FUSAFE individual patient data meta-analysis (MA) to assess the performance of dihydropyrimidine dehydrogenase (DPD) gene polymorphisms for predicting grade 4-5 fluoropyrimidine (FP) toxicity

Abstract

Background: Despite decades of research, performance of DPYD genotyping to predict FP toxicity (tox) is poorly documented. GPCO-UNICANCER and RNPGx groups initiated the FUSAFE MA on individual patient data (IPD) to assess the prognostic value of consensual deleterious DPYD variants on grade (G) 4-5 FP tox. Methods: Eligibility criteria included unbiased recruitment of Caucasian patients (pts) without FP dose adjustment based on DPD status. Main endpoint was 12 weeks hematological or digestive G4-5 tox. Age, sex, body mass index (BMI), advanced stage (M- vs M+), FP drug (5FU vs capecitabine), FP administration (bolus±continuous vs continuous alone or p.o.) and associated anticancer drugs (AAD) were collected. Multivariable logistic models were applied. Performance was assessed by AUC and diagnostic indices maximizing Youden index. Results: From the 18 identified eligible studies (10230 pts), 14 were included (9030 pts), with complete IPD collected for 6403 pts (84% colorectal, 16% M+, 66% 5FU, 80% AAD). G4-5 tox prevalence was 8% (518 events). DPYD variants *2A, D949V, *13 and HapB3 were carried by 0.9%, 1.2%, 0.2% and 3.9% of pts, respectively. The clinical model (M1) retained age, sex, BMI, FP-administration, AAD. Adding variants *2A/D949V/*13 (at least one mutated allele) significantly (p