Early Precise Prediction and Severe Risk Stratification of Intrahepatic Cholestasis of Pregnancy: Advances and Future
Wenting Xu1, Minghe Wang2, Xiang Li1
1Department of Gastroenterology, Department of Obstetrics and Gynecology, Guangdong Provincial Key Laboratory of Major Obstetric Diseases, Guangdong Provincial Clinical Research Center for Obstetrics and Gynecology, Guangdong-Hong Kong-Macao Greater Bay Area Higher Education Joint Laboratory of Maternal-Fetal Medicine, The Third Affliated Hospital, Guangzhou Medical University, Guangzhou 510150, China.
Insights
Intrahepatic cholestasis of pregnancy (ICP) is a serious liver disorder in pregnancy. Advanced AI and multi-omics can improve early prediction and management of ICP, enhancing maternal and fetal health.
Area of Science:
- Obstetrics and Gynecology
- Hepatology
- Genetics and Genomics
- Artificial Intelligence in Medicine
Background:
- Intrahepatic cholestasis of pregnancy (ICP) is a pregnancy-specific liver disorder linked to adverse maternal and fetal outcomes, including preterm birth and fetal death.
- Risk factors include genetics, hormones, metabolism, environment, geographical location (e.g., southern China), and co-infection with hepatitis B virus.
- Elevated serum total bile acid (TBA) levels (≥ 40 μmol/L) are associated with severe ICP, but current prediction models lack generalizability due to small sample sizes and limited validation.
Purpose of the Study:
- To review the current state of prediction models for intrahepatic cholestasis of pregnancy (ICP).
- To highlight the potential of advanced computational methods, including nomograms, machine learning, and deep learning, for early ICP prediction and risk stratification.
- To identify challenges and propose future research directions for developing robust and clinically applicable ICP prediction tools.
Main Methods:
- Review of existing literature on ICP pathogenesis, risk factors, and prediction models.
- Analysis of the application of nomograms, machine learning (ML), and deep learning (DL) in ICP risk stratification.
- Discussion of multi-omics integration (genomics, metabolomics, gut microbiome) combined with AI for enhanced predictive accuracy.
Main Results:
- Nomograms and ML methods show promise for early ICP prediction and risk stratification.
- Deep learning and multi-omics integration further improve predictive performance.
- Existing models often suffer from limitations such as small sample sizes, retrospective design, and insufficient external validation.
Conclusions:
- Advanced AI, particularly deep learning, combined with multi-omics data offers significant potential for improving ICP prediction accuracy.
- Future research should focus on large-scale, multi-center prospective studies integrating diverse data types (genomics, metabolomics, microbiome) and AI.
- Developing clinically interpretable and generalizable risk prediction tools integrated into healthcare systems is crucial for early ICP identification and management to improve maternal and neonatal outcomes.
Abstract:
Intrahepatic cholestasis of pregnancy (ICP) is a liver disorder unique to pregnancy, closely associated with severe adverse maternal and fetal outcomes such as preterm birth and intrauterine fetal death. Its pathogenesis involves a complex interplay of genetic, hormonal, metabolic, and environmental factors, with a higher risk observed in southern China and among individuals co-infected with hepatitis B virus. Substantial evidence demonstrates a significant dose-response relationship between serum total bile acid (TBA) levels and perinatal outcomes, with TBA ≥ 40 μmol/L commonly used as a criterion for severe ICP. However, most existing prediction models are based on single-center, retrospective studies with small sample sizes and insufficient external validation, limiting their clinical generalizability. In recent years, nomograms and machine learning methods have demonstrated advantages in the early prediction and risk stratification of ICP. Deep learning models and multi-omics integration strategies have further enhanced predictive accuracy. Nevertheless, challenges remain regarding model interpretability, data standardization, and cross-population applicability. Future research should leverage large-scale, multi-center prospective cohorts, integrating multi-omics technologies, including genomics, metabolomics, gut microbiome profiling with artificial intelligence to develop clinically actionable and interpretable risk prediction tools. Integrating these tools into electronic health record systems and mobile platforms may facilitate the early identification and individualized management of ICP, ultimately improving maternal and neonatal outcomes.