Risk Factors and Prediction Model for Liver-Related Outcomes in Elderly Patients With Steatotic Liver Disease

Overview

This is a single-center, retrospective cohort study based on data from the Nanjing Elderly Steatotic Liver Disease Cohort. The study aims to investigate risk factors for liver-related adverse outcomes (including significant fibrosis, advanced fibrosis, cirrhosis, hepatocellular carcinoma, and liver-related death) and extrahepatic outcomes (new-onset type 2 diabetes, chronic kidney disease, and cardiovascular disease) in elderly patients (aged ≥60 years) with steatotic liver disease. A total of approximately 10,000 participants will be included. Baseline and annual follow-up data on demographics, lifestyle, anthropometric measurements, laboratory tests, abdominal ultrasound, and medication use will be collected. Risk prediction models will be developed using machine learning algorithms. The study is observational and does not involve any intervention.

Background: Steatotic liver disease (SLD) is highly prevalent among the elderly and can progress to cirrhosis and hepatocellular carcinoma. However, large-scale longitudinal studies focusing on risk prediction in Chinese elderly populations are limited. Objectives: Primary objective is to identify risk factors and develop a prediction model for significant fibrosis. Secondary objectives include models for advanced fibrosis, cirrhosis, hepatocellular carcinoma, liver-related death, and extrahepatic outcomes (type 2 diabetes, chronic kidney disease, cardiovascular disease), as well as comparison of outcomes across SLD subtypes (MASLD, MetALD, ALD). Methods: This is a single-center, retrospective cohort study using data from the Nanjing Elderly Steatotic Liver Disease Cohort (initiated in 2018). Approximately 10,000 participants aged ≥60 years with imaging or biopsy-proven hepatic steatosis will be included. Baseline and annual follow-up data include demographics, lifestyle factors (smoking, alcohol, diet, physical activity), anthropometric measurements, laboratory tests (glucose, lipids, liver and kidney function), abdominal ultrasound, and medication use. The primary outcome is significant fibrosis (FIB-4 ≥2.67); secondary outcomes include advanced fibrosis, cirrhosis, hepatocellular carcinoma, liver-related death, and extrahepatic outcomes. Cox regression will be used for univariate and multivariate analyses. Machine learning algorithms (random forest, XGBoost, Cox-boost) will be applied to develop prediction models, with performance evaluated by time-dependent ROC curves, calibration curves, and decision curve analysis. A competing risk model will account for death as a competing event. The study has been approved by the Ethics Committee of the Affiliated Nanjing Drum Tower Hospital of Nanjing University Medical School.