Discrimination of DILI and AIH by Artificial Intelligence

Development and Validation of a Machine Learning Model to Differentiate Drug-induced Liver Injury and Autoimmune Hepatitis

A retrospective, multi-center, non-interventional cohort study has been going to explore whether artificial intelligence can discriminate Drug-induced liver injury and Autoimmune hepatitis.

A machine learning-based tool will be developed and validated to help clinicians to differentiate between Drug-induced liver injury and Autoimmune hepatitis

Study Overview

• Status: Completed
• Conditions: Autoimmune Hepatitis; Drug-induced Liver Injury

Detailed Description

Research Objectives:

1. To develop a machine learning-based model from retrospective data.
2. To validate the machine learning-based model from internal dataset and external datasets nationwide.
3. To setup a website or application based on the above model to discriminate Drug-induced liver injury and Autoimmune hepatitis.

• Study Type: Observational
• Enrollment (Actual): 2583

Contacts and Locations

Study Locations

• China
  • Beijing, China, 100050
    • Beijing Friendship Hospital, Capital Medical University

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Child; Adult; Older Adult
• Accepts Healthy Volunteers: No

• Sampling Method: Non-Probability Sample

Study Population

Hospitalized patients with a definite diagnosis of DILI and AIH were recruited from all hospitals between January 2009 to December 2025. This study was approved by the Institutional Ethical Review Board (2022-P2-263-01).

Description

Inclusion Criteria:

• Drug-induced liver injury

  1. RUCAM ≥6 and met one of the following biochemical conditions: a.) ALT≥5 ULN, b.) or ALP ≥2 ULN, iii) or ALT≥3 ULN and TBil≥2 ULN.
  2. RUCAM was between 3-5, the medical records were further reviewed by the three authors to determine the eligibility.

• Autoimmune hepatitis

  1. The revised International Autoimmune Hepatitis Group (IAIHG) diagnostic score≥6 points.
  2. Liver biopsy available, which is compatible with typical features of AIH.
  3. If liver biopsy was unavailable, patients who achieved biochemical resolution after sustained immunosuppressive therapy.

Exclusion Criteria:

• Drug-induced liver injury

  1. Hepatotropic viral infection: hepatitis A, B, C, D and E.
  2. Non-hepatotropic viral infection: cytomegalovirus (CMV) and Epstein-Barr virus (EBV), etc.
  3. Hypoxic ischemic hepatitis and congestive liver disease.
  4. Alcohol consumption: male >40g/d, female >20g/d, and ≥5 years.
  5. Biliary obstruction, primary biliary cholangitis; primary sclerosing cholangitis.
  6. Autoimmune hepatitis.
  7. Parasitic infection.
  8. Sepsis.
  9. Previous liver transplantation or bone marrow transplantation.
  10. Pregnancy or lactation.
  11. Genetic and metabolic liver diseases.

• Autoimmune hepatitis

  1. Hepatotropic viral infection: hepatitis A, B, C, D and E.
  2. Non-hepatotropic viral infection: cytomegalovirus (CMV) and Epstein-Barr virus (EBV), etc.
  3. Hypoxic ischemic hepatitis and congestive liver disease.
  4. Alcohol consumption: male >40g/d, female >20g/d, and ≥5 years.
  5. Biliary obstruction, primary biliary cholangitis; primary sclerosing cholangitis.
  6. Drug-induced liver injury.
  7. Parasitic infection.
  8. Sepsis.
  9. Previous liver transplantation or bone marrow transplantation.
  10. Pregnancy or lactation.
  11. Genetic and metabolic liver diseases.

Study Plan

How is the study designed?

Design Details

• Observational Models: Other
• Time Perspectives: Retrospective

Number of groups / cohorts

3

Cohorts and Interventions

Group / Cohort
Development group, retrospectively collecting; The training set includes 80% of the development dataset from Beijing Friendship Hospital (Retrospectively collecting DILI 649 cases and AIH 180 cases in total)
Validation group, retrospectively collecting; The internal validation group includes 20% of the development dataset from Beijing Friendship Hospital (Retrospectively collecting DILI 649 cases and AIH 180 cases in total). External validation groups include Fifth Medical Center of Chinese PLA Medical Center(Retrospectively collecting DILI 585 cases and AIH 400 cases in total), Beijing You'an Hospital (Retrospectively collecting DILI 266 cases and AIH 100 cases in total) and additional seven tertiary hospitals throughout China (Retrospectively collecting DILI 182 cases and AIH 92 cases in total, including Tianjin Second People's Hospital, Heilongjiang Provincial Hospital, Affiliated Hospital of Qingdao University, the First Affiliated Hospital of Xiamen University, Traditional Chinese Medical Hospital of Xinjiang Uygur Autonomous Region, Lanzhou University Second Hospital, Qinghai Provincial People's Hospital).
Real-World validation group, prospectively and retrospectively collecting; Real-World validation data will be prospectively and retrospectively collected from Beijing Friendship Hospital (78 cases for DILI and 51 cases for AIH)

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Accuracy of the model in the differential diagnosis of DILI and AIH	The ratio of the correct number of forecasts to the total number of forecasts	May 31, 2023
The confidence of the model in the differential diagnosis of DILI and AIH	The confidence and 95% confidence internal of the model in determining whether each case is DILI or AIH	May 31, 2023

Collaborators and Investigators

• Sponsor: Beijing Friendship Hospital
• Investigators: Study Chair: Xinyan Zhao, Dr., Beijing Friendship Hospital; Principal Investigator: Ying Sun, Dr., The Fifth Medical Department of the PLA; Principal Investigator: Jing Zhang, Dr., Capital Medical University Affiliated Beijing You'an Hospital; Principal Investigator: Jia Li, Dr., Tianjin Second People's Hospital; Principal Investigator: Liang Wang, Dr., Lanzhou University Affiliated Second Hospital; Principal Investigator: Jingshou Chen, Dr., The First Affiliated Hospital of Xiamen University; Principal Investigator: Feng Guo, Dr., Xinjiang Uygur Autonomous Region Hospital of Traditional Chinese Medicine; Principal Investigator: Pingying Li, Dr., Qinghai People's Hospital; Principal Investigator: Qiuju Tian, Dr., The Affiliated Hospital Of Qingdao University; Principal Investigator: Xiaoli Hu, Dr., Heilongjiang Provicial Hospital

Publications and helpful links

General Publications

• Wang Y, Lin X, Sun Y, Liu J, Li J, Tian Q, Guo F, Hu X, Wang L, Li P, Chen J, Wang Y, Ma Z, Jia J, Zhang J, Zou Z, Zhao X. Development and Validation of a Novel Model to Discriminate Idiosyncratic Drug-Induced Liver Injury and Autoimmune Hepatitis. Liver Int. 2025 Feb;45(2):e16239. doi: 10.1111/liv.16239.

Study record dates

Study Major Dates

• Study Start (Actual): July 1, 2022
• Primary Completion (Actual): May 31, 2023
• Study Completion (Actual): May 31, 2023

Study Registration Dates

• First Submitted: August 29, 2022
• First Submitted That Met QC Criteria: September 4, 2022
• First Posted (Actual): September 8, 2022

Study Record Updates

• Last Update Posted (Actual): July 6, 2026
• Last Update Submitted That Met QC Criteria: July 2, 2026
• Last Verified: July 1, 2023

More Information

Terms related to this study

• Keywords: Artificial intelligence; Differential diagnose
• Additional Relevant MeSH Terms: Autoimmune Diseases; Immune System Diseases; Digestive System Diseases; Liver Diseases; Chemically-Induced Disorders; Drug-Related Side Effects and Adverse Reactions; Poisoning; Hepatitis, Chronic; Hepatitis; Chemical and Drug Induced Liver Injury; Hepatitis, Autoimmune
• Other Study ID Numbers: 2022-P2-263-01

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No