Aims Cancer patients with reduced dihydropyrimidine dehydrogenase (DPD) activity are at increased risk of severe fluoropyrimidine (FP)‐related adverse events (AE). Guidelines recommend FP dosing adjusted to genotype‐predicted DPD activity based on four DPYD variants (rs3918290, rs55886062, rs67376798 and rs56038477). We evaluated the relationship between three further DPYD polymorphisms: c.496A>G (rs2297595), *6 c.2194G>A (rs1801160) and *9A c.85T>C (rs1801265) and the risk of severe AEs. Methods Consecutive FP‐treated adult patients were genotyped for “standard” and tested DPYD variants, and for UGT1A1*28 if irinotecan was included, and were monitored for the occurrence of grade ≥3 (National Cancer Institute Common Terminology Criteria) vs. grade 0–2 AEs. For each of the tested polymorphisms, variant allele carriers were matched to respective wild type controls (optimal full matching combined with exact matching, in respect to: age, sex, type of cancer, type of FP, DPYD activity score, use of irinotecan/UGT1A1, adjuvant therapy, radiotherapy, biological therapy and genotype on the remaining two tested polymorphisms). Results Of the 503 included patients (82.3% colorectal cancer), 283 (56.3%) developed grade ≥3 AEs, mostly diarrhoea and neutropenia. Odds of grade ≥3 AEs were higher in c.496A>G variant carriers (n = 127) than in controls (n = 376) OR = 5.20 (95% CI 1.88–14.3), Bayesian OR = 5.24 (95% CrI 3.06–9.12). Odds tended to be higher in c.2194G>A variant carriers (n = 58) than in controls (n = 432) OR = 1.88 (0.95–3.73), Bayesian OR = 1.90 (1.03–3.56). c.85T>C did not appear associated with grade ≥3 AEs (206 variant carriers vs. 284 controls). Conclusion DPYD c.496A>G and possibly c.2194G>A variants might need to be considered for inclusion in the DPYD genotyping panel.