Meta-analysis of performance of Kings's College Hospital Criteria in prediction of outcome in non-paracetamol-induced acute liver failure

Background & aims: Current techniques for predicting outcome and requirement for emergency liver transplantation (ELT) in acute liver failure (ALF) are imperfect, though The Kings College Criteria (KCC) are the most commonly applied tools for this purpose. Their performance in identification of patients with non-paracetamol-induced ALF (non-POD ALF), who would not survive without ELT, has recently been questioned. Using quantitative techniques, we therefore performed a meta-analysis of outcome data of the KCC for prediction of survival in non-POD ALF.

Methods: A systematic database search was performed and retrieved articles graded according to a pre-agreed pro-forma of methodological quality. Collated data was meta-analysed for summary sensitivity, specificity, diagnostic odds ratio (DOR) and ROC curve analysis. Pre-specified sub-group analysis was performed on the basis of methodological quality, the severity of hepatic encephalopathy (HE) of reported patients, timing of KCC application and exclusion of those who underwent ELT.

Results: Eighteen studies with data on 1105 patients were available for production of 2x2 tables. Summary sensitivity was 68 (95% CI 59-77)%, specificity 82 (75-88)% and DOR 12.6 (6.5-26.1). Heterogeneity was detected in the DOR and related to methodological quality (I(2)=64% for all studies versus 47% for 'good' quality studies) and was lower in studies considering high grade HE or dynamic application of KCC (I(2)=0%). For data where ELT were excluded (13 studies) summary sensitivity was 68 (57-79)%, specificity 81 (72-90)% and DOR 12.2 (4.9-30.1) and a symmetric summary ROC curve was produced. Specificity was highest in studies of patients with high grade HE (93 (80-100)%) and where KCC were applied dynamically through the clinical course (88 (78-97)%). Sensitivity was reduced in studies published post 2005 compared with studies pre 1995 (58 (46-71)% versus 85 (76-82)%).

Conclusions: KCC for outcome in non-POD ALF have good specificity and more limited sensitivity. There is significant heterogeneity in the published data partially related to methodological quality. KCC perform best in groups with high grade encephalopathy and in historically earlier studies suggesting modern medical management of ALF may modify performance of KCC.