A Combined Strategy of SAGE and Quantitative PCR Provides a 13-Gene Signature that Predicts Preoperative Chemoradiotherapy Response and Outcome in Rectal Cancer

PURPOSE: Preoperative chemoradiotherapy (CRT) is the treatment of choice for rectal cancer (RC), but half of the patients do not respond, suffer unnecessary toxicities, and surgery delays. We aimed to develop a model that could predict a clinically meaningful response to CRT by using formalin-fixed paraffin-embedded (FFPE) biopsies. EXPERIMENTAL DESIGN: We first carried out an exploratory screening of candidate genes by using SAGE technology to evaluate dynamic changes in the RC transcriptome in selected refractory patients before and after CRT. Next, 53 genes (24 from SAGE and 29 from the literature) were analyzed by qPCR arrays in FFPE initial biopsies from 94 stage II/III RC patients who were preoperatively treated with CRT. Tumor response was defined by using Dworak's tumor regression grade (2-3-4 vs. 0-1). Multivariate Cox methods and stepwise algorithms were applied to generate an optimized predictor of response and outcome. RESULTS: In the training cohort (57 patients), a 13-gene signature predicted tumor response with 86% accuracy, 87% sensitivity, and 82% specificity. In a testing cohort (37 patients), the model correctly classified 6 of 7 nonresponders, with an overall accuracy of 76%. A signature-based score identified patients with a higher risk of relapse in univariate (3-year disease-free survival 64% vs. 90%, P = 0.001) and multivariate analysis (HR = 4.35 95% CI: 1.2-15.75, P = 0.02), in which it remained the only statistically significant prognostic factor. CONCLUSIONS: A basal 13-gene signature efficiently predicted CRT response and outcome. Multicentric validation by the GEMCAD collaborative group is currently ongoing. If confirmed, the predictor could be used to improve patient selection in RC studies.