A molecular profile of colorectal cancer to guide prognosis and therapy after resection of primary or metastatic disease.

Abstract

339 Background: Biomarkers used to identify patients at high risk for recurrence after surgical resection of colorectal cancer (CRC) lack predictive capacity and are applicable only to localized disease. As the use of adjuvant therapy in this setting can be controversial, the development of a biomarker that is prognostic, predictive, and applicable after resection of primary or metastatic disease may dramatically improve patient outcomes. Methods: We compiled microarray data from 850 patients with predominantly primary CRC from four publically available datasets to predict activity of 19 oncogenic pathways to generate patterns of pathway deregulation for each patient tumor. Mixture modeling was applied to the samples using affinity propagation to subgroup tumors bearing unique patterns of pathway deregulation. The model was then applied to a dataset of 133 predominantly metastatic CRC samples from patients undergoing curative surgical resection at the Duke University Medical Center. Results: Mixture modeling resulted in the identification of 6 distinct molecular subgroups of CRC (MSCC) based on pathway deregulation for both primary and metastatic CRC. Kaplan-Meier analysis of the primary data set revealed differences in recurrence free survival among the 6 MSCC (p=0.0004) with patients assigned to MSCC3 having the poorest prognosis and patients assigned to MSCC4 having the best prognosis. A similar result was obtained when the model was applied to the metastatic data set (p<0.05). Comprehensive analysis of 41 oncogenes showed that no specific mutations were associated with the any of the subgroups. Finally, we observed that MSCCs demonstrate differential sensitivity to specific targeted agents such as MSCCs 1, 2 and 3 to inhibition of Her2 by Lapatinib and EGFR by erlotinib (p<0.05). Conclusions: Prediction of oncogenic signaling pathway deregulation is a powerful tool that may be used to molecularly sub-classify CRCs into clinically relevant subgroups. These subgroups have prognostic implications for recurrence following surgical resection and may help to identify therapies in the adjuvant setting after resection of either primary or metastatic disease.