Gut Microbiome Signatures Predict Hepatocellular Carcinoma Development in High-Risk Individuals
Gut Microbiome Signatures Predict Hepatocellular Carcinoma Development in High-Risk Individuals in a Prospective Cohort
Sekar, K., Lee, J. W. J., Sim, Y. K., Chong, S. L., Chew, S. L., Lee, G. H., Chong, O. F., Choo, J. W. S., Ithnin, S. B., Chang, J. P. E., De Roza, M. A., Chandrasekaran, K. R., Ong, K. K., Chia, P. C. Y., Tan, K. T., Tan, N. C., Toh, H. C., Wong, Y. J., Yang, W. L., … Chow, P., on behalf of the Asia-Pacific Hepatocellular Carcinoma (AHCC) Trials Group. (2026). Gut microbiome signatures predict hepatocellular carcinoma development in high-risk individuals in a prospective cohort [Poster presentation]. 11th Microbiome R&D and Business Collaboration Congress.
BACKGROUND: Hepatocellular carcinoma (HCC), the third leading cause of cancer-related mortality worldwide, is frequently diagnosed at advanced stages, limiting curative treatment options. Current surveillance strategies, including ultrasound and alpha-fetoprotein (AFP), have suboptimal sensitivity and do not effectively stratify risk among individuals with chronic liver disease. There is therefore an urgent need for non-invasive biomarkers that can improve early detection and predict HCC risk.
METHODS: We performed a prospective shotgun metagenomic analysis within the AHCC10 ELEGANCE cohort (N = 2002), a multi-centre longitudinal study in Singapore comprising high-risk individuals with chronic liver disease, including cirrhosis (40%), chronic hepatitis B (47.2%), hepatitis C (0.8%), and MASLD/MASH (12%). An independent PROSECT HCC cohort (n = 101) was included for comparison. Participants underwent 6-monthly HCC surveillance with longitudinal biospecimen collection. Over a median follow-up of 35.5 months, 62 individuals developed HCC.
RESULTS: After adjusting for clinical covariates (age, gender, ethnicity, cirrhosis, HBV, diabetes, and MASLD), pre-diagnostic stool samples from individuals who developed HCC (n = 62) showed a distinct microbiome profile compared to controls (n = 1916). This was characterized by enrichment of oral-associated taxa (e.g., Streptococcus, Veillonella) and depletion of short-chain fatty acid–producing commensals (e.g., Roseburia, Ruminococcus, Anaerobutyricum).
Established HCC cases (n = 149) exhibited further enrichment of facultative pathogens (e.g., Phocaeicola, Escherichia, Hungatella) alongside persistent depletion of butyrate-producing taxa, suggesting impaired intestinal barrier function and disruption of the gut–liver axis.
Using LASSO logistic regression with 5-fold cross-validation, we developed predictive models for HCC risk and diagnosis. The risk model (15 microbial species) achieved an AUC of 0.80 (95% CI: 0.74–0.86), while the diagnostic model (40 microbial species) achieved an AUC of 0.83 (95% CI: 0.78–0.85). Notably, the microbiome-based risk score identified individuals up to 4 years prior to HCC diagnosis.
CONCLUSIONS: Gut microbiome alterations precede clinical HCC and may serve as early, non-invasive biomarkers for risk stratification. Integration into surveillance programmes could enhance early detection and enable personalized monitoring.