Meta-analysis of proton pump inhibitors induced risk of community-acquired pneumonia

Abstract

Purpose

Proton pump inhibitors (PPIs), one of the most widely used medications, are commonly used to suppress several acid-related upper gastrointestinal disorders. Acid-suppressing medication use could be associated with increased risk of community-acquired pneumonia (CAP), although the results of clinical studies have been conflicting.

Data sources

A comprehensive search of MEDLINE, EMBASE and Cochrane library and Database of Systematic Reviews from the earliest available online year of indexing up to October 2018.

Study selection

We performed a systematic review and meta-analysis of observational studies to evaluate the risk of PPI use on CAP outcomes.

Data extraction

Included study location, design, population, the prevalence of CAP, comparison group and other confounders. We calculated pooled odds ratio (OR) using a random-effects meta-analysis.

Results of data synthesis

Of the 2577 studies screening, 11 papers were included in the systematic review and 7 studies with 65 590 CAP cases were included in the random-effects meta-analysis. In current PPI users, pooled OR for CAP was 1.86 (95% confidence interval (CI), 1.30–2.66), and in the case of recent users, OR for CAP was 1.66 (95% CI, 1.22–2.25). In the subgroup analysis of CAP, significance association is also observed in both high-dose and low-dose PPI therapy. When stratified by duration of exposure, 3–6 months PPIs users group was associated with increased risk of developing CAP (OR, 2.05; 95% CI, 1.22–3.45). There was a statistically significant association between the PPI users and the rate of hospitalization (OR, 2.59; 95% CI, 1.83–3.66).

Conclusion

We found possible evidence linking PPI use to an increased risk of CAP. More randomized controlled studies are warranted to clarify an understanding of the association between PPI use and risk of CAP because observational studies cannot clarify whether the observed epidemiologic association is a causal effect or a result of unmeasured/residual confounding.

clinical study, epidemiology, gastrointestinal diseases, pneumonia, proton pump inhibitors

Introduction

Community-acquired pneumonia (CAP) is one of the most common infectious diseases and is also an important cause of mortality and morbidity worldwide. A recent study in the USA reported that the age-adjusted death rate caused by influenza and pneumonia was 20.3 per 100 000 persons [1]. The incidence of CAP ranged from 4 to 5 million cases per year, and of those about 25% are required hospitalization [2]. Fung et al. [3] reported that subjects aged more than 60 years with alcoholism, heart and lung diseases and immunosuppressive therapy were independent risk factors for pneumonia. Several studies show other risk factors such as low body mass index, history of respiratory infection and pneumonia, smoking, diabetics and asthma [4–6].

Recently, several studies mentioned that acid-suppressing medication use has been associated with increased risk of CAP. It is not fully clear the mechanism in which acid suppressor medication might increase the risk of CAP but change in gastric PH due to medications can alter normal gastrointestinal and oropharyngeal flora, leading to decreased elimination of or increased colonization of various pathogens [7]. Indeed, the elevation of gastric PH by acid suppressor promotes the proliferation of bacteria in the mouth and oropharynx [8, 9]. As gastric acid has a major role in protecting against infection and therefore, its attenuation provides a feasible mechanism to explain the role of PPIs to increase CAP [10, 11].

Whether the use of PPI therapy on patients is affected the outcome of CAP is still much debated. Some epidemiologic studies have examined the association between PPIs and the risk of CAP, but the results of previous studies were inconsistent. Despite limited definitive evidence on the magnitude of risk associated with PPI use and widely varying results of the available data, regulatory authorities have recently notified healthcare professionals and patients regarding a possible increased risk of CAP with the use of these medications [12]. Recently, Lambert et al. [11] reported that the outpatient PPI use has been associated with a 1.5-fold increased risk of CAP, within the first 30 days after initiation of therapy. Because of the widespread use of the PPIs, it is necessary to determine whether there is a relationship between the use of PPI and the risk of CAP. We, therefore, performed a systematic review with meta-analysis of existing observational studies to evaluate the association between PPI use and risk of CAP.

Figure 1

Diagram of study selection, adapted from PRISMA group 2009 flow diagram. Note: Studies could be excluded for more than one reason; therefore, the sum of exclusion reasons exceeds the number of total studies. Examples of studies lacking extractable data included abstracts without full results (such as abstracts reporting OR without exposure or outcome definitions) or articles summarizing outcomes quantitatively in the text, such as manuscripts reporting adverse event categories that grouped dementia among other events.

Methods

This systematic review and meta-analysis were conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [13] and Meta-analysis of Observational Studies in Epidemiology (MOOSE) [14] (PRISMA checklist reported in Table S1 in appendix).

Data sources

To identify studies, we did complete a comprehensive search of MEDLINE, EMBASE and Cochrane library and Database of Systematic Reviews from the earliest available online year of indexing up to October 2018. Our search strategies were designed and performed by experienced authors with input from study investigators. We used the following text words as search terms: ‘proton pump inhibitors’, ‘omeprazole’, ‘pantoprazole’, ‘rabeprazole’, ‘lansoprazole’, ‘esomeprazole’, ‘community-acquired pneumonia’ and ‘pneumonia’. Our search included studies published in English and non-English language. Further, we also scanned the bibliographies of all retrieved articles for additional relevant articles. The authors of potentially eligible abstracts, posters or manuscripts were contacted via email to clarify study information and additional data were obtained. There is no any other unpublished data included in our analysis.

Study selection and eligibility criteria

Two authors (Y.C.W. and C.C.C.) did the research, data extraction and assessment of risk bias, which were subsequently reviewed by the other authors (P.A.N. and M.I.). All disagreement was resolved by discussion with the main investigator (U.I.).

We included studies that met following inclusion criteria: (i) observational studies, (ii) non-PPI users or past users were used as the reference group for comparison, (iii) reported CAP outcome risks associated with PPIs use the outcome of interest was CAP and (iv) odds ratio (OR)/risk ratio (RR) with 95% confidence interval (CI) or adequate data provided with exact information to analyze them. We excluded the studies if they met any of the following criteria: (i) case report, editorial and review, (ii) studies that did not provide outcome with OR/RR, (iii) studies only used CAP diagnostic as the outcome and (iv) associations considered with non-preceding CAP.

Data extraction

All studies based on study characteristics including first author’s last name, publication year, source of study, participant’s characteristics, study location, sample size, number of cases and participants, variable adjustment and the risk estimated with 95% CI and the same two authors independently screened for final inclusion. When any problem arose, the main investigator (Y.C.L.) adjudicated discordant assessments. While we screened the title and abstract and the full text in a similar fashion. However, specific exclusion reasons were documented only during the full-text screening.

Assessment of methodological quality

The main outcome of this meta-analysis was the use of PPIs and the risk of CAP among the population. In this study, we assessed the methodological quality of the included studies based on the Newcastle-Ottawa Scale (NOS) for the quality of observational studies in meta-analyses [15] (Table S2 in appendix). This scale uses a star system to evaluate the quality of included studies based on three broad perspectives: the selection of study cohort, the comparability of the study cohort, and the ascertainment of the outcome of interest, length of follow-up and loss to follow-up rate. After judging these three perspectives, a maximum of nine stars could be assigned to each study. A study with seven or more stars was thought to be of high quality. We used a combination of validity criteria suggested by Loke et al. [16] and Levine et al. [17] to assess the risks for bias in the studies.

Statistical analysis

We computed pooled estimates of the odds of CAP among PPI users compared with non-users using a random-effects model meta-analysis, which assumes that the true underlying effect varies between observational studies. We used risk estimates obtained with random-effects meta-analysis instead of fixed-effects models because this approach provides a more conservative assessment (i.e. wide CIs) of the average effect size. In the case of systematic analysis, we carefully examined study countries, study design, study period, number of cases and control and adjustments. For the case of meta-analysis, we followed the definition of the drug exposure for the use of PPIs as ‘current use’ referred to as a prescribed therapy that lasted until the index date or had ended within the previous 30 days. ‘Recent use’ described a prescription that ended between 31 and 90 days before the index date. ‘Past use’ was defined as a prescription ending between 91 days and 1 year before the index date. To resolve the proportion of variability in the effect estimates due to heterogeneity, we calculated Cochrane’s Q test (reported with a λ²and P-value) the I² statistic [18, 19]. We observed the accepted cutoff of 50% to define significant heterogeneity and additionally calculated the associated P-value (via χ² test).

The forest plots for each of our exposure–outcome comparisons were generated. Further, we used a random-effects based model to calculate the weighted mean, OR and associated 95% CI and P-value due to significant clinical, methodological and statistical heterogeneity across the included studies. So, all effect estimates were assumed to approximate the OR and were pooled in meta-analyses. To test for publication bias, we formed a funnel plot and undertook the Egger test.

Results

Study selection

Initially, we planned to include randomized controlled trials (RCTs) in this systematic and meta-analysis; however, as of the end of our search period, no RCTs on this topic had been reported. A total of 2577 records were identified through our initial database search. Of these, 2532 studies were excluded after preliminary review and 45 studies underwent detailed full-text evaluation. These 11 studies met our inclusion criteria for the systematic review. No additional abstracts were identified by hand searches of conference proceedings. Fig. 1 shows the number of studies by reason for exclusion at each stage of the eligibility assessment. Seven studies included for meta-analysis involved 98 397 CAP cases.

Systematic review

This systematic review identified 5 239 371 subjects, and most of the studies pointed out the specific male and female numbers in both CAP cases and control subjects, except for two studies for no information on gender [20, 21]. The studies meeting our inclusion criteria are summarized in Table 1. The majority of the studies were of case–control (n = 9) and two were cohort studies. The total number of CAP cases in the included studies was 107 322. Six studies conducted in Europe, four in North America and only one study in Asia. Data collection spanned was around 3 decades (1987–2018). A modified NOS was used to assess the methodological quality of included observational studies (Table 2).

Table 1

Characteristics of included studies

Authors, Ref.	Country	Study period	Study design	Participants	Adjustment
Chen, 2015	Taiwan	1997–2002	Cohort	619/8076	Age, sex, COPD, asthma
Ernst, 2012	UK	1997–2009	Case–control	1835/17 923	Age, sex, Parkinson’s disease
Filion, 2013	Canada, UK, & USA	1997–2010	Cohort	5135/4 238 504	Age, sex, asthma, COPD
Euric, 2010	Canada	2000–2002	Case–control	248/2476	Age, sex, smoking status, heart disease
Almirall, 2008	Spain	1999–2000	Case–control	1336/1326	Age, sex, smoking status
Rodríguez, 2009	Sweden	2000–2005	Case–control	7297/9993	Age, sex, BMI, smoking status
Hermos, 2012	USA	1998–2007	Case–control	1544/15 440	Age, sex, liver disease, kidney disease
Gulmez, 2007	Denmark	2000–2004	Case–control	7642/34 176	Age, diabetes, NSAIDs
Laheij, 2012	Netherland	1995–2002	Case–control	475/4690	Age, sex, diabetes, heart failure
Sarkar, 2008	UK	1987–2002	Case–control	80 066/799 872	Age, sex, alcoholism, stroke, diabetes
Dublin, 2010	USA	2000–2003	Case–control	1125/2235	Age, asthma, smoking status

BMI, body mass index; NSAIDs, non-steroidal anti-inflammatory drugs.

Table 2

Assessment of methodological quality of the case–control studies

Study	Selection				Comparability	Exposure			Total
	Definition adequate	Representativeness of the cases	Selection of controls	Definition of controls	Control for important factor or additional factor	Ascertainment of exposure	Same method of ascertainment for cases and controls	Non-response rate	(0–9)
Dublin, 2010	*	*	*	*	*	*	*		7
Euric, 2010	*	*	*	*	*	*	*		7
Gulmez, 2007	*	*	*	*	**	*	*	*	9
Hermos, 2012	*	*	*	*	*	*	*		7
Laheij, 2012	*	*	*	*	**	*	*		8
Rodriguez, 2009	*	*	*	*	**	*	*	*	9
Sarkar, 2008	*	*	*	*	**	*	*	*	9

Note: A ‘star (*)’ system of the NOS has been developed for the methodological quality assessment: each study can be awarded a maximum of one star for each numbered item within the selection and exposure categories, while a maximum of two stars can be given for the comparability category.

Meta-analysis

Our meta-analysis pooled data from 7 of 11 studies included in our systematic review. These seven studies include 967 279 participants and 65 590 cases of CAP. Four studies were not included in the meta-analysis due to substantially overlapping participant populations.

Seven studies were included in the meta-analysis for the current use of PPIs and the risk of CAP, involving 28 289 CAP developed cases among 967 279 participants. When we pooled adjusted OR, the current use of PPI users showed an increased risk of CAP (OR, 1.86; 95% CI, 1.30–2.66) with high heterogeneity (I² = 99%, P < 0.00001). Fig. 2A shows an overview of the current use of PPIs and the risk of CAP in the meta-analysis of case–control studies (N = 7). The size of each square is proportional to the study’s weight. The open diamond is centered at the summary effect estimate from the pool studies with OR and 95% CI.

Figure 2

The use of different PPIs and the risk of CAP in the random-effects model. Note: (A) Current PPI use; (B) recent PPI use; (C) past PPI use.

Our review included three studies for the recent use of PPIs and the risk of CAP, yielding 971 CAP developed cases among 64 279 participants (Fig. 2B). Recent users had 1.66-fold higher risk of CAP (OR, 1.66; 95% CI, 1.22–2.25) with moderate heterogeneity (I² = 72%, P = 0.03).

Meta-analysis for past use of PPIs and risk of CAP included seven studies, involving 36 330 CAP developed cases among 967 279 participants. Past users had a moderate risk of CAP (OR, 1.20; 95% CI, 0.90–1.60) than that of non-users when we pooled adjusted ORs. Fig. 2C provides the overview recent use of PPIs and the risk of CAP in the meta-analysis of case–control studies (N = 3). The size of each square is proportional to the study’s weight. The open diamond is centered at the summary effect estimate from the pool studies with OR and 95% CI.

Subgroup meta-analyses

We performed subgroup analyses to assess the influence of age, comorbidity, co-medication, hospitalization, dosage and duration of PPI exposure on the risk of CAP and to evaluate whether these characteristics could be the possible sources of heterogeneity (Table 3).

Table 3

Summary of subgroup analysis

Features	Number of studies	Pooled effect [OR (95% CI)]	I2 (%)	P value for heterogeneity
Age
<60	3	0.96 (0.91–1.02)	45	0.160
≥60	3	1.02 (0.95–1.08)	51	0.130
Gender
Male	5	2.40 (1.50–3.86)	97	<0.00001
Female	2	0.95 (0.82–1.10)	60	0.120
PPIs duration
<1 month	5	1.94 (1.07–3.51)	98	<0.00001
1–6 months	5	2.05 (1.22–3.45)	99	<0.00001
>6 months	5	1.91 (1.22–3.00)	99	<0.00001
PPIs dose
Low	5	1.67 (0.84–3.30)	99	<0.00001
High	5	2.4 (1.50–3.80)	97	<0.00001
Hospitalization	4	2.59 (1.83–3.66)	97	<0.00001
COPD	5	2.71 (1.96–3.74)	98	<0.00001
CHF	5	2.94 (1.85–4.67)	99	<0.00001
Diabetes	5	1.72 (1.41–2.09)	94	<0.00001
Stroke	3	2.7 (1.14–6.38)	100	<0.00001
Antibiotic	5	3.61 (2.42–5.39)	99	<0.00001
Corticosteroid	5	5.08 (3.29–7.84)	99	<0.00001
Antipsychotic	4	3.23 (1.09–9.57)	100	<0.00001

Five studies assessed separately the impact of low- and high-dose PPI therapy on CAP. There was an increased risk of CAP both in low-dose users (OR, 1.67; 95% CI, 0.84–3.30) with significance heterogeneity (P < 0.00001, I² = 99%) and in high-dose users (OR, 2.40; 95% CI, 1.50–3.86) with significance heterogeneity (P < 0.00001, I² = 97%) (Fig. S1 in appendix).

Four studies estimated the rate of hospitalization due to CAP. There was an enhanced rate of hospitalization of CAP patients with PPI therapy (OR, 2.59; 95% CI, 1.83–3.66) with significance heterogeneity (P < 0.00001, I² = 97%) (Fig. S2 in appendix).

Five studies examined the impact of the duration of PPI therapy and CAP risk (Fig. S3 in appendix). All studies provided risk estimates of shorter duration of PPI therapy, defined as a duration of exposure <30 days, and the pooled OR was 1.94 (95% CI, 1.07–3.51) with significant heterogeneity (P < 0.00001, I² = 98%). In the case of 30–180 days, the pooled OR was 2.05 (95% CI, 1.22–3.45) with significant heterogeneity (P < 0.00001, I² = 99%) while a cumulative duration of exposure of >180 days also showed a positive association with statistically significant (OR, 1.91; 95% CI, 1.22–3.00) and heterogeneity between studies remain same (P < 0.00001, I² = 99%).

The pooled OR for the risk of CAP in the case of the patient <60 years was 0.96 (95% CI, 0.91–1.02) while heterogeneity was estimated (P = 0.16, I² = 45%). The pooled OR for the patients ≥60 years was 1.02 (95% CI, 0.95–1.08) while heterogeneity was preserved (P = 0.13, I² = 51%) (Fig. S4.in appendix). Five studies analyzed the risk of CAP between males and two studies analyzed the risk of CAP in female patients (Fig. S5 in appendix). The pooled OR for the risk of CAP in males was statistically significant 2.40 (95% CI, 1.50–3.86); significant heterogeneity (P < 0.00001, I² = 97%). In contrast, the associated risk of females was no longer statistically significant (pooled OR, 0.95; 95% CI, 0.82–1.10) while heterogeneity between studies further decreased (P = 0.12, I² = 60%).

For the commodities, five studies evaluated separately the association of CAP risk and chronic obstructive pulmonary disease (COPD), cardiac disease and diabetes. The pooled OR was (2.71, 95% CI, 1.96–3.74; 2.94, 95% 1.85–4.67; 1.72, 95% CI 1.41–2.09) respectively with significant heterogeneity. Also, ischemic heart disease (OR, 1.78; 95% CI, 1.24–2.55), renal disease (OR, 3.67; 95% CI, 1.79–7.54) and stroke (OR, 2.70; 95% CI, 1.14–6.38) were positively associated with CAP risk (Fig. S6 in appendix).

Furthermore, in the case of co-medication and the risk of CAP (Fig. S7 in appendix), five studies examined the association between corticosteroid users and CAP risk. The pooled OR was 5.08 (95% CI, 3.29–7.84) with significant heterogeneity (P < 0.00001, I² = 99%). The pooled OR for antibiotics users was 3.61 (95% CI, 2.42–5.39) and the heterogeneity between studies was significant (P < 0.00001, I² = 99%).

Sensitivity analysis

The results of sensitivity analyses are shown in Table S3 in the appendix. To assess whether a single study had a substantial influence on the main results, we included each study and evaluated its effect on the summary estimates and heterogeneity of the main analysis. When we analyzed PPIs along with other gastric medications like H₂A, the pooled OR showed moderate association but not statistical significance (OR, 1.30; 95% CI, 1.08–1.55; P = 0.32, I² = 12%). For the association between H₂A and CAP risk, the pooled OR showed a 1.59-fold higher risk with significant heterogeneity (OR, 1.59; 95% CI, 0.98–2.60; P < 0.00001; I² = 99%).

Publication bias

We found no statistically significant asymmetry in funnel plots, and the result of the Egger test was non-significant. Specifically, the Begg’s funnel plot was not asymmetrical (P < 0.05) for PPIs, indicating no significant influence of publication bias among the included studies.

Discussion

In this systematic review of 11 observational and meta-analysis of 7 observational studies, we identified high association between PPI therapy and the risk of CAP. However, these results should be interpreted with caution due to the significant statistical and clinical heterogeneity among studies. The inherent inability of observational studies is to clarify whether the observed epidemiologic association is a causal effect or a result of unmeasured confounding [12]. When stratified by hospitalization, the use of PPI therapy significantly increased the risk of CAP, whereas there was no significant difference between groups of more or less than 60 years. The risk of CAP differed depending on the duration of PPI therapy. Current users (a month) of PPI therapy had ~2-fold risks of developing CAP when compared with non-users, whereas chronic users (3–6 months) were highly associated with CAP risk. Similarly, long-term use and a low dose of PPIs moderately increased the risk of CAP, but high-dose therapy was significantly higher among the users.

Interestingly, the subgroup meta-analyses by the number of comorbidities showed a significant association. Diabetics is a risk factor of developing CAP, and this result is supported by several studies. It is associated with increased susceptibility to infection and an independent contributor factor [22–24]. COPD is the most common comorbidities of CAP; however, this meta-analysis found a positive association between COPD and CAP. Some studies suggest that COPD should be used as prognostic factors and the measurement of paCO₂ also adds prognostic information in hospitalized patients with CAP [25–27]. Besides, the severity of CAP was accountable for the prescription of corticosteroid drugs, antibiotics and hospitalization for a CAP exacerbation.

PPI therapy may truly increase CAP risk, but the study design and the complexity of patient care limit our ability to identify the expected dose and temporal relationship with adverse outcomes. CAP risk from PPI therapy we observed could represent residual confounding. All included studies were observational studies and most relied on retrospective databases; a higher OR would have been required to prove an association. Adjusting for confounding in observational studies may not be adequate as there could be other unknown confounders for which full adjustment is not possible [12]. In this study, the pooled estimates of unadjusted or least-adjusted risks were much higher than the pooled estimates of the most complete adjusted risks. In words, if further confounders had been identified and adjusted for, the estimated OR may have been even closer to unity. Furthermore, patients treated with PPIs may have a higher CAP risk for other reasons.

Biological plausibility

CAP is an infection of the pulmonary parenchyma, and it may be caused by numerous pathogens including bacteria, viruses, fungi and parasites [11]. CAP covers a group of specific infections, each with its specific clinical features. It is noted that pathophysiological mechanisms, which contribute to an increased risk of CAP during PPI therapy, are not well established. Several supposed mechanisms like anti-infective, anti-inflammatory and immunomodulatory effects could potentially affect the susceptibility to bacterial infections in patients using PPIs including CAP but also enteric infections [28–30]. Laheij et al. mentioned that PPI therapy impairs the immune system leading to an increased susceptibility to infections [31]. The low pH of the intra-gastric environment constitutes a major non-specific defense mechanism of the body against pathogen invasion of the gastrointestinal tract. As the use of PPIs causes to decrease the gastric acidity that may result in insufficient eradication of ingested pathogens through several mechanisms like alteration of the gut microflora, enhanced bacterial translocation altering various immunomodulatory and anti-inflammatory effects [32–34]. Rabe et al. [35] already revealed a reduction of gastric acid secretion led in almost 60% of patients to bacterial overgrowth in the stomach, and particular pathogens were normally found in the oropharyngeal cavity. Several studies confirmed that patients using PPIs more often are being or becoming infected with microorganisms originating from the oropharyngeal cavity [36–38].

Limitations

Our systematic review and meta-analysis have several limitations that need to be addressed. Firstly, there was substantial heterogeneity across the studies in the main analysis of CAP. Although this could be partly explained by different study design, dose and duration of PPI use, we could not thoroughly explore other possible sources of variability due to the small number of included studies. However, we decided that it could be informative to pool these data with a random-effects model, which in part accounts for heterogeneity among studies. Secondly, the evaluation of publication bias was not performed because of a small number of studies. In such cases, the tests for publication bias typically have low power and may be inappropriate [39]. Instead, we attempted to minimize the possibility of publication bias through manual searches of conference abstracts for studies that were not published as full papers. Thirdly, because of inadequate data in the primary studies, we were unable to evaluate differences among different PPIs.

Conclusion

In conclusion, our results suggest a potential association between the use of PPIs and the risk of CAP. However, to confirm or refute the results of our meta-analysis, a large long-term prospective RCT specifically designed to assess the effect of PPIs on CAP is needed. While awaiting further studies, because CAP is associated with substantial morbidity and mortality, clinicians should ensure that PPIs are only prescribed for patients with a clear indication. Conversely, until higher quality evidence is available, the majority of patients appropriately prescribed PPIs should not discontinue this medication solely based on CAP risk.

Funding

This research is sponsored in part by Taipei Medical University under grant TMU107-AE1-B18 and Ministry of Science and Technology project number MOST 108-2410-H-038-010-SSS and MOST107-2218-E-038-004-MY2.