Comparison of dynamic changes among various prognostic scores in viral hepatitis-related acute liver failure

Introduction and aim. Multiple prognostic scores are available for acute liver failure (ALF). Our objective was to compare the dynamicity of model for end stage liver disease (MELD), MELD-sodium, acute liver failure early dynamic model (ALFED), chronic liver failure (CLIF)-consortium ACLF score and King’s College Hospital Criteria (KCH) for predicting outcome in ALF. Materials and methods. All consecutive patients with ALF at a tertiary care centre in India were included. MELD, MELD-Na, ALFED, CLIF-C ACLF scores and KCH criteria were calculated at admission and day 3 of admission. Area under receiver operator characteristic curves (AUROC) were compared with DeLong method. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), likelihood ratio (LR) and diagnostic accuracy (DA) were reported. Results. of the 115 patients included in the study, 73 (63.5%) died. The discrimination of mortality with baseline values of prognostic scores (MELD, MELD-Na, ALFED, CLIF-C ACLF and KCH) was modest (AUROC: 0.65-0.77). The AUROC increased on day 3 for all scores, except KCH criteria. On day 3 of admission, ALFED score had the highest AUROC 0.95, followed by CLIF-C ACLF 0.88, MELD 0.81, MELD-Na 0.77 and KCH 0.52. The AUROC for ALFED was significantly higher than MELD, MELD-Na and KCH (P < 0.001 for all) and CLIF-C ACLF (P = 0.05). ALFED score ≥ 4 on day 3 had the best sensitivity (87.1%), specificity (89.5%), PPV (93.8%), NPV (79.1%), LR positive (8.3) and DA (87.9%) for predicting mortality. Conclusions. Dynamic assessment of prognostic scores better predicts outcome. ALFED model performs better than MELD, MELD, MELD-Na, CLIF-C ACLF scores and KCH criteria for predicting outcome in viral hepatitis-related ALF.