Consensus molecular subtypes, mutational features, and circulating tumor DNA dynamics in colorectal cancer: A real-world clinico-genomic analysis.

3652 Background: Consensus molecular subtypes (CMS) define biologically distinct colorectal cancer (CRC) groups with differences in tumor microenvironment, genomic features, and prognosis. While CMS classification is well established in tissue-based transcriptomic studies, its relationship with circulating tumor DNA (ctDNA) dynamics and molecular features in real-world clinical practice remains incompletely characterized. We evaluated CMS subtype distribution, mutational characteristics, and ctDNA detection patterns in a large real-world CRC cohort. Methods: Patients with CRC who underwent whole-transcriptome sequencing via Altera (Natera, Inc.) testing were identified from Natera’s real-world clinico-genomic database. CMS subtype assignment was based on the whole-transcriptome sequencing, and then correlated with Signatera results, a personalized, tumor-informed ctDNA assay. In patients with stage II–III CRC, ctDNA positivity was assessed during the MRD window (1–10 weeks post-surgery) and during post-MRD surveillance (>10 weeks post-surgery). Overall survival (OS) was estimated using the mortality database (Veritas Data Research, Fact of Death Mortality Data Index). Results: Among 8,456 CRC cases with available molecular data, CMS distribution was 10% CMS1, 26% CMS2, 12% CMS3, 22% CMS4, and 29% inconclusive. While canonical anatomic patterns were recapitulated (CMS1 right-sided, CMS2 left-sided, CMS3 rectal), CMS4 tumors were enriched in advanced-stage disease. CMS1 demonstrated high MSI-H prevalence (53%), frequent BRAF V600E (35%) and KMT2D (44%) alterations, and enrichment of MMR deficiency and APOBEC and POLE mutational signatures. CMS2 tumors were almost exclusively microsatellite stable (99.7%) and enriched for APC (84%) and TP53 (80%) mutations, whereas CMS3 showed intermediate MSI-H prevalence (10%) with frequent APC (74%) and KRAS (61%) alterations. CMS4 tumors exhibited low MSI-H prevalence (2%) but frequent TP53 (60%) and APC (59%) mutations. Among patients with stage II–III CRC, ctDNA positivity during the early post-operative MRD window was not associated with CMS subtype. In contrast, during later surveillance, ctDNA positivity was more frequent in CMS4 tumors compared with CMS1–CMS2 (36.8% vs 28.1%). Among MRD-positives, CMS1 was associated with the shortest OS, whereas CMS2 demonstrated the longest OS. No difference in OS was noted in MRD-negative patients. Conclusions: In this large real-world cohort, CMS subtypes were recapitulated with expected clinicogenomic and mutational associations and demonstrated distinct ctDNA detection patterns during surveillance. Integration of CMS biology, mutational features, and longitudinal ctDNA dynamics provides insights into residual disease and recurrence risk beyond static tissue-based classification.