A nested case-control study on the relationship between gut microbiota and non-alcoholic fatty liver disease

The gut microbiota plays an important role in non-alcoholic fatty liver disease (NAFLD). This study aimed to explore gut microbiota-associated biomarkers for incident NAFLD and evaluate their predictive value, providing preliminary evidence for early risk identification.

From an initial cohort of 1,772 participants, 404 were included in the final analysis, comprising 101 incident NAFLD cases and 303 age-matched controls (1:3). Gut microbiota composition and function were characterized by metagenomic sequencing. Microbial features were compared between participants who developed NAFLD (NAFLD -/+) and those who did not (NAFLD -/-). A CatBoost model integrating gut microbiota, functional pathways, and clinical indicators was constructed to predict NAFLD, with a random forest (RF) model on the same features for comparison. Progressively enriched feature sets assessed the incremental value of microbiota-related features beyond clinical indicators, and SHapley additive exPlanations (SHAP) values interpreted the model outputs.

The NAFLD (-/+) group had significantly lower species-level beneficial gut microbiota (P < 0.05), and the two groups differed markedly in gut microbiota metabolic enrichment. The CatBoost model achieved an area under the curve (AUC) of 0.762, indicating moderate predictive performance for incident NAFLD. We also identified significant intergroup differences in gut microbiota composition within the diagnostic model, including the key species Coprobacillus unclassified and the metabolic pathway PWY-5030.

Gut microbiota alterations may serve as potential early biomarkers for incident NAFLD and contribute to its pathogenesis. The CatBoost model integrating multi-omics and clinical data shows moderate predictive ability, supporting the potential of gut microbiota-based approaches for non-invasive risk stratification.