A two-step pre-treatment screening strategy combining enzyme phenotyping and targeted DPYD genotyping: A pragmatic approach for preventing fluoropyrimidine toxicity.

Abstract

306 Background: Severe fluoropyrimidine toxicity is linked to reduced DPD activity. Universal DPYD genotyping has low sensitivity, high cost, long turnaround, and still fails to prevent toxicity in up to one-third of patients. Phenotyping with serum uracil and dihydrouracil:uracil (DHU:U) ratio is faster and more sensitive but less specific. We evaluated a two-step workflow—phenotyping all patients, then genotyping only phenotype-abnormal cases—to improve toxicity prediction and reduce genotyping requirements. Methods: In this study, 160 adults starting fluoropyrimidines underwent baseline fasting uracil and DHU:U testing. An abnormal phenotype was defined as uracil > 14.5 ng/mL or DHU:U < 6. The first 79 patients had universal DPYD phenotyping and genotyping; in the remaining patients, only abnormal phenotypes were genotyped. Grade ≥3 hematologic or gastrointestinal toxicities within cycles 1-3 were recorded along with dose reductions of 5-FU. Test performance for toxicity prediction was calculated for phenotype alone and the two-step strategy. Results: The median age was 47 (IQR 38–58), with 59% male and 73% having gastrointestinal cancers. Treatment included doublet chemotherapy (56%) and triplet (25%). Higher uracil increased toxicity risk (OR 1.15, p = 0.025); a higher DPYD activity score was protective (OR 0.79, p = 0.003). In the initial 79-patient cohort, 6 (7.6%) had an abnormal phenotype and 1 (1.2%) an abnormal genotype. Grade ≥3 toxicity occurred in 31 (39.2%): universal genotyping identified none, while the two-step strategy identified 4 (12.9%). No normal-phenotype patient with toxicity had an actionable variant. In the later 81-patient cohort undergoing two-step testing, 35 (43.2%) had an abnormal phenotype and 4 (4.9%) an abnormal genotype. Grade ≥3 toxicity was noted in 25 (30.8%), despite pre-emptive cycle 1 dose reduction in those with an abnormal phenotype (48.5%). In the Overall cohort (n = 160), 56 (35.0%) developed toxicity. The two-step strategy identified 16 (28.6%). Phenotype alone yielded 28.6% sensitivity, 76.2% specificity, 40.0% PPV, and 65.8% NPV. The two-step strategy improved specificity to 88.6% and PPV to 66.7%, keeping sensitivity and NPV unchanged. Under a sensitivity analysis projecting universal genotyping: assuming all untested were variant-negative gives 3/56 (5.4 %) identified; assuming all untested variant-positive gives 10/56 (17.9 %). We observed that only 2.8% of normal phenotype pateints with toxicity carried an actionable genotype variant. Conclusions: A two-step testing strategy predicts more severe toxicities, reduces genotyping requirements, and lowers costs compared to universal genotyping, offering a scalable, evidence-based method for safer fluoropyrimidine dosing.