Laboratory-based nutritional risk assessment including GDF15 in cirrhosis: a multicenter retrospective study

Abstract Malnutrition is a common complication of cirrhosis and is associated with adverse outcomes. However, current nutritional risk assessment tools may not fully incorporate objective laboratory biomarkers. This study aimed to develop a biomarker-based prediction model for nutritional risk assessment in patients with cirrhosis. We retrospectively collected the clinical data of 551 patients with liver cirrhosis from two centers between April 2023 and April 2026. The Royal Free Hospital–Nutritional Prioritizing Tool (RFH-NPT) classification was used as the reference standard for nutritional risk categorization. Candidate predictors were selected using univariate logistic regression and feature selection methods. Model performance was evaluated by discrimination, calibration, and clinical utility. Model interpretability was evaluated using SHapley Additive exPlanations (SHAP). The added predictive value of the predictors was assessed using the net reclassification improvement (NRI) and integrated discrimination improvement (IDI). Hepatic decompensation was assessed using survival analysis. The laboratory-based approach incorporating growth differentiation factor 15 (GDF15), carbohydrate antigen 19 − 9 (CA19-9), gamma-glutamyl transferase (GGT), Child–Turcotte–Pugh (CTP) score, estimated glomerular filtration rate (eGFR), and triglyceride–glucose (TyG) index demonstrated satisfactory discrimination for nutritional risk stratification in cirrhosis (area under the receiver operating characteristic curve [AUROC] = 0.857). Calibration showed acceptable agreement, and decision curve analysis (DCA) suggested potential clinical utility. GDF15 demonstrated the strongest association with nutritional risk. Its addition improved discrimination (AUROC: 0.850 to 0.862; NRI = 0.2273, IDI = 0.0236; P < 0.05). Higher GDF15 levels were associated with increased risk of hepatic decompensation (hazard ratio [HR] = 2.4). We developed and externally validated a laboratory-based approach for nutritional risk assessment in patients with cirrhosis. This approach integrates routinely available laboratory indicators with clinical parameters and may serve as a complementary approach to support nutritional risk stratification in clinical settings.