Letter: proton pump inhibitors - severity of liver disease and mortality in patients with cirrhosis.

Abstract

We read the article by Dultz et al.1 with interest. In this prospective study involving 272 cirrhotic patients, the author compared the overall survival rate between proton pump inhibitor (PPI) and non-PPI users, and drew the conclusion that PPI was an independent risk factor for mortality in those patients. However, we still have some questions regarding this research. First, the baseline characteristics of PPI users and nonusers were not comparable: the PPI users had higher median MELD scores (16 vs. 12, P < 0.001) and Child-Pugh scores than the nonusers, and decompensated cirrhosis predominated in the former group (Child-Pugh B+C 84.0% vs. 59.9%). It is strange that the authors did not give the P values for the difference in these baseline characteristics between the two groups. Based on the excellent review by Cholongitas et al.,2 mortality is positively associated with high Child–Pugh or MELD scores, which means the PPI users will have a lower survival rate. These differences between the two groups compromised the reliability of the findings. Second, according to the study of Borzio et al.,3 patients with decompensated cirrhosis have a trend to increased rates of bacterial infection, which increases mortality. In fact, the rates of infectious complications and bacterial colonisation were higher in PPI users in Dultz's research. Although there is much evidence that PPI therapy is associated with the development of spontaneous bacterial peritonitis in cirrhotic patients,4, 5 Dultz et al. did not compare all possible risk factors between the two groups that might significantly influence the mortality rate, such as hepatorenal syndrome, gastrointestinal bleeding, albumin use or antibiotic therapy. Third, the authors did not provide the duration of PPI therapy. Was it short or long-term? The median follow-up time was 266 days with a range of 1–1382 days. We find it hard to believe that short-term PPI therapy will lead to a poor outcome several years later, and therefore wonder whether it was appropriate to include individuals that were only followed up for a very short period of time (like 1 day) in their statistical analysis? Finally, in this research, 89 of 213 PPI takers had recent gastrointestinal bleeding, endoscopic ligation of varices, severe reflux or peptic ulcer disease, and the remaining also presented with epigastric pain or abdominal discomfort. Actually, the 6-week mortality rate due to variceal bleeding is 15–20%, and in patients with severe decompensated cirrhosis, the mortality rate increases up to 30%.6-8 The authors did not describe the bleeding or re-bleeding rate during follow-up, although variceal bleeding was one of the main reasons of death in that study. As mentioned in the paper, PPI was given to patients without strong indications (epigastric pain, nausea or vomiting). If the authors could present the mortality data based on subgroup analysis (patients with or without strong indications for PPI), the results may be more convincing. Due to all these biases and limitations inherent in their study, it is hard to draw any firm conclusions that PPI was the main risk factor for mortality. We hope the authors can provide more details in response to our points. Declaration of personal and funding interests: None.