An evaluation of the diagnostic accuracy of a panel of variants in dpyd and a single variant in enosf1 for predicting common capecitabine related toxicities

Abstract

Efficacy of 5‐Fluorouracil (5‐FU)‐based chemotherapy is limited by significant toxicity. Tests based upon variants in the Clinical Pharmacogenetics Implementation Consortium (CPIC) guidelines with high level evidence of a link to dihydropyrimidine dehydrogenase (DPD) phenotype and 5‐FU toxicity are available to identify patients at high risk of severe adverse events (AEs). We previously reported associations between rs1213215, rs2612091, and NM_000110.3:c.1906‐ 14763G>A (rs12022243) and capecitabine induced toxicity in clinical trial QUASAR 2. We also iden-tified patients with DPD deficiency alleles NM_000110.3: c.1905+1G>A, NM_000110.3: c.2846C>T, NM_000110.3:c.1679T>G and NM_000110.3:c.1651G>A. We have now assessed the frequency of thirteen additional DPYD deficiency variants in 888 patients from the QUASAR 2 clinical trial. We also compared the area under the curve (AUC)—a measure of diagnostic accuracy—of the high-level evidence variants from the CPIC guidelines plus and minus additional DPYD deficiency variants and or common variants associated with 5‐FU toxicity. Including additional DPYD deficiency variants retained good diagnostic accuracy for serious adverse events (AEs) and improved sensitivity for predicting grade 4 haematological toxicities (sensitivity 0.75, specificity 0.94) but the improve-ment in AUC for this toxicity was not significant. Larger datasets will be required to determine the benefit of including additional DPYD deficiency variants not observed here. Genotyping two common alleles statistically significantly improves AUC for prediction of risk of HFS and may be clini-cally useful (AUC difference 0.177, sensitivity 0.84, specificity 0.31).