Introduction Outcome prediction is a cornerstone of the management of Acute Liver Failure (ALF) where Emergency Liver Transplantation (ELT) is indicated for predicted death. The King's College Criteria (KCC) for paracetamol overdose (POD) and non-POD ALF are the benchmark prognostic scores but recent reports have suggested the Modified End Stage Liver Disease (MELD) score could replace KCC.

Aim To meta-analyse and compare diagnostic performance for outcome prediction of KCC and MELD in ALF.

Method A systematic database search was performed and retrieved articles graded according to a pre-agreed proforma of methodological quality. Collated data were meta-analysed for summary sensitivity, specificity, Diagnostic OR, DOR meta-regression and ROC curve analysis. Pre-specified subgroup analysis was performed on the basis of methodological quality, the severity of hepatic encephalopathy (HE) of reported patients, and exclusion of those who underwent ELT.

Results 32 studies published between 1992 and 2009 with data on 3008 patients (2464 from KCC and 544 MELD) were available for production of 2×2 tables. Taking data where transplanted patients were excluded summary sensitivity for KCC was 63 (95% CI 60 to 66) %, specificity 91 (90 to 93) % and DOR 18 (8.9 to 37). Summary sensitivity for MELD was 82 (95% CI 77 to 86) %, specificity 65 (59to71) % and DOR 11 (5.3 to 23; RDOR 0.75 (0.18 to 3.13))). Despite different MELD cut-offs between studies no statistical evidence of threshold was found and the AUROC for MELD was 0.83 (SE 0.05) and 0.88 (0.03) for KCC. A lack of patient level data prevented statistical comparison between these areas. The DOR for KCC in POD ALF was 27 (9–83) and 13 (5–31) in non-POD ALF. Heterogeneity (using the I² statistic) in the DOR for MELD was 49% and 80% for KCC although this was not dependent on aetiology. The lower sensitivity for KCC could be overcome by dynamic application of the criteria.

Conclusion MELD is not superior to KCC on the basis of quantitative analysis of published evidence. While both scores may give complementary sensitivity and specificity (particularly in cases where bilirubin may have more prognostic value) the DOR and AUROC superiority of KCC suggest they should remain the preferred method of outcome prediction and listing decision for ELT in ALF, particularly in cases of POD-ALF.