Association between proton pump inhibitors and respiratory infections : A systematic review and meta-analysis of clinical trials

London Health Sciences Centre, Division of Gastroenterology, Department of Medicine, Schulich School of Medicine and Dentistry, London, Ontario Correspondence: Dr Nabil Sultan, London Health Sciences Centre, 339 Windermere Road, London, Ontario N6A 5A5. Telephone 519-668-7974, fax 519-668-7974, e-mail nmsultan@uwo.ca Received for publication February 9, 2008. Accepted May 28, 2008 N Sultan, J Nazareno, J Gregor. Association between proton pump inhibitors and respiratory infections: A systematic review and meta-analysis of clinical trials. Can J Gastroenterol 2008; 22(9):761-766.

P roton pump inhibitors (PPIs) have become the mainstay of treatment and prophylaxis for gastroesophageal reflux disease (GERD) and peptic ulcer disease (PUD), and are one of the most prescribed classes of medications in North America because of both their excellent efficacy and safety profiles (1)(2)(3).
The most commonly reported side effects of PPIs include headache, nausea and diarrhea, which occur only slightly more frequently than with placebo in randomized, controlled trials (RCTs).Long-term safety issues with PPIs have become progressively less discussed in recent years, with no long-term complications reported in up to 11 years of continuous use (4).Initial concerns regarding the possible increased risk for atrophic gastritis in Helicobacter pylori-positive patients receiving PPIs and the potential for cancer risk have been refuted by a clinical trial (4).Dosage reduction of PPIs is not necessary, even in patients with compromised renal or hepatic function; all of this has contributed to the soaring popularity of PPI use in the general population.
While initial investigations into complications of PPI use focused on dysplasia and cancer, there has been growing evidence that PPI use may not be completely free of risk.Recent cohort and case-control studies investigating the risk for Clostridium difficile infection with in-hospital PPI use have shown a positive association, with ORs ranging from 2.4 to 2.7 (5,6,7).More recent data suggest that the association may even extend into the community setting (8).
Laboratory and clinical evidence suggests that a less acidic gastric pH may be linked to increased bacterial colonization of the stomach (9).This colonization is thought to be an important source of pathogens that could cause pneumonia by translocating to the lungs via the upper digestive and upper respiratory tracts (10).Until recently, studies looking specifically at clinical outcomes such as ventilator-associated pneumonia in the critical care setting have not been definitive in either demonstrating or entirely refuting a link to acid suppressive drugs (ASDs) (11,12).Although it accounts for 60,000 deaths annually in the United States alone, until recently, virtually no attention has been paid to any association between community-acquired pneumonia (CAP) and ASD use.Over the past few years, European population-based studies (13)(14)(15) have suggested that an association may exist.
The objective of the present study was to examine the association between the outpatient use of PPIs and respiratory infections, including CAP.

Study identification
Relevant articles evaluating the efficacy of PPIs published before August 2007 were identified through PubMed, MED-LINE and Cochrane databases.Individual searches using the key words "proton pump inhibitors", "PPI", "esomeprazole", "rabeprazole", "lansoprazole", "pantoprazole" and "omeprazole" were conducted.Studies identified by this method were used to identify additional citations.

Study selection
The meta-analysis inclusion criteria included randomized, single-or double-blinded, placebo-controlled studies investigating PPIs in various gastrointestinal diseases.Unblinded studies were excluded.Most studies identified had evaluated the use of PPIs in patients with GERD or PUD.Selected studies were required to have clearly presented the prevalence of the specific adverse events of respiratory infection, upper respiratory infection or pneumonia in both treatment and placebo arms.Studies stating no difference in adverse events without explicit mention of respiratory infections were excluded.

Data extraction
Data regarding adverse events were required to be given explicitly, including clearly stating the adverse events of respiratory infections and upper respiratory infections (or equivalent), with quantification of both active and placebo arms.
Data analysis χ 2 analyses evaluating active and placebo arms, as well as their association with respiratory infections, were performed to assess the significance of association.A meta-analysis of outcomes was performed by combining trials using the Mantel-Haenszel method (RevMan 4.2; The Cochrane Collaboration, United Kingdom).Statistical heterogeneity was evaluated and P<0.1 was considered significant.Primary outcomes were summarized.

RESULTS
The initial key word search identified 7457 studies.The majority were excluded because they were either not RCTs or because the control group did not receive a placebo.Most studies identified evaluated the use of PPIs in patients with GERD or PUD.Only 70 RCTs met the major inclusion criteria.Of these 70 studies, 63 were excluded because it was not possible to isolate the adverse events of respiratory infection, upper respiratory infection or pneumonia for the active group, control group or both.
Seven studies (16)(17)(18)(19)(20)(21)(22) met all predefined inclusion criteria.All seven studies were peer-reviewed publications.Three studies used esomeprazole, two used omeprazole, one used pantoprazole and one used rabeprazole.All studies compared PPIs with placebos.They ranged in duration from four weeks to six months.A total of 2586 patients (1943 active treatment, 643 placebo) were included in all of the studies.Table 1 summarizes the characteristics of the trials.
Table 2 summarizes the prevalence of respiratory infections in the included studies.The mean prevalence of respiratory infections in the active treatment groups was 4.2% (range 0% to 6.6%) and in the placebo groups was 2.9% (range 0% to 8%).
χ 2 analyses (Table 3) revealed that the majority of trials showed no statistically significant association between PPIs and respiratory infections, although five of seven studies did demonstrate a greater number of infections in the active treatment arm.Only a single treatment arm in one study showed a significant association; this occurred in the esomeprazole 40 mg arm when tested against placebo in a study by Vakil et al (17), revealing a P-value of 0.05.
A meta-analysis of included studies (Figure 1) revealed no significant association between PPI use and respiratory infection.A total of 84 respiratory infections (of 1943 subjects) occurred in treatment groups and 22 (of 643 subjects) occurred in placebo groups; 4.3% and 3.4%, respectively.The total OR for developing a respiratory infection with PPI treatment was found to be 1.42 (95% CI 0.86 to 2.35).The test for overall effect revealed a P-value of 0.17 (deemed insignificant), and the test for heterogeneity revealed a P-value of 0.51.

DISCUSSION
The present study failed to show a significant association between the use of PPIs and respiratory tract infections over the course of use during which these medications are normally initially prescribed.While statistically insignificant, the absolute incidence of respiratory infections was estimated to be 1.3% higher in patients receiving PPIs than in those receiving placebo.This would translate into a number needed to harm of approximately 77, suggesting that even if a true association exists, it may have limited clinical significance, except possibly in the most vulnerable patients.
Only a minority of the 70 trials identified actually tracked respiratory infections such as pneumonia.Another limitation of our study is that most studies used the term 'respiratory infections' without clearly defining the diagnostic criteria used (history, radiological, positive cultures, antibiotic use, etc) for those events and, as a result, there may have been a significant degree of subjectivity related to those outcomes.We could not determine the proportion of viral infections in each study, which would be important given that the hypothesis forwarded is that ASDs may cause bacterial respiratory infections through bacterial translocation.There are no randomized trials of PPIs investigating respiratory infections as an end point and we therefore identified respiratory infections followed in RCTs as adverse events.Reporting bias may have influenced the results because only a minority of studies reported the presence or absence of respiratory infections.Despite these limitations, we were able to pool the results of more than 2000 patients from placebo-controlled trials, the strongest level of evidence available.
Our investigation was stimulated in part by the results of two Dutch studies by Laheij et al (14,15).The first, a retrospective questionnaire (15) of 700 outpatients, suggested that ASD users were 2.3 times more likely to have had symptoms of a respiratory infection in the month preceding assessment.In the second, a large cohort of nearly one million patients was studied (14).In this study, the retrospective cohort analysis revealed that ASD users were nearly 4.5 times more likely to develop pneumonia than non-users.A case-control analysis of the same cohort, in an attempt to control for confounders, revealed an adjusted OR of 1.89 for PPIs and 1.63 for histamine-2 receptor antagonists; both values were statistically significant.The authors estimated an attributable risk of one extra case of pneumonia for every 100 years of PPI use.
With respect to dose-response, the results of Laheij et al (14) showed that "more than one defined daily dose (DDD)" of PPI has a significantly higher OR for pneumonia than "less than one DDD" (OR 2.63 versus 1.21, respectively).While our study could not provide any conclusive evidence for a doseresponse relationship, one study (17) did find a statistically significant association between PPIs and respiratory infections in χ 2 testing, using higher dose esomeprazole (40 mg versus 10 mg or 20 mg).However, other trials comparing different doses of PPI failed to show any association across all doses.
Although diseases such as nonvariceal upper gastrointestinal bleeding have a significant case fatality rate of approximately 5% (57), the case fatality rate of CAP is not insignificant.Of the one million CAP patients requiring hospitalization in the United States annually (58), 2% to 21% will    Figure 1) Results of a meta-analysis testing the association between proton pump inhibitors (PPIs) and respiratory infections.df Degrees of freedom succumb to their illness; this figure rises to 50% among patients with severe disease (59).If PPIs do cause some cases of CAP, the risk-benefit ratio of chronic PPI administration clearly needs to be taken into account, particularly when administered to asymptomatic individuals for prophylactic indications such as the prevention of nonsteroidal antiinflammatory-induced ulcers.This concern is highlighted by recent evidence that up to 60% of primary care PPI prescriptions may not be fulfilling established PPI indications (60), and by other data showing the explosive growth of the PPI market, with a British study (61) showing a growth of 456% in PPI prescriptions from 1992 to 1997.The development and use of PPIs has led to a revolution in the management and prophylaxis of some of the most serious and debilitating diseases of the upper gastrointestinal tract over the past two decades.While it is not always possible to perform placebo-controlled trials to answer every clinical question, particularly those pertaining to relatively rare adverse events, the indications for the use of PPIs are sufficiently common that such a trial is likely feasible.Such a study would aid physicians and their patients in making informed decisions regarding the chronic administration of ASD therapy.

TABLE 1
Studies meeting inclusion criteria for the present study