Prognostic benefit of the addition of a quantitative index of hepatic encephalopathy to the MELD score: the MELD-EEG.

AbstractBackground & Aims: A slowed electroencephalogram (EEG) is indicative ofthe presence of hepatic encephalopathy (HE). Since HE is not reflected in theMELD score and is an important prognostic parameter, we assess the prog-nostic benefit of the addition of an EEG-based HE index to theMELD. Methods: Three hundred and ninety-two patients with cirrhosisunderwent EEG and automated determination of its mean dominant fre-quency (MDF). MELD was calculated at the time of EEG recording. Patientswere monitored for up to 18 months in relation to the occurrence of death/transplantation. The prognostic value of the stand-alone/combined MELDand MDF was calculated using standard survival analysis techniques. Patientstransplanted for hepatic decompensation were considered dead on the day oftransplantation, those transplanted for hepatocellular carcinoma were cen-sored. The findings were validated using a split-sample technique (referencegroup: n = 256; test group: n = 136). During the follow-up period, 107patients died/were transplanted for hepatic decompensation. Results: BothMELD and MDF predicted mortality on Kaplan–Meier analysis, and bothwere independent predictors of mortality on a Cox model. Based on Coxregression parameters, a novel prognostic index was devised, as follows:MELD-EEG = 0.087*MELD–0.306*MDF. On ROC curve analysis, MELD-EEG had higher prognostic accuracy in predicting 12- and 18-month mortal-ity compared to MELD (P = 0.016 and P = 0.018, respectively). In addition,it had better sensitivity and reduced the misclassification rate for given levelsof specificity. On validation, no significant differences were observed betweenthe reference/test groups. Conclusions: The addition of an automaticallyobtained EEG-based index improves the prognostic accuracy of the MELDscore.Hepatic encephalopathy (HE) is not part of the MELDscore, which is widely used to assess the severity of hepa-tic failure and to estimate the need/timing for hepatictransplantation (1, 2). This is mostly related to difficul-ties and interoperator variability in the clinical diagnosisof HE, and its grading (3). In addition, while part of theChild-Pugh score (4), overt HE did not seem to signifi-cantly improve the prognostic accuracy of MELD whenthis was originally devised (1). However, there is evi-dence that HE is a good prognostic marker of survivalin patients with cirrhosis (5–7). HE has also been shownto carry prognostic information which is additive to thatof MELD, and it has been suggested that patients withHE may not receive a transplant in a timely fashion ifMELD is used as a stand-alone (8). In recent years, toolshave become available to quantify HE across its spec-trum of severity in a non-operator dependent, docu-mentable fashion, thus providing ‘MELD-like’ summaryHE indices. These tools include the automated assess-ment of the electroencephalographic (EEG) slowingwhich characterizes HE (Fig. 1) (9, 10). This has beenshown to have a good relationship with laboratorymarkers of HE (11), a history of HE/its developmentover time (12), and survival, especially in patients withmoderate/severe hepatic decompensation (6, 8).The prognostic efficiency of MELD can probably beimproved, especially in patients with non-fulminanthepatic failure (13, 14). A number of proposals have