Aim Dihydropyrimidine dehydrogenase (DPD) deficiency can be detected by phenotyping (measurement of plasma uracil [U], with U ≥ 16 μg/L defining a partial deficiency) and/or by genotyping (screening for the four most frequent DPYD variants). We aimed to determine the proportion of discrepancies between phenotypic and genotypic approaches and to identify possible explanatory factors. Methods Data from patients who underwent both phenotyping and genotyping were retrospectively collected. Complementary genetic analyses (genotyping of the variant c.557A>G and DPYD sequencing) were performed for patients with U ≥ 16 μg/L without any common variants. The characteristics of patients classified according to the congruence of the phenotyping and genotyping approaches were compared (Kruskal–Wallis test), and determinants of U levels were studied in the whole cohort (linear model). Results Among the 712 included patients, phenotyping and genotyping were discordant for 12.5%, with 63 (8.8%) having U ≥ 16 μg/L in the absence of a common variant. Complementary genetic investigations marginally reduced the percentage of discrepancies to 12.1%: Among the nine additional identified variants, only the c.557A>G variant, carried by three patients, had been previously reported to be associated with DPD deficiency. Liver dysfunction could explain certain discordances, as ASAT, ALP, GGT and bilirubin levels were significantly elevated, with more frequent liver metastases in patients with U ≥ 16 μg/L and the absence of a DPYD variant. The impact of cytolysis was confirmed, as ASAT levels were independently associated with increased U ( p < 0.001). Conclusion The frequent discordances between DPD phenotyping and genotyping approaches highlight the need to perform these two approaches to screen for all DPD deficiencies.