Gene expression profiles and molecular markers to predict recurrence of Dukes' B colon cancer.

The 5-year survival rate of patients with Dukes' B colon cancer is approximately 75%. Identification of the patients at high risk of recurrence in this group would allow better staging and more informed use of adjuvant chemotherapy. In this study, we used DNA chip technology to systematically identify new prognostic markers for tumor relapse in Dukes' B patients. Using Affymetrix U133a GeneChip containing approximately 22,000 transcripts (Affymetrix, Santa Clara, CA), RNA samples from 74 patients with Dukes' B colon cancer were analyzed. Thirty-one patients developed tumor relapse in less than 3 years, whereas 43 patients remained disease-free for more than 3 years after surgery. Two supervised class prediction approaches were used to identify gene markers that can best discriminate between patients who would experience relapse and patients who would remain disease-free. A multivariate Cox model was built to predict recurrence. Gene expression profiling identified a 23-gene signature that predicts recurrence in Dukes' B patients. This signature was validated in 36 independent patients. The overall performance accuracy was 78%. Thirteen of 18 relapse patients and 15 of 18 disease-free patients were predicted correctly, giving an odds ratio of 13 (95% CI, 2.6 to 65; P =.003). The log-rank test indicated a significant difference in disease-free time between the predicted relapse and disease-free patients (P =.0001). The clinical value of these markers is that the patients at a high predicted risk of relapse (13-fold risk) could be upstaged to receive adjuvant therapy, similar to Dukes' C patients. Our data highlight the feasibility of a prognostic assay that could focus more intensive treatment for localized colon cancer.