Fluoropyrimidines combined with other agents are commonly used for gastrointestinal cancer treatment. Considering that severe toxicities occur in 30% of patients, we aimed to structure a nomogram to predict toxicity, based on metabolic parameter and patients' characteristics. We retrospectively enrolled patients affected by gastrointestinal tract cancers. Pretreatment 5-fluorouracil (5-FU) degradation rate and DPYD, TSER, MTHFR A1298T, and C677T gene polymorphisms were characterized. Data on toxicities were collected according to CTCAE v3.0. Multivariate logistic regression analysis was used to structure a nomogram. 642 patients were enrolled (384 men; 258 female; median age: 67 years, range: 27-87): 449 (69.9%) patients were affected by colorectal cancer; 118 (18.4%) by gastroesophageal cancer; 66 (10.3%) by pancreatic cancer; and nine (1.4%) by other cancers. Grade 3-4 toxicities were observed in 118 (18.4%) patients and were most frequently observed in patients with altered 5-FU degradation rate (43.5 and 26.7% of the patients in the poor metabolizer and in the ultrarapid metabolizer group respectively, vs. 17% in the normal metabolizer group) and in DPYD heterozygous mutated patients (83.3% of the patients). Age, DPYD status, the number of drugs administered, and 5-FU degradation rate value were associated to severe toxicities. On the basis of these findings, we structured a nomogram to assess a score to predict the risk of developing severe toxicity. Compared with the available pharmacogenetic tests, this approach can be applied to the whole population, predicting the risk for severe toxicity, with an easy, low-cost, and not invasive technique.