Pharmacogenomic (PG) analysis for prediction of individual response to paclitaxel in 5-FU-refractory metastatic gastric cancer: Prediction formula of tumor response using novel marker genes and genotypes associated with the toxicity

Abstract

2538 Background: Taxanes offer hope for improving outcomes of metastatic gastric cancer patients including 5-FU failure cases, but the response remarkably varies among patients. We conducted this prospective PG study for paclitaxel monotherapy to develop a prediction formula of efficacy and to identify potent genetic markers of toxicity. Methods: Paclitaxel was intravenously given on Days 1, 8, and 15, every 4 weeks in 5-FU failure metastatic gastric cancer patients with typical eligibility criteria. Tumor and blood samples were collected before the initial paclitaxel administration for PG. PK analysis was done on day 1 of cycle 1. Tumor response and toxicity were evaluated by RECIST and CTCAE, respectively. Results: Forty-nine pts were enrolled, and up to date, 48 and 44 pts have been determined for worst toxicity and best tumor response, respectively. Observed response rate was 27.3% (12/44), and common grade 3/4 toxicities were neutropenia (31.3%) and leucopenia (20.8%). PG analysis demonstrated that CYP2C8*1C (w/v=27, v/v=5) and CYP2C8 -411T>C (w/v=24, v/v=10) were weakly associated with grade 3/4 leucopenia (p=0.087 and 0.092), and CYP2C8 IVS7+49T> A (w/v=9, v/v=6) might correlate with grade 3/4 anemia (p=0.039), although none of them correlated with any PK parameter. We identified 4 novel potent marker genes (SEPT5, MARN2, PER3, and PISA3) for paclitaxel efficacy in vitro through microarray expression analysis, and then successfully developed the best linear models, which converted the quantified expression data into objective clinical response, in terms of best tumor response (r=0.985, AICS=-5.269) and overall survival (r=0.980, AICS=6.953), using 15 and 13 data sets of gene expression and clinical response, respectively. At present, utility-confirmation analyses using other clinical samples appeared to show that the formulae could accurately predict tumor response. Conclusions: Polygenetic strategies using several known polymorphisms for toxicity and a prediction formula using 4 novel genes for efficacy would be of predictive value for individual response to paclitaxel. No significant financial relationships to disclose.