Prevalent acute-on-chronic liver failure and response to corticosteroids in alcoholic hepatitis

We read the paper by Sersté et al. regarding the significance of acute-on-chronic liver failure (ACLF) in alcoholic hepatitis (AH) with interest. They conclude that ACLF in AH is associated with a poor outcome and that the Lille response to corticosteroids was reduced in those with prevalent ACLF.

In the light of this we have interrogated the STOPAH database to determine the role of ACLF in AH. The calculation of the CLIF-OF score in STOPAH patients was limited by the absence of data relating to respiratory function. However, in the Sersté study only 4.2% of patients in the Erasme cohort had pulmonary failure, which is assumed to have been infrequent in the STOPAH cohort. For the purposes of calculating the CLIF-C ACLF score, the respiratory sub-score was therefore assumed to be 1. Based on this assumption, the CLIF-OF score, the ACLF grade and the CLIF-C ACLF score were calculable in 1,019 of 1,069 patients who were previously described in the STOPAH trial.^,  Comparisons between groups were analysed using chi-squared (χ²) tests and between survival curves using Kaplan-Meier logrank tests for trends.

In addition to the absence of respiratory data, the STOPAH protocol excluded patients with severe renal impairment (creatinine >500 µmol/L) and those on inotropic support, thereby not including many patients with CLIF-OF sub-scores of 3 for renal and circulatory systems. Therefore, the proportion of those recruited with ACLF was less than that seen by Sersté et al. who noted that more than 20% of patients had circulatory and/or respiratory failure in the Erasme cohort. In the STOPAH cohort, the numbers of patients with ACLF were 737 (72.3%), 152 (14.9%), 115 (11.3%) and 15 (1.5%) for grades 0, 1, 2 and 3, respectively. The median CLIF-OF score was 8 (range 6–13) and mean CLIF-C ACLF score was 41.4 (standard deviation 6.8). For the purpose of analyses, ACLF Grade 2 and 3 patients were grouped together. Overall survival rates at 28 days were 90.2%, 75.7% and 56.9%, and at 90 days were 80.3%, 65.8% and 43.8% for ACLF 0, ACLF 1 and ACLF (2+3) respectively. The area under the curve for the CLIF-C ACLF score relating to prediction of 28-day outcome was 0.793 (95% CI 0.767–0.818) and of 90-day outcome was 0.727 (95% CI 0.698–0.754), similar to that described for other scores measured at baseline. The proportions of those with a Lille response are shown (Table 1), indicating a greater chance of Lille response with corticosteroid therapy in all groups, although not significantly so for those with ACLF. Irrespective of the grade of ACLF, a Lille response to corticosteroids was associated with a similarly favourable 90-day outcome. For ACLF grades 0, 1 and (2+3) the 90-day survival rates for responders were 83.7%, 80.0% and 83.3% respectively (p <0.892; χ² 0.23). For Lille-non-responders the equivalent survival rates were 68.1%, 45.8% and 36.7% (p <0.004; χ² 10.9). More patients with ACLF developed infection in the first week: 64 (8.7%) for ACLF 0, 26 (17.1%) for ACLF 1 and 25 (19.2%) for ACLF (2+3); p <0.001, χ² 18.0 ACLF vs. no ACLF. Similarly, for infection at 28 days: 150 (20.4%) for ACLF 0, 48 (31.6%) for ACLF 1 and 43 (33.1%) for ACLF (2+3); p <0.001, χ² 16.0 ACLF vs. no ACLF.

In conclusion the STOPAH data, although incomplete with regards to a full CLIF-OF score, confirms a higher risk of infection, a lesser chance of Lille response and greater mortality with higher ACLF grades. However, if a Lille response to corticosteroids is achieved, the survival benefit is maintained irrespective of ACLF grade.