Predictive Ability of Different Scoring Models in Severe Alcoholic Hepatitis

Abstract

ABSTRACT Introduction: Alcoholic hepatitis is a major health problem which carries a high mortality despite treatment. Predicting severity of disease is of utmost importance to avoid overzealous and unnecessary treatment with steroids in each and every case. Maddrey’s discriminant function (DF) may not always be reliable in predicting such prognosis. Aim of the Study: To observe the predictive ability of Glasgow alcoholic hepatitis score (GAHS), age, bilirubin, international normalized ratio (INR) and creatinine (ABIC) score, model for end stage liver disease (MELD) score, and Lille’s score in determining the prognosis in severe alcoholic hepatitis. Materials and Methods: Patients of severe alcoholic hepatitis were enrolled in this study as per inclusion and exclusion criteria and a DF score of >32. Once the decision to initiate corticosteroids was made, DF scores, MELD scores, GAHS scores, and ABIC scores were calculated on the day on initiation of therapy. On day 7, patients were re-assessed and Lille score was calculated in addition. Follow-up was obtained on repeat out-patient visit or telephonically after every 2 weeks till 90 days or till mortality, whichever occurred first. Statistical Analysis: The scores were compared among the survival and mortality groups to see association between mortality and severity of scores using Chi-square test of significance. Cut-offs providing a composite of best sensitivity, specificity, negative predictive value, and positive predictive value were derived on the basis of data of our study. P value of ≤ 0.05 was taken as a cut-off for significance. Results: Our study comprised 30 patients with severe alcoholic hepatitis. At the end of 90-day follow-up, 23 patients survived with 7 (23%) mortalities because of liver failure and sepsis. On day 7, barring the ABIC score (0.713), the areas under curve of all scores were above 0.8 and were hence of comparable efficacy. Lille score performed the best at assessing treatment response but had the limitation of not being applicable at day 1. Conclusion: Severe alcoholic hepatitis carries a bad prognosis. Cut-offs of DF and MELD need to be raised, whereas they need to be lowered in the case of GAHS, ABIC, and Lille score to improve the predictive value of score. Steroid-treated patients should be evaluated preferably with Lille score for a better prediction of survival.