Ability of King's College Criteria and Model for End-Stage Liver Disease Scores to Predict Mortality of Patients With Acute Liver Failure: A Meta-analysis
TL;DR: Based on a meta-analysis of studies, the KCC more accurately predicts hospital mortality among patients with AALF, whereas MELD scores more accurately predict mortalityamong patients with NAALF.
Abstract: Background & Aims Several prognostic factors are used to identify patients with acute liver failure (ALF) who require emergency liver transplantation. We performed a meta-analysis to determine the accuracy of King's College criteria (KCC) versus the model for end-stage liver disease (MELD) scores in predicting hospital mortality among patients with ALF. Methods We performed a systematic search of the literature for articles published from 2001 through 2015 that compared the accuracy of the KCC with MELD scores in predicting hospital mortality in patients with ALF. We identified 23 studies (comprising 2153 patients) and assessed the quality of data, and then performed a meta-analysis of pooled sensitivity and specificity values, diagnostic odds ratios (DORs), and summary receiver operating characteristic curves. Subgroups analyzed included study quality, era, location (Europe vs non-Europe), and size; ALF etiology (acetaminophen-associated ALF [AALF] vs nonassociated [NAALF]); and whether or not the study included patients who underwent liver transplantation and if the study center was also a transplant center. Results The DOR for the KCC was 5.3 (95% confidence interval [CI], 3.7–7.6; 57% heterogeneity) and the DOR for MELD score was 7.0 (95% CI, 5.1–9.7; 48% heterogeneity), so the MELD score and KCC are comparable in overall accuracy. The summary area under the receiver operating characteristic curve values was 0.76 for the KCC and 0.78 for MELD scores. The KCC identified patients with AALF who died with 58% sensitivity (95% CI, 51%–65%) and 89% specificity (95% CI, 85%–93%), whereas MELD scores identified patients with AALF who died with 80% sensitivity (95% CI, 74%–86%) and 53% specificity (95% CI, 47%–59%). The KCC predicted hospital mortality in patients with NAALF with 58% sensitivity (95% CI, 54%–63%) and 74% specificity (95% CI, 69%–78%), whereas MELD scores predicted hospital mortality in patients with NAALF with 76% sensitivity (95% CI, 72%–80%) and 73% specificity (95% CI, 69%–78%). In patients with AALF, the KCC's DOR was 10.4 (95% CI, 4.9–22.1) and the MELD score's DOR was 6.6 (95% CI, 2.1–20.2). In patients with NAALF, the KCC's DOR was 4.16 (95% CI, 2.34–7.40) and the MELD score's DOR was 8.42 (95% CI, 5.98–11.88). Conclusions Based on a meta-analysis of studies, the KCC more accurately predicts hospital mortality among patients with AALF, whereas MELD scores more accurately predict mortality among patients with NAALF. However, there is significant heterogeneity among studies and neither system is optimal for all patients. Given the importance of specificity in decision making for listing for emergency liver transplantation, MELD scores should not replace the KCC in predicting hospital mortality of patients with AALF, but could have a role for NAALF.
Summary (2 min read)
- Acute Liver Failure (ALF) is a rare, but devastating illness with a high risk of progression to multi-organ failure and death1-3.
- The key clinical issue remains to accurately identify patients with ALF who will die without ELT, and those who will survive with medical management alone.
- One particularly salient difference is the treatment of transplanted patients.
- To date, there have been three meta-analyses23 of the performance of the KCC in ALF.
- The first included only Acetaminophen-induced ALF (AALF) identifying nine studies in total, and concluded that the KCC had limited sensitivity13.
- All potential articles were assessed independently by two researchers (HF, MM) according to prospectively defined eligibility criteria, and disagreements were resolved by consensus or consultation with a third author (WB).
- If this was not possible or there was doubt over the 2 x 2 calculation, the study was excluded from the subsequent analysis.
- The DerSimonian-Laird random effects method was used to produce summary estimates of sensitivity, specificity, likelihood ratios (LR) and diagnostic odds ratio (DOR, defined as the ratio of positive to negative likelihood ratios).
- A funnel plot and effective sample size (ESS) regression analysis (the logarithm of the DOR plotted against 1/√ESS) was used to investigate publication bias.
- Data analyses were performed using the freeware Meta-Disc version 1.4 (Universidad Complutense, Madrid, Spain) and Eggers statistic calculated in Excel (Microsoft Corporation, Redmond WA)33.
- The search strategy identified 4,063 potentially relevant studies.
- Subgroup analysis was performed to assess differences in heterogeneity and diagnostic accuracy between the groups specified earlier.
- Furthermore Egger's statistic was not significant again suggesting publication bias was not present.
- This meta-analysis confirms that when comparing KCC and MELD for outcome prediction in ALF KCC have lower sensitivity and MELD lower specificity.
- The sROC analysis is therefore a more valid way to pool the results of studies with varying thresholds.
- This is no doubt a consequence of the fact that the KCC were derived from an ALF cohort, whereas MELD was developed from results in chronic liver disease patients undergoing TIPS.
- This may be why KCC is preferred in countries facing such organ shortages and with high rates of AALF.
- Clearly such delays are relatively short but in cases of fulminant hepatic failure it is clearly advantageous to use simpler bedside tests during the evolution of disease.
- Information on prothrombin time measurements and assay details were not available in all studies and may have contributed to heterogeneity or threshold effects.
- The potential benefits of combining the specificity of the KCC with the sensitivity of MELD are attractive.
- Such novel methods would require data for each patient rather than summative as presented for publication.
- Many new biomarkers have been proposed in ALF but have failed to be validated in larger studies or are deemed not ready for widespread distribution.
- Neither KCC nor MELD are optimal in all circumstances so there remains an urgent need for more accurate outcome prediction systems in ALF.
Did you find this useful? Give us your feedback
Cites background from "Ability of King's College Criteria ..."
...In patients with DILI who developed ALF, the King’s College criteria or the US ALF Study Group criteria for non-APAP ALF can be applied for assessing the prognosis and for timing liver transplant evaluation, but these models are not specific for DILI (73,74)....
Cites background from "Ability of King's College Criteria ..."
...A recent metaanalysis has revealed its prognostic ability in comparison with the MELD score: for acetaminophen-related ALF, sensitivities were 58% and 80%, respectively, and specificities were 89% and 53%, respectively; for non-acetaminophen etiologies, sensitivities were 58% and 76%, respectively, and specificities were 74% and 73%, respectively ....
Related Papers (5)
Frequently Asked Questions (2)
Q1. What have the authors contributed in "Meta-analysis of king's college criteria and model for end stage liver disease to predict outcome in acute liver failure" ?
The authors assessed the accuracy of King 's College Criteria ( KCC ) versus the Model-forEnd-Stage-Liver-Disease ( MELD ) in ALF through meta-analysis of studies which report the accuracy of both tests.
Q2. What have the authors stated for future works in "Meta-analysis of king's college criteria and model for end stage liver disease to predict outcome in acute liver failure" ?
The authors hope these data help inform such decisions and future research. A worsening grade of HE can be detected at the bedside and incorporated into KCC without awaiting further biochemical analysis.