Up to 30 % of patients undergoing 5-fluorouracil (5FU)-based chemotherapy experience severe toxicity. Dihydropyrimidine dehydrogenase (DPD) deficiency explains 36-61 % of cases. Predicting toxicity is an unmet challenge. Uracil breath test (UraBT) consists of measuring (13)CO2 in exhaled breath after ingestion of 2-(13)C-uracil to evaluate pyrimidine (and 5FU) catabolism. We studied 33 gastrointestinal cancer patients previously exposed to 5FU: Thirteen had grade 3-4 and 20, grade 0-1 toxicity. The following tests were used to evaluate pyrimidine catabolism: (1) sequencing of three exons of DPYD; (2) plasma dihydrouracil/uracil ratio (UH2/U); and (3) UraBT. We tested the performance of UraBT to discriminate patients who had grade 0-1 toxicity versus grade 3-4 toxicity and patients with and without proven DPD deficiency. Of the thirteen patients, four grade 3-4 toxicity patients were proved to be DPD-deficient: Three had deleterious mutations (IVS14 + 1G>A in one; single nucleotide polymorphism 2846A>T in two), and one had low UH2/U ratio. Mean delta over baseline in 50 min (DOB50) significantly differed between groups. DOB50≤161.4 discriminated individuals with grade 3-4 versus grade 0-1 toxicity (sensitivity = 61.5 %; specificity = 85 %) and DPD-deficient versus non-DPD-deficient (sensitivity = 75 %; specificity = 85 %). UraBT has moderate accuracy in discriminating individuals who manifested severe toxicity from those who had mild or no toxicity to 5FU.