The efficacy and toxicity equilibrium of emodin for liver injury: A bidirectional meta-analysis and machine learning
Sihan Hu1, Xiaochuan Guo2, Lang Tu3
1Department of Integrated Traditional Chinese and Western Medicine, Peking University First Hospital; Institute of Integrated Traditional Chinese and Western Medicine, Peking University, Beijing, PR China; State Key Laboratory of Southwestern Chinese Medicine Resources, Chengdu University of Traditional Chinese Medicine, Chengdu, PR China.
Background:
Emodin, a hepatoprotective agent derived from various herbs, exhibits dual effects on liver injury, necessitating further investigation into its therapeutic and toxic properties. Traditional meta-analyses lack predictive capability for dose- and duration-dependent effects. This study uniquely employs meta-analysis to confirm both hepatoprotective and hepatotoxic effects of emodin and uses machine learning to predict critical thresholds where these effects invert.
Purpose:
We aimed to unravel the balance between emodin's hepatoprotective and hepatotoxic effects in rodent models, focusing on identifying dose- and duration-dependent responses. By dissecting emodin's efficacy and toxicity and elucidating the underlying mechanisms, our project contributes to developing a more rational dosing regimen and provides insights for the judicious and standardized use of traditional medicine in clinical pharmacology.
Methods And Materials:
A systematic review and meta-analysis, registered with INPLASY (202330123), were conducted to evaluate the bidirectional effects of emodin on liver injury. Relevant preclinical studies were searched in the Cochrane Library, PubMed, EMBASE, and Web of Science up until December 1, 2023. From an initial pool of 695 records, 28 pertinent rat and mouse studies were ultimately included. Data analysis for the meta-analysis was performed using STATA 17.0, while machine learning models were implemented in R 4.2.1 and Python 3.9 to assess the impact of intervention variables (dose and duration) on serum alanine aminotransferase (ALT) levels.
Results:
This meta-analysis incorporated 28 studies with 537 rodents, confirming emodin's dual effects on liver injury. Controlled doses and durations of emodin significantly reduced aspartate aminotransferase (AST) (SMD = -3.29, 95 % CI [-4.33, -2.25], p < 0.001), ALT (SMD = -2.65, 95 % CI [-3.44, -1.86], p < 0.001), and alkaline phosphatase (ALP) (SMD = -1.70, 95 % CI [-2.59, -0.80], p < 0.001) levels, primarily by inhibiting cytochrome P450 2E1 (CYP2E1) expression and activating the farnesoid X receptor/bile salt export pump (FXR/BSEP) pathway. Conversely, higher doses and prolonged durations were associated with increased hepatotoxicity, as indicated by a significant rise in AST (SMD = 2.19, 95 % CI [0.91, 3.47], p < 0.001) in healthy animals, with ALT (SMD = 0.59, 95 % CI [-0.18, 1.35], p > 0.05) and ALP (SMD = -0.35, 95 % CI [-1.00, 0.30], p > 0.05) levels showing no significant changes. Furthermore, machine learning targeting serum ALT levels suggests that a dosage exceeding 45.74 mg/kg/day or a duration beyond 30.41 days may represent the critical thresholds at which emodin transitions from hepatoprotective to hepatotoxic. This provides a more objective reference for minimizing the risk of hepatotoxicity while maximizing therapeutic efficacy.
Conclusions:
Emodin demonstrates significant potential in treating liver injury within specific therapeutic windows. The integration of meta-analysis with machine learning in this study not only confirms the bidirectional effects of emodin but also offers a framework for explaining preclinical intervention variables, thereby advancing its clinical applications in diseases.
