Logistic Regression Model of the Clinical Response to 5-Fluorouracil Based Chemotherapy for Metastatic Colorectal Cancer Patients

Abstract

Several genes have been involved in drug resistance but none are currently used in the drug decision process. To address this problem, mRNA levels were measured for the 5-fluorouracil metabolism-related genes, thymidylate synthase, thymidine phosphorylase and dihydropyrimidine dehydrogenase in tumor samples of 40 patients with synchronous metastatic colon cancer by quantitative RT-PCR. Drug response and overall survival were also obtained for each patient. A logistic regression model was defined to calculate a response predicting score (RPS) with gene expression levels. This RPS split responders from non-responders as, at the best statistical threshold (0.35), the area of receiver operating characteristic (ROC) curve established with this method was 0.82 and sensitivity and specificity were respectively 100 % and 65.4 %. Furthermore patients with scores above 0.35 tended to have better overall survival than those with a score less than 0.35 (p=0.09).