3611 Background: Current surveillance strategies for postoperative recurrence in colorectal cancer predominantly depend on tumor markers, imaging modalities, and endoscopic procedures; however, these approaches are challenged by suboptimal diagnostic accuracy. Consequently, the identification and validation of novel biomarkers represent an urgent clinical necessity. Sensitive microbiome-based approaches for postoperative risk stratification, surveillance optimization, and early recurrence detection may substantially influence clinical decision-making and resource allocation for colorectal cancer patients. Methods: This cohort study included CRC fecal and tumor samples from four independent cohorts (n = 615) in the Fudan University Shanghai Cancer Centre (Shanghai, China) between January 2017 and December 2020, with follow-up through December 2024. Fecal samples from the discovery (n = 250), training-validation cohorts (n = 153), as well as tumors with paired adjacent normal tissues from the investigation cohort (n = 244), were subjected to multi-omics analyses. We developed prediction models of disease-free survival by additionally incorporating dual-microbial signature including the abundance of Roseburia and Bifidobacterium. Both models were evaluated by internal validation cohort, and visual nomograms of prediction models were constructed accordingly. Results: We found that the combined high abundance of Roseburia and Bifidobacterium was associated with favorable outcomes (P = 0.039), especially in male (P = 0.018) and early-onset colorectal cancer (P = 0.047). Integration of clinical variables with a dual-microbial signature achieved a mean AUC > 0.8 for recurrence prediction. A high combined abundance of Roseburia and Bifidobacterium was linked to reduced enrichment of primary bile acid biosynthesis pathways and increased enrichment of propanoate metabolism pathways. Conclusions: This study proposes a novel strategy for CRC recurrence risk assessment and highlights the potential of gut microbiota as a tertiary preventive approach in CRC management. However, the proposed model has not yet been implemented in clinical practice and requires validation in large, multicenter prospective cohorts before clinical translation.
Chen et al. (Wed,) studied this question.