Reply

We thank Dr. Ali Khan and his colleagues for their interest in our research. They question our conclusion that proton pump inhibitors (PPIs) pose a threat to patients with cirrhosis by increasing the risks of hepatic encephalopathy and spontaneous bacterial peritonitis, arguing that the association may be due to confounding. We find it unlikely that residual confounding can account for our relative risk estimate of 1.36.1 Without confounder adjustment the hazard ratio was 1.50 (95% confidence interval 1.12-1.99), so the confounders we adjusted for were actually not strong confounders. This is also evident from the fact that the PPI users and nonusers were very similar, as shown in our article's Table 1. We disagree with Dr. Ali Khan's statement that in observational epidemiology relative risks below 2 are generally due to bias. We find that the review by Sørensen et al. offers a more balanced view of the pros and cons of observational and randomized studies.2

Despite our differences of opinion, we agree with Dr. Ali Khan and colleagues on the most important and self-evident aspect: patients with a valid indication for PPIs can use them; others should not. Such practice would probably spare patients from many hepatic encephalopathy episodes, and any risk of hepatic encephalopathy attributable to PPI use would at least be countered by beneficial effects of PPIs. When we accomplish this, we can consider how to handle the excess hepatic encephalopathy risk for those patients who need PPIs. We have no evidence base for this situation, but it is a possibility that lactulose could be preemptively coadministered together with PPIs.3 The way to obtain an evidence base seems to be a randomized PPI discontinuation trial for cirrhosis patients with a valid indication for PPIs. We believe that such a trial should be undertaken.