ALADDIN: A Machine Learning Approach to Enhance the Prediction of Significant Fibrosis or Higher in Metabolic Dysfunction-Associated Steatotic Liver Disease.

Summary

Across 2,630 biopsy-characterized MASLD patients, the ALADDIN ensemble (with VCTE) improved external validation AUC to 0.791 for ≥F2 fibrosis and outperformed VCTE-only and established scores; a labs-only version (AUC 0.706) performed best among non-VCTE methods. Web calculators support biopsy-free selection of resmetirom-eligible patients.

Key Findings

• External validation AUC 0.791 (95% CI 0.764–0.819) for ALADDIN-F2-VCTE, superior to VCTE alone (0.745), FAST (0.710), and Agile-3 (0.740).
• ALADDIN-F2-Lab (no VCTE) achieved AUC 0.706, outperforming FIB-4 and other lab-based scores.
• Decision curve analysis and calibration favored ALADDIN models, supporting clinical utility.
• Public web calculators enable immediate clinical integration and reproducibility.

Clinical Implications

ALADDIN-F2-VCTE can refine referrals for treatment and reduce unnecessary biopsies; ALADDIN-F2-Lab offers a pragmatic alternative where VCTE access is limited.

Limitations

• Retrospective datasets and potential spectrum bias of referred populations
• Biopsy reference standard subject to sampling variability; generalizability beyond study centers needs further proof

Future Directions

Prospective impact studies to test ALADDIN-guided care pathways on biopsy rates and treatment outcomes; local recalibration and fairness assessments across demographics.