Revolutionizing hepatic fibrosis staging: A machine learning approach combining clinical, biochemical, and microbiome insights.

Random Forest and XGBoost models achieved excellent classification accuracy for hepatic fibrosis staging in NASH patients.

• Random Forest achieved a balanced accuracy of 99.1%
• XGBoost achieved an area under the curve (AUC) value of 1.0
• Models were trained on 1834 patients with biopsy-confirmed NASH
• Performance was assessed using 10-fold cross-validation with the primary training cohort and external validation on an independent hospital database
• Fibrosis stages classified were F0, F1, F2, F3, and F4

The addition of microbiome features enhanced the predictive capability of the machine learning models for fibrosis staging.

• Microbiome features included diversity indices and relative abundance of certain taxa
• Microbiome data was obtained via 16S rRNA gene sequencing
• Enhancement of predictive capability indicates that the gut-liver axis plays a significant role in the development of NASH
• SHAP (SHapley Additive Explanations) analysis was used to identify which clinical and microbiome features affected model predictions for fibrosis stages

Advanced stages of fibrosis (F3 and F4) were associated with significant gut microbiome dysbiosis.

• Advanced fibrosis stages showed increased relative abundance of pathogenic bacteria including Escherichia-Shigella and Enterococcus
• Advanced fibrosis stages showed decreased relative abundance of Akkermansia and Ruminococcus
• Dysbiosis patterns were identified using SHAP analysis of the machine learning models
• Microbiome profiling was performed via 16S rRNA gene sequencing across all fibrosis stages

The machine learning models demonstrated superiority over traditional scoring systems for hepatic fibrosis staging.

• Traditional scoring systems compared included APRI (AST to Platelet Ratio Index) and FIB-4
• The ML models were described as superior for guiding clinical decision making and risk assessment
• The models provide a non-invasive method for determining hepatic fibrosis stage in NASH
• The study cohort comprised patients from multiple healthcare systems with biopsy-confirmed NASH

The study included a retrospective cohort of 1834 patients with biopsy-confirmed NASH from multiple healthcare systems.

• Total sample size was 1834 patients
• All patients had biopsy-confirmed NASH with a stated fibrosis stage (F0, F1, F2, F3, F4)
• Patients were drawn from multiple healthcare systems
• Clinical variables included liver function tests, demographics, and microbiome profiles
• An independent hospital database was used for external validation