Letter: Predictors of Corticosteroid Response in Alcohol-Related Hepatitis

Abstract

We read with interest the observational study by Idalsoaga et al. [1], which analysed a group of 289 patients treated with corticosteroids for alcohol-related hepatitis (AH). Models to determine response to corticosteroid therapy were assessed. It found that the Lille-7 score was the best-performing model, although none of the tested models had a high degree of discrimination. Determining response to corticosteroids for AH is important in order to abbreviate treatment in those not deriving benefit and so avoid corticosteroid-related complications. A dynamic model is one which looks at a change in variables after exposure to corticosteroids to determine their effectiveness. This is certainly true of the Lille model, which includes change in bilirubin from baseline (pre-treatment) to either Day 4 or Day 7 of treatment. However, the authors have also tested the trajectory of serum bilirubin (TSB), the neutrophil-to-lymphocyte ratio (NLR) and the change in NLR (Delta NLR), none of which have been proposed as indicators of response to already initiated corticosteroid therapy.

The TSB was originally described to categorise the trajectory of bilirubin before exposure to corticosteroids [2]. Therefore, it is a characteristic of a patient pre-treatment and not a dynamic model reflecting the response to already initiated corticosteroid treatment. If the authors wished to look at the change in bilirubin after exposure to corticosteroids, they should have used either ‘early change in bilirubin’ (ECBL) [3], or the percentage change in bilirubin over 1 week [4].

Similarly, the NLR has only been described as a baseline characteristic to identify those likely to benefit from corticosteroid treatment compared with those not receiving such treatment. The NLR has never been advocated as a simple prognostic or dynamic model, but as a baseline variable with a narrow ‘window’ (between 5 and 8) to predict improved outcomes with corticosteroid treatment [5]. This was demonstrated in 789 patients enrolled in the STOPAH trial and validated in a separate group of 237 patients. As there is a ‘window’ of NLR which identifies those who may benefit from corticosteroids, it is to be expected that it would perform poorly when analysed as a linear prognostic score. Its value can only be determined by comparing treated with untreated patients, which the current study does not assess. Delta NLR has never previously been suggested as a model to assess corticosteroid treatment in AH, and the well-recognised effects of corticosteroids upon circulating leucocytes will make this difficult to interpret.

In conclusion, Idalsoag and colleagues have identified the Lille score at Day 7 as the most useful model to determine response in those who have already started corticosteroids. However, their study has not been designed to investigate TSB or NLR as useful pre-treatment characteristics of patients. These readily assessable parameters still have relevance for the clinical management of patients with severe AH.