Background

Prednisolone is considered the cornerstone treatment for severe alcoholic hepatitis (AH). However, its use is limited by the increased risk of infection in an already immunocompromised patient population. Among patients with severe AH, there exists a group of non-responders who do not benefit from prednisolone therapy. Day-4 Lille score is a widely employed prognostic model used to identify this non-responder subgroup. The present study evaluates the prognostic ability of the inflammatory marker, the neutrophil-lymphocyte ratio (NLR), as a stand-alone model and in conjunction with the day-4 Lille score.

Methods

We retrospectively reviewed the electronic medical records of patients diagnosed with AH. Demographic and biochemical data at diagnosis were collected to calculate Maddrey’s discriminant function (MDF) and model for end-stage liver disease (MELD) score upon admission and also on day 4. Receiver operating characteristic (ROC) curves were plotted for day-4 NLR and day-4 Lille score for prediction of 90-day mortality, and optimal cut-off values were determined. Patients were then subcategorized into groups based on the generated optimal cut-off values. Categorization was validated by comparing the mortality rate in each group with the chi-squared test. We then performed a multivariate analysis for prediction of 90-day mortality using day-4 Lille score and day-4 NLR, constructing a new prediction score based on the odds ratio (OR). The ROC curve of the new score was plotted and the area under a curve (AUC) was reported and compared with previously validated scores.

Results

Our analysis demonstrated that both day-4 NLR and Lille score individually predicted 90-day mortality with statistical significance (p: 0.049, p: <0.001, respectively). The ROC analysis of day-4 Lille score for the prediction of 90-day mortality revealed an AUC of 0.819 with an optimal cut-off value of 0.45 (sensitivity: 83.3%, specificity: 76.1%). Day-4 NLR had an AUC of 0.756 with an optimal cut-off value of 12.3 (sensitivity: 66.7%, specificity: 78.1%) The combined day-Lille-NLR model with a cut-off of 0.55 had an AUC of .889, which was higher than day-4 Lille score and NLR independently.

Conclusion

Day-4 NLR is an easily assessed prognostic model of mortality in alcoholic hepatitis. However, it often underperforms relative to day-4 Lille score. Combining these two models to create a "modified" Lille score adds increased performance characteristics to the prediction of outcomes/mortality. The "modified" Lille score presented in this study can be used to further cut down the number of non-responders who are often forced to undergo costly and potentially harmful treatment courses.