Pre-therapeutic DPYD pharmacogenetic test to prevent fluoropyrimidines (FL)-related toxicities is not yet common practice in medical oncology. We aimed at investigating the clinical validity of DPYD genetic analysis in a large series of oncological patients. Six hundred three cancer patients, treated with FL, have been retrospectively tested for eight DPYD polymorphisms (DPYD-rs3918290, DPYD-rs55886062, DPYD-rs67376798, DPYD-rs2297595, DPYD-rs1801160, DPYD-rs1801158, DPYD-rs1801159, DPYD-rs17376848) for association with Grade ≥3 toxicity, developed within the first three cycles of therapy. DPYD-rs3918290 and DPYD-rs67376798 were associated to Grade ≥3 toxicity after bootstrap validation and Bonferroni correction (p = 0.003, p = 0.048). DPYD-rs55886062 was not significant likely due to its low allelic frequency, nonetheless one out of two heterozygous patients (compound heterozygous with DPYD-rs3918290) died from toxicity after one cycle. Test specificity for the analysis of DPYD-rs3918290, DPYD-rs55886062 and DPYD-rs67376798 was assessed to 99%. Among the seven patients carrying one variant DPYD-rs3918290, DPYD-rs55886062 or DPYD-rs67376798 allele, not developing Grade ≥3 toxicity, 57% needed a FL dose or schedule modification for moderate chronic toxicity. No other DPYD polymorphism was associated with Grade ≥3 toxicity. Our data demonstrate the clinical validity and specificity of the DPYD-rs3918290, DPYD-rs55886062, DPYD-rs67376798 genotyping test to prevent FL-related Grade ≥3 toxicity and to preserve treatment compliance, and support its introduction in the clinical practice.