Multiomics-based risk prediction tool for early identification of AIH patients

Background and Aims: Autoimmune hepatitis (AIH) is a chronic liver disease characterised by an autoimmune response against hepatocytes. Early diagnosis is crucial in preventing disease progression. However, it remains challenging due to the limited availability of appropriate diagnostics, such as liver biopsy. In this study, we aim to develop a risk prediction tool based on widely available biomarkers to identify patients-at-risk populations, that could benefit from a more elaborated or AIH-focused diagnostic approach.

Methods: Using the 319 participants within the UKBiobank, that were diagnosed with an autoimmune hepatitis at least one year after the comprehensive baseline examination, we built a predictive risk estimation model based on the coded ICD-10 diagnoses within the National Health Service (NHS) hospital records. We split the dataset into a training (80%) and testing set (20%). The training was performed in 5 fold cross validation to allow hyperparameter tuning and the 5 trained models were further combined into one mean voting algorithm.

Results: The evaluation of the predictive performance on the test-set showed that, depending on the time of readout, we can archive areas under the receiver operating characteristics curve for the binary classification of up to 0.8, with AST, sex, GGT and CRP being upon the most important features. The predictive performance decreases over time after the peak at around 6 years hinting at the limited prognostic horizon ([Fig. 1]).

Fig. 1  

Conclusion: In this study, we developed an AIH risk prediction tool based on widely available biomarkers, with potential for clinical application to identify at-risk patients who may benefit from referral to academic centers.

Informationen zum Einsatz von KI: Bei der Erstellung dieses Abstracts wurde unterstützend künstliche Intelligenz (KI) eingesetzt.

Publication History

Article published online:
04 September 2025