To systematically evaluate the performance of prognostic prediction models for mortality in patients with acute-on-chronic liver failure (ACLF), emphasizing their clinical applicability.
Key Findings:
185 studies included, covering 241 external validation cohorts and 75 distinct prognostic models, highlighting the implications of the high risk of bias.
99.51% of analysis units had a high risk of bias due to insufficient outcome events and inappropriate performance assessment.
24 models met criteria for meta-analysis, with c-statistics ranging from 0.58 to 0.84.
ACLF-specific models showed better discrimination than general models, with CLIF-C ACLF demonstrating stability across subgroups.
Interpretation:
Current evidence supports relative comparisons of model discrimination but is insufficient for precise clinical decision-making, highlighting the limitations of existing models.
Limitations:
High risk of bias in external validation studies, with specific examples of how the lack of calibration affects model use.
Lack of calibration reports in most studies.
Performance of models declined with longer prediction horizons.
Conclusion:
Further validation of ACLF prognostic models is needed through high-quality multicenter studies focusing on calibration and long-term predictive performance, specifically addressing identified gaps.
