The Diagnostic and Prognostic Significance of Liver Histology in Alcoholic Hepatitis

Ewan Forrest; Gemma Petts; Andrew Austin; Kirsty Lloyd; Mark Wright; Nikhil Vergis; Stephen Atkinson; Steven Masson; David Patch; Alberto Quaglia; Mark Thursz; Robert Goldin


Aliment Pharmacol Ther. 2021;53(3):426-431. 

In This Article



The STOPAH trial recruited 1068 patients; from whom biopsies were taken from 182 (17%) patients from 17 recruitment centres (26% of 65 centres) during the index admission. Of these, 21 (12%) were inadequate for histological assessment. Therefore analysis of histological findings was based upon 161 patients (15% of the total STOPAH trial cohort).

There were no significant differences between key clinical features in those with an adequate biopsy sample cohort compared to the main STOPAH trial cohort however the mortality of the biopsied patients was lower in those with an adequate biopsy compared with the non-biopsied STOPAH patients (Table 1). The majority of liver biopsies (100, 62%) were performed in five centres, where the overall biopsy rate for trial participants was 40%. Within these five centres the 28-day and 90-day mortalities for biopsied patients were 8% and 16% compared with 16% and 26% for non-biopsied patients (P = 0.08 and 0.05 respectively).

Clinical vs Histological Diagnosis

A definite histological diagnosis of ASH was reported in 140 of the 161 adequate biopsies (87%). In 18 cases (11%) there were histological features associated with alcohol-related liver injury but ASH was not diagnosed, most commonly due to an absence of sufficient hepatocyte ballooning. There were only three biopsies (2% of cohort) without histological features of alcohol-related liver injury. Of the 140 biopsies diagnostic of ASH, 99 (71%) had definite cirrhosis (Laennec fibrosis score 5–6) and only two patients had mild or minimal fibrosis (Laennec fibrosis score 1–2).

The timing of liver biopsy varied with 40 (25%) cases biopsied before randomisation, 26 within 48 hours of randomisation (16%), 45 (28%) between day 2 and day 7 and 50 (31%) between day 7 and day 28 of randomisation. Early liver biopsies obtained before randomisation showed histological features of ASH in 37 cases (92.5%). The rate of ASH confirmation fell to 88% (23 cases) if biopsied within 48 hours after randomisation and fell further to 84% (80 cases) if biopsied more than 2 days after randomisation. Patients not exposed to prednisolone at the time of biopsy had features of ASH in 83 of 91 biopsies (91%), whereas biopsies from prednisolone-treated patients had ASH in 57 of 70 cases (81%). Higher clinical prognostic scores (GAHS and MELD) were associated with a higher rate of confirmed ASH. Those with MELD scores between 21 and 24 had confirmed ASH in 32 of 40 biopsies (80%) whereas ASH was identified in 61 of 67 biopsies (91%) when the MELD ≥25. Combination of these factors affecting the presence of ASH (biopsy obtained before randomisation, no exposure to prednisolone treatment and either a GAHS ≥9 or MELD ≥25) identified ASH in 100% of cases (Table 2).

Prognostic Information Based on Liver Biopsy

In the 60 liver biopsies taken before or within 48 hours of randomisation with confirmed histological features of ASH, overall survival was 83% at 90 days. Using the AHHS scoring system, 37 cases (62%) were classified as Severe, 20 (33%) as Moderate and three (5%) Mild. By 90 days Mild, Moderate and Severe groups had survival rates of 100%, 90% and 78% respectively (Figure 1). These survival differences between groups were not significant: comparison of mild plus moderate (91%: 21 of 23 patients) with severe (78%: 29 of 37 patients) groups: P = 0.18; HR 2.35 (0.66, 8.33).

Figure 1.

Survival curves for Alcoholic Hepatitis Histological Score (AHHS) severity groups

There were no statistically significant differences in AUC on comparison of the AHHS with clinical scoring systems (MELD and GAHS): 90-day values of 0.654, 0.652 and 0.634 for AHHS, MELD and GAHS respectively (Table 3). The AHHS score did not correlate significantly with the DF (ρ = 0.213; P = 0.10) but did correlate with both GAHS (ρ = 0.340; P = 0.008) and MELD (ρ = 0.273; P = 0.03).