The consensus molecular subtypes of colorectal cancer as a predictive factor for chemotherapies against metastatic colorectal cancer.

736 Background: Since treatment options of chemotherapy for metastatic colorectal cancer (mCRC) have been diversified, biomarkers which could predict efficacy of chemotherapies are needed for optimal treatment. In this study, we analyzed the association of the treatment outcomes by several chemotherapies and the consensus molecular subtypes (CMS), which was developed from large-scale comprehensive gene expression analysis of colorectal cancer (Guinney J et al. Nat Med 21:1350-6, 2015). Methods: A total of 193 patients with previously treated mCRC were classified in 4 subtypes, CMS1 to CMS4 by using the microarrays data of primary tumors and the “CMS classifier” package for the R software. Then the association between the subtypes and the treatment outcomes was analyzed retrospectively. Results: The193 cases were classified into 21 CMS1, 53 CMS2, 69 CMS3 and 50 CMS4. The molecular characteristics of each subtype such as mutation status of KRAS and BRAF genes, and DNA methylation status were consistent with those of the previous study. First-line chemotherapies of the 193 patients were oxaliplatin (OX)-based treatment (121 patients), irinotecan (IRI)-based treatment (59 patients) and others (13 patients). In CMS4, the median progression-free survival (PFS) of the IRI-based group was significantly longer than that of the OX-based group (21.8 months vs. 10.5 months, p < 0.01). The median overall survival (OS) and objective response rate of the IRI-based group were also better than that of the OX-based group in CMS4. Among 193 patients, 103 were RAS wild-type and treated with anti-EGFR antibody. The effect of the anti-EGFR treatment was low in CMS1 (median PFS: 2.4 months, p < 0.01, median OS: 5.7 months, p < 0.01) and whereas was high in CMS2 (median PFS: 8.0 months, p = 0.05, median OS: 26.6 months, p = 0.02). CMS significantly correlated with PFS and OS after anti-EGFR treatment by univariate analysis. However the most useful predictive biomarker was a DNA methylation status by multivariable analysis. Conclusions: The biological characteristics of CMS may affect the therapeutic effect of anti-cancer agents in mCRC and, therefore, might be a new predictive factor of anti-cancer treatment.