{
  "abstract": "Introduction Non-invasive assessment of liver fibrosis is essential for the management of chronic hepatitis B (CHB). Several simple biomarkers; such as bilirubin, alkaline phosphatase (ALP), the AST/ALT ratio (AAR), the GGT-to-platelet ratio (GPR), and the FIB-4 score, have been proposed to predict severe fibrosis (F3–F4). With the emergence of artificial intelligence, medical chatbots represent new tools that may assist data analysis. The aim of this study was to compare the performance of two chatbots against reference results calculated using Jamovi.Methods This was a retrospective, comparative, analytical study conducted over an 8-year period using a database of patients with chronic hepatitis B. Biological variables required to compute the biomarkers (bilirubin, ALP, AAR, GPR, FIB-4) were collected. The same standardized prompt was submitted to two chatbots (ChatGPT-5 and Gemini) to obtain, for each marker, the area under the ROC curve (AUROC), sensitivity, and specificity for detecting severe fibrosis (F3–F4). The chatbot outputs were then compared with reference analyses performed using Jamovi version 2.6.13.Results A total of 123 patients were included. The mean age was 46.7 ± 12.3 years, with a male predominance (M/F sex ratio = 1.15). The main comorbidities were smoking (12.2%), alcohol use (7.3%), diabetes (9.7%), and hypertension (10.5%). According to the METAVIR classification, 12.5% of patients were F0, 60% F1, 18.3% F2, 1.7% F3, and 7.5% F4.Based on Jamovi results, the AUROC for bilirubin and ALP was 0.856 and 0.71, respectively, with sensitivities of 87.5% and 66.6%, and specificities of 79.2% and 73%. The AAR showed a sensitivity of 100%, a specificity of 59%, and an AUROC of 0.79. The GPR demonstrated a sensitivity of 81.8% and a specificity of 91.1% (AUROC = 0.85). FIB-4 provided the best performance, with an AUROC of 0.913, sensitivity of 90.9%, and specificity of 91.1%.In comparison, ChatGPT-5 produced values very close to those obtained with Jamovi, notably for bilirubin (0.86 vs 0.856), ALP (0.72 vs 0.71), and FIB-4 (0.94 vs 0.913). Gemini generally underestimated AUROCs, particularly for bilirubin (0.57 vs 0.856) and FIB-4 (0.87 vs 0.913).Conclusions The diagnostic performance generated with Jamovi confirms the value of these biomarkers, particularly FIB-4 and GPR for identifying severe fibrosis in chronic hepatitis B. Among the two chatbots tested, ChatGPT-5 was closer to the reference values, suggesting better reliability for automated medical data analysis. However, based on our findings, human verification remains essential to validate and interpret these analyses to ensure reliability in clinical practice.Abstract P92 Table 1MarkerJamovi (AUROC)SensitivitySpecificityChatGPT-5 (AUROC)SensitivitySpecificityGemini 2.5 (AUROC)SensitivitySpecificityBilirubin0.85688%79%0.8688%79%0.5767%54%ALP0.7167%73%0.7275%70%0.6667%66%AAR0.79100%59%0.82100%71%0.7980%68%GPR0.8582%91%0.8788%74%0.8080%71%FIB-40.91391%91%0.9488%96%0.8783%78%",
  "authors": [
    {
      "affiliations": [
        "Medecine C Hepatogastroenterology Departement, Ibn Sina University Hospital, Faculty Of Medicine And Pharmacy of Rabat., Rabat, Morocco"
      ],
      "name": "Hamza Benazzouz"
    },
    {
      "affiliations": [
        "Medecine C Hepatogastroenterology Departement, Ibn Sina University Hospital, Faculty Of Medicine And Pharmacy of Rabat., Rabat, Morocco"
      ],
      "name": "Mohamed Borahma"
    },
    {
      "affiliations": [
        "Medecine C Hepatogastroenterology Departement, Ibn Sina University Hospital, Faculty Of Medicine And Pharmacy of Rabat., Rabat, Morocco"
      ],
      "name": "Fatimazahra Chabib"
    },
    {
      "affiliations": [
        "Medecine C Hepatogastroenterology Departement, Ibn Sina University Hospital, Faculty Of Medicine And Pharmacy of Rabat., Rabat, Morocco"
      ],
      "name": "Nawal Lagdali"
    },
    {
      "affiliations": [
        "Medecine C Hepatogastroenterology Departement, Ibn Sina University Hospital, Faculty Of Medicine And Pharmacy of Rabat., Rabat, Morocco"
      ],
      "name": "Fatimazahra Ajana"
    },
    {
      "affiliations": [
        "Medecine C Hepatogastroenterology Departement, Ibn Sina University Hospital, Faculty Of Medicine And Pharmacy of Rabat., Rabat, Morocco"
      ],
      "name": "Maryeme Kadiri"
    }
  ],
  "title": "P92 Comparative reliability of AI-assisted analysis for noninvasive detection of severe fibrosis in chronic Hepatitis B",
  "uid": "11375102-7716-53c0-8a76-820401022c7c"
}