Background & Aims: Carriage of rs738409:G in PNPLA3 is associated with an increased risk of developing alcohol-related cirrhosis and has a significant negative effect on survival. Short-term mortality in patients with severe alcoholic hepatitis is high; drinking behaviour is a major determinant of outcome in survivors. The aim of this study was to determine whether carriage of rs738409:G has an additional detrimental effect on survival in this patient group. Methods: Genotyping was undertaken in 898 cases with severe alcoholic hepatitis, recruited through the UK Steroids or Pentoxifylline for Alcoholic Hepatitis (STOPAH) trial, and 1,188 white British/Irish alcohol dependent controls with no liver injury recruited via University College London. Subsequent drinking behaviour was classified, in cases surviving ≥90 days, as abstinent or drinking. The relationship between rs738409 genotype, drinking behaviour and survival was explored. Results: The frequency of rs738409:G was significantly higher in cases than controls (29.5% vs. 18.9%;). Case-mortality at days 28, 90 and 450 was, 16%, 25% and 41% respectively. There was no association between rs738409:G and 28-day mortality. Mortality in the period day 90-450 was higher in survivors who subsequently resumed drinking (Hazard Ratio [HR] 2.77, 95% Confidence Interval [CI] 1.79-4.29; p<0.0001) and individuals homozygous for rs738409:G (HR 1.69, 95% CI 1.02-2.81, p=0.04). Conclusion: Homozygosity for rs738409:G in PNPLA3 confers significant additional risk of medium-term mortality in patients with severe alcoholic hepatitis. Rs738409 genotype may be an additional factor when considering treatment options in these patients.Response to Reviewers: Reviewer statistics 1.Imputing missing drinking behavior at 90 days based on observed values at 1 year is tricky, even in sensitivity analyses. Actually, this corresponds to sort of conditioning on the future, which should be avoided because it may lead to biased results (see works by Terry Therneau, for instance). Since it is a sensitivity analyses, the issue is less severe than for a primary analysis, but the limit should be stated.
2.The authors explain that now they only kept subgroup analyses where a significant interaction was found, but in the manuscript, it remains "Separate models were fitted for clinically relevant features and biochemical parameters", thus without conditioning on a significant interaction.
3.My point on the sentence "Individuals homozygous for rs738409:G had a significantly lower 450-day survival, even if abstinent" has been misunderstood, and it is certainly because I was not clear enough. I did not mean that there was no significant difference in any of the whole population and the subgroup analyses, but that it was not possible to claim that 450 days mortality would be lower in patients homozygous for rs738409:G because this was not analyzed. What the study shows is that among patients still alive at 90 days, the 450 days mortality was different, which is different. Simply stating "450 day...