Effect of Proton Pump Inhibitors on the Risk and Prognosis of Infections in Patients With Cirrhosis and Ascites

Gitte Dam; Hendrik Vilstrup; Per Kragh Andersen; Lars Bossen; Hugh Watson; Peter Jepsen


Liver International. 2019;39(3):514-521. 

In This Article


Infections are common and clinically important among patients with decompensated cirrhosis. In this detailed international multicentre data set, we found that they were associated with a more than three-fold increase in mortality. It is therefore worrying that so many patients—54% during these trials—used PPI in the face of mounting evidence that such use increases the risk of infections and death. We showed that users of PPI at the time of study inclusion had higher mortality than nonusers, and we found that PPI use was a risk factor for all types of infections, except upper respiratory tract infections which were probably caused by virus and clinically unimportant because they did not increase mortality. Interestingly, PPI use did not affect mortality among those who had an infection.

The data for this study were collected according to a detailed protocol, and it is highly unlikely that the validity of the collected data depends on PPI use or on any other patient characteristic. The number of trial participants and the rigour of the data collection during the trials are strong points of our study. Another strong point is that we used two different methods to examine the effect of PPI use on the incidence of infections and that they gave similar results. Marginal structural models are an important tool to examine the effect of a time-varying treatment (here: PPI use) on the cumulative risk of an outcome (here: infection). The standard cumulative incidence function does not suffice.[27] A final strong point of our study is that we were able to adjust for confounding by cirrhosis severity. We showed that markers of severe cirrhosis were associated with a high risk of infections and with a poor prognosis among the infected, but PPI users and nonusers were similar with respect to cirrhosis severity. We were concerned that PPI use was a marker of severe cirrhosis in itself; that is, that PPI users had more severe cirrhosis than nonusers in ways that were not described by the MELD score, albumin, etc. The absence of an association between PPI use and mortality following an infection provides compelling evidence that this was not the case. We do not think that there are strong uncontrolled confounders; hence, we believe that PPI use is indeed a risk factor for infections.

We showed that the effect of PPI use decreased during the study period. However, the decrease was not statistically significant, and we cannot see a plausible reason for such a decrease. We are confident that the associations we present, 1.43 from the Cox model and 1.57 from the marginal structural model, are valid estimates of the average effect of PPI use on the incidence of infections.

Proton pump inhibitor use appeared to be a risk factor for all kinds of infection except the harmless, presumably viral infections of the upper respiratory tract. One causal mechanism is that PPI use raises the gastric pH and promotes proliferation of bacteria such as Streptococcus and Enterococcus.[8,12,30] The latter have been shown to translocate from the stomach to the gut following acid suppression,[8] and this translocation increases the risk of spontaneous bacterial peritonitis. Moreover, Streptococcaceae originating from the oral cavity and proliferating in the stomach following acid suppression can colonize the tracheobronchial tract and result in pulmonary micro-aspiration of the gastric fluids.[30–32] PPI use could cause urinary tract infections, skin infections and other infections outside the gastrointestinal and respiratory tracts via bacteraemia resulting from the combination of bacterial translocation and reduced hepatic clearing of bacteria.[33] Kedika et al[34] have reviewed the adverse anti-inflammatory effects of PPI that could explain such reduced clearing. Finally, it has been suggested that PPI use can promote antibiotic resistance,[35] and such an effect could indeed contribute to the increased risk of infections. On the other hand, if PPI use conferred resistance, we would expect PPI users to have a higher mortality among the infected, but that was not the case.

Our study was conducted within a cohort of patients participating in a trial. Such patients are generally healthier than nonparticipants,[36] meaning that the cumulative risk of infections may be even higher in everyday clinical practice than reported here. The prevalence of PPI use might be different, too, and our estimate is in the lower end of the range of previous estimates.[17–22] Importantly, when we compare trial participants with trial participants, such as when we examined the relative risk of infections for users vs nonusers of PPI, our findings likely generalize to nonparticipants.

The generalizability of our relative risk estimates is supported by the fact that they are consistent with estimates from previous studies,[9–14] including those from an Austrian study of 607 patients with cirrhosis, of whom no less than 86% used PPI.[21] The authors concluded that there was no association between PPI use and spontaneous bacterial peritonitis or other infections, but in fact, they found that PPI users had a higher cumulative risk of spontaneous bacterial peritonitis (subdistribution hazard ratio = 1.38, 95% CI 0.63–3.01) and of spontaneous bacterial peritonitis or other infections (subdistribution hazard ratio = 1.71, 95% CI 0.85–3.44). These associations closely resemble those that we report; the only difference is that theirs were not statistically significant due to the smaller number of infections. No other studies have specifically examined the impact of PPI use on the risk of infections other than spontaneous bacterial peritonitis, but three studies have come close. Firstly, Bajaj et al[11] reported that PPI use increased the risk of serious infections overall, but it is unclear whether an association existed between PPI use and infections other than spontaneous bacterial peritonitis. Secondly, the Italian study by Merli et al[37] found that PPI use increased the risk of infections overall and at the same time presented the distribution of infection sites. This distribution hinted that PPI did not increase the risk of infections other than spontaneous bacterial peritonitis, but their analyses were not designed to address this question. Thirdly, O'Leary et al[38] reported that PPI use increased the risk of a subsequent infection among cirrhosis patients with an infection and that the second infection was often in a different location from the initial one.

Not all studies have found an association between PPI use and spontaneous bacterial peritonitis: an Argentinian study examined the association between PPI use and prevalence of spontaneous bacterial peritonitis among patients with cirrhosis and ascites, 43% of whom used PPIs. Although they did not control for confounding, it appears that there was in fact no association between PPI use and spontaneous bacterial peritonitis in their patients.[17] It is not clear why this study disagrees with most other studies, but—on balance—there is accumulating evidence, also from mechanistic studies, that PPI use changes the gut microbiota composition[8,39,40] and that this can have downstream adverse effects increasing the risk of infections.

The two previous studies on the effect of PPI use on mortality among infected patients reached opposite conclusions.[22,23] The larger of them, from Korea, included 533 patients with cirrhosis and spontaneous bacterial peritonitis, and it reported a higher short-term mortality for PPI users.[22] It was notable that only 15.3% of their cohort had used PPI within the previous 30 days and that PPI users had more severe cirrhosis than nonusers. We speculate that, in the Korean study unlike in ours, PPI use was restricted to those with more severe cirrhosis, so that PPI use remained a marker of severe cirrhosis after adjustment for MELD score and presence of HCC. Clearly, the huge variation in indications for PPI use makes it difficult to compare findings across studies.

To conclude, we showed that patients who used PPI at inclusion had a higher mortality than those who did not. We also showed that infections are associated with a sharp increase in mortality and that PPI users have more infections. Hence, the excess mortality among PPI users is partly explained by their higher risk of infections, although—as we showed—PPI use does not affect mortality among the infected. Our previous work showed that PPI use can cause hepatic encephalopathy[15] and that complication provides another explanation to their higher mortality. We recognize the central role of PPI to manage diseases caused by excess gastric acid, but are concerned by its widespread use among patients with cirrhosis and ascites. Clinicians should prescribe PPI on a proper indication or stop prescribing them.