Adoption of personalized treatment approaches in metastatic colorectal cancer (mCRC), which takes into consideration factors such as primary tumor location, RAS mutation status, and presence of microsatellite instabilities (MSI) for treatment selection, combined with an increased number of treatment options, has led to significant improvements in survival. However, the prognosis of patients with mCRC generally remains poor. Development of both new treatment options and more sensitive predictive biomarkers that account for mCRC heterogeneity is a high unmet need. Recently, consensus molecular subtypes (CMS) have emerged as a novel way to classify patients with mCRC. Under the CMS system, patients are sorted into 1 of 4 groups on the basis of disease biology: CMS1 (MSI immune), CMS2 (canonical), CMS3 (metabolic), and CMS4 (mesenchymal). These categories reflect biologic processes related to gene expression, rather than just presence or absence of mutations.
In order to test the validity of CMS as a prognostic and predictive biomarker in patients with mCRC, a retrospective analysis compared CMSs to outcomes in patients enrolled on the phase III CALGB/SWOG 80405 study, a trial that compared the addition of bevacizumab or cetuximab to first-line chemotherapy in patients with mCRC. CMS data from 581 patients were classified using a NanoString gene expression platform. CMS2 was the most common group (41.65%), followed by CMS4 (28.74%), CMS1 (17.9%), and CMS3 (11.7%). However, in right-sided tumors CMS1 was dominant (37.34%). CMS classification was significantly prognostic for overall survival, regardless of treatment regimen (P<.001), with median OS of 15 months, 40.3 months, 24.3 months, and 31.4 months for CMS1 through CMS4, respectively. A similar association between CMS and progression-free survival (PFS) was observed (P<.001). The association between CMS and survival was less pronounced in patients receiving bevacizumab than in those receiving cetuximab.
CMS classification could also be used to determine which patients would respond to bevacizumab or cetuximab treatment. Patients in the CMS1 group had prolonged OS (22.5 months vs 11.7 months; P<.001) and PFS (8.7 months vs 5.7 months; P<.001) with bevacizumab, compared to cetuximab. Conversely, in the CMS2 group, patients treated with cetuximab had superior OS compared to those treated with bevacizumab (42 months vs 36 months; P = .0046).
Among patients who were KRAS wildtype or had wildtype gene expression signatures, treatment with cetuximab resulted in significant improvements in both PFS (HR 0.67, P = .0362) and OS (HR 0.62, P = .0256) compared to treatment with bevacizumab, regardless of CMS group. Patients with KRAS mutant tumors did not derive a benefit from cetuximab treatment. Similarly, MSI status also appeared to influence outcomes on treatment. Among patients in CMS1 who were microsatellite stable, there was no benefit for bevacizumab over cetuximab, while those with high MSI had significantly improved OS on bevacizumab compared to cetuximab (P = .0091).
The investigators concluded that these data indicate the potential clinical utility of CMSs as a prognostic and predictive biomarker that may help guide selection of anti-VEGF and anti-EGFR therapy in patients with mCRC. Future refinement of this system, including evaluation in a prospective clinical trial is needed.