A pharmacogenetic model predicting low paclitaxel clearance based on the DMET platform

• Published in: JOURNAL OF CLINICAL ONCOLOGY Vol. 31; no. 15_suppl; p. 2597
• Main Authors: Nieuweboer, Annemieke J.M., de Graan, Anne-Joy M., Elens, Laure, Smid, Marcel, Martens, John W. M., Sparreboom, Alex, Friberg, Lena E., Elbouazzaoui, Samira, Wiemer, Erik A.C., van der Holt, Bronno, Verweij, Jaap, van Schaik, Ron H.N., Mathijssen, Ron H.J.
• Format: Journal Article Conference Proceeding
• Language: English
• Published: 20.05.2013
• Subjects: Medicin och hälsovetenskap
• ISSN: 0732-183X; 1527-7755
• DOI: 10.1200/jco.2013.31.15_suppl.2597

Abstract only 2597 Background: Paclitaxel (PTX) is a commonly used cytotoxic agent. It is metabolized by P450 cytochrome iso-enzymes CYP3A4 and CYP2C8 and has high interindividual variability in pharmacokinetics (PK) and toxicity. Here, we present a genetic prediction model to identify patients with low PTX clearance (CL) using the new Drug-Metabolizing Enzyme and Transporter (DMET; Affymetrix) platform, capable of detecting 1,936 genetic variants (SNPs) in 225 genes. Methods: In a PK study, 270 Caucasian cancer patients were treated with PTX. PK parameters were determined using a limited sampling strategy. HPLC or LC-MS/MS were used to determine PTX plasma concentrations and non-linear mixed effects modelling (NONMEM) was used to estimate individual unbound CL from previously developed PK population models. Subsequently, the cohort of patients was randomly split into a training and validation set. In all patients, the presence of SNPs in metabolic enzymes and transporters was determined using the DMET platform. Selected SNPs were subsequently validated in the validation set. Results: Baseline characteristics were comparable in both sets. The mean CL of the total cohort was 488 ± 149 L/h and the threshold for low CL was set at 339 L/h (1 SD < total mean CL). 14 SNPs were selected to be included in the prediction model and validated in the validation set. For none of these 14 SNPs, evidence for a biological plausible link to taxane metabolism exists. The developed prediction model had a sensitivity of 95% to identify low PTX CL, a positive predictive value of 22% and remained significantly associated with low CL after multivariate analysis correcting for age, gender and Hb levels at start of therapy (P=0.024). Conclusions: This is the first considerably-sized application of the DMET platform to explain PK variability of a widely used anti-cancer drug. Although this validated prediction model for PTX CL had a high sensitivity, its positive predictive value is too low to be of direct clinical use. Likely, genetic variability in DMET genes alone does not sufficiently explain PTX CL, as for example environmental factors may also influence PTX metabolism.