{
  "abstract": "Introduction Primary biliary cholangitis (PBC) is a rare autoimmune cholestatic liver disease with a heterogeneous clinical course. Long-term prognosis is commonly assessed using validated scores such as the GLOBE and UK-PBC scores. However, these tools rely on predefined models and may not fully capture the complexity of multidimensional clinical data. Artificial intelligence (AI) offers an opportunity to integrate heterogeneous baseline variables for improved prognostic stratification. This study aimed to develop an AI-based prognostic model in PBC and to compare its performance with established scores in a real-world setting.Methods We conducted a retrospective single-centre study including 142 patients with PBC followed between 2010 and 2025. Baseline data included clinical features, biochemical parameters, immunological markers, and histological findings. Prognosis was assessed using the GLOBE and UK-PBC scores, which served as benchmarks for AI model evaluation. Categorical variables were encoded, numerical variables normalized, and missing data handled. Three supervised models (logistic regression, penalized logistic regression (LASSO), and Random Forest) were developed. Models were internally validated and compared using ROC curve analysis, area under the curve (AUC), sensitivity, specificity, precision, accuracy, and calibration.Results The mean age at diagnosis was 49.8 ± 13.3 years, with a strong female predominance (90%). Clinically, jaundice was present in 52%, pruritus in 40%, and signs of portal hypertension in 45% of patients. Fibrosis assessment was available in 104 patients, of whom 54.8% had advanced fibrosis (≥F3). Among 57 patients with available 10-year prognostic scores, 22 (38.6%) were classified as high risk according to GLOBE and/or UK-PBC scores.The Random Forest model demonstrated the best prognostic performance, with an AUC of 0.83 (95% CI: 0.69–0.94), sensitivity of 63.6%, specificity of 88.6%, precision of 77.8%, and overall accuracy of 78.9%. Logistic regression and LASSO models showed lower discrimination, with AUCs of 0.68 and 0.69, respectively. The most influential baseline predictors of adverse prognosis included age, cholestatic biomarkers, advanced ­fibrosis, and markers of portal hypertension. Calibration analysis indicated good agreement between predicted probabilities and observed risk categories.Conclusion In this real-world Moroccan cohort of PBC, AI-based models enabled meaningful prognostic stratification using baseline multidimensional data. This approach may ­facilitate earlier identification of high-risk patient profiles and support personalized long-term monitoring and management strategies.",
  "authors": [
    {
      "affiliations": [
        "Medicine Department C, Ibn Sina University Hospital Center. Mohammed V University, Rabat, Morocco"
      ],
      "name": "Jaouhara Lahsini"
    },
    {
      "affiliations": [
        "Medicine Department C, Ibn Sina University Hospital Center. Mohammed V University, Rabat, Morocco"
      ],
      "name": "Mohamed Borahma"
    },
    {
      "affiliations": [
        "Medicine Department C, Ibn Sina University Hospital Center. Mohammed V University, Rabat, Morocco"
      ],
      "name": "Fatimazahra Chabib"
    },
    {
      "affiliations": [
        "Medicine Department C, Ibn Sina University Hospital Center. Mohammed V University, Rabat, Morocco"
      ],
      "name": "Nawal Lagdali"
    },
    {
      "affiliations": [
        "Medicine Department C, Ibn Sina University Hospital Center. Mohammed V University, Rabat, Morocco"
      ],
      "name": "Selma Sabbah"
    },
    {
      "affiliations": [
        "Medicine Department C, Ibn Sina University Hospital Center. Mohammed V University, Rabat, Morocco"
      ],
      "name": "Fatima Zohra Ajana"
    },
    {
      "affiliations": [
        "Medicine Department C, Ibn Sina University Hospital Center. Mohammed V University, Rabat, Morocco"
      ],
      "name": "Maryeme Kadiri"
    }
  ],
  "title": "P114 Artificial intelligence for prognostic stratification in primary biliary cholangitis: a moroccan single centre experience",
  "uid": "85799dd0-fe43-57e1-ada0-2bb20d560ebf"
}