Letter: the effects of rifaximin in hepatic encephalopathy

SIR, We thank Payer et al. for their comments on our article 2 describing the performance of MELD and organ failure (OF) scores to predict outcome in patients with cirrhosis and active variceal bleeding (VB). As they rightly point out, in patients with cirrhosis and OF, the number and severity of organs in failure are the better of predictors of outcome; hence SOFA and APACHE are preferred over MELD in these cases. Variceal bleeding is a special circumstance. Cardiovascular and respiratory compromise are common complications, and intensive care admission allows invasive monitoring and vasoactive medication to be given; airway protection by endotracheal intubation is often preferred prior to, or at the time of, gastroscopy. Nevertheless, as has been shown recently, secondary infections and second insults while in intensive care are common precipitants of multi-OF and subsequent mortality. In patients with VB such second insults may be aspiration pneumonia and bacterial translocation. The latter especially may be related to portal hypertension and severity of the underlying liver disease. This is more likely to be reflected in the MELD scores on admission rather than OF scores on day one in our cohort. To that end, we have re-analysed our data by subdividing into patients with or without more than one organ in failure on admission. In patients with two or more OFs the area under the receiver operating curve to predict hospital survival for MELD was 0.867 (95% CI 0.791–0.943), APACHE II 0.810 (0.719–0.901), and SOFA 0.791 (0.695–0.888), P > 0.05 for all comparisons. For patients with one or no organs in failure the AUROC curves were 0.679 (0.527–0.830) for MELD, 0.712 (0.544–0.880) for APACHE II, and 0.734 (0.597–0.870) for SOFA, P > 0.05 for all comparisons. Thus, we were not able to demonstrate a significant difference between the accuracy of the scoring systems but the sample sizes in these subgroups were 86 and 66, respectively, so we are not powered for this comparison in these subgroups. Nevertheless, in patients with a higher MELD score, it may be that microor macro-aspiration and endotoxaemia from bacterial translocation during the index bleeding episode may be more likely to precipitate secondary insults that increase mortality. Further data on how OF scores later during intensive care stay in VB patients predict outcome would be of interest.