# Analysis of risk factors for herb-induced liver injury: a retrospective study from China

## TL;DR

This study identifies risk factors for liver injury caused by herbal medicines in China, using clinical data from thousands of patients.

## Contribution

The study introduces a predictive model for herb-induced liver injury using clinical and laboratory variables.

## Key findings

- ALP is an independent risk factor distinct from WILI in herb-induced liver injury.
- Age, alcohol, urea, platelet distribution width, and monocyte ratio are significant predictors.
- The model shows strong performance in AUC and DCA metrics across training and validation sets.

## Abstract

To analyse the clinical and laboratory characteristics of patients experiencing liver injury following herbal medication use, and to identify herb-induced liver injury in patients with risk factors.

This retrospective study included patients admitted to Nanjing Hospital of Chinese Medicine affiliated to Nanjing University of Chinese Medicine, between December 2014 and December 2023 who developed liver injury following herbal medicine consumption. The cohort was divided into a training set and an validation set. Patients admitted to Jiangsu Province Hospital of Chinese Medicine between January 2024 and March 2025 who developed liver injury following herbal medicine consumption were included as the test set. Predictor variables were screened using univariate analysis and binary logistic regression, and predictive models were constructed for risk factors in patients with herb-induced liver injury versus non-drug-induced liver injury. Evaluate model performance using multiple assessment metrics, including AUC and DCA.

The analysis included 3,914 patients with abnormal liver function who had a history of herbal medicine use, of whom 176 were assessed as having herb-induced liver injury through the RUCAM causality assessment. Research has found that ALP is an independent risk factor distinct from WILI. Incorporate risk factors for both the HILI group and NON-DILI group: age, alcohol, urea, platelet distribution width, and monocyte ratio. Construct a predictive model and evaluate its performance. The model demonstrates favourable performance in terms of AUC and DCA across both the training and validation sets.

Compared with patients suffering from non-drug-induced liver injury, those of advanced age, with concomitant hepatic metabolic dysfunction, or underlying immune disorders exhibit a significantly higher risk of developing herb-induced liver injury.