduction in hematologic toxicity (P = 0.031). In contrast, haplotype II, which contains four high-risk alleles, showed significantly higher hematologic toxicity than the other haplotypes (P = 0.010). Six out of seven patients who were homozygous for UGT1A1*28 or *6 experienced severe hematological toxicity despite the fact that their response rate was not impaired (42.9%). We concluded that UGT1A polymorphisms, especially UGT1A1*6, are important for the prediction of severe toxicity of FOLFIRI in northeast Asian populations. In this regard, haplotype analyses should substantially impact the prediction of severe hematological toxicities of FOLFIRI. (Clinical Trial Registration: UMIN000002388 and UMIN000002476).