Laheij RJF, Sturkenboom MCJM, Hassing R, Dieleman J, Stricker BHC, Jansen JBMJ. Risk of Community-Acquired Pneumonia and Use of Gastric Acid–Suppressive Drugs. JAMA. 2004;292(16):1955-1960. doi:10.1001/jama.292.16.1955
Author Affiliations: Department of Gastroenterology,
University Medical Center St. Radboud, Nijmegen, the Netherlands (Drs Laheij
and Jansen); and Department of Medical Informatics (Drs Laheij, Sturkenboom,
Dieleman, and Stricker and Mr Hassing), Pharmacoepidemiology Unit, Department
of Epidemiology and Biostatistics (Drs Sturkenboom and Stricker), and Internal
Medicine (Dr Dieleman), Erasmus MC University, Medical Center Rotterdam, Rotterdam,
Context Reduction of gastric acid secretion by acid-suppressive therapy allows
pathogen colonization from the upper gastrointestinal tract. The bacteria
and viruses in the contaminated stomach have been identified as species from
the oral cavity.
Objective To examine the association between the use of acid-suppressive drugs
and occurrence of community-acquired pneumonia.
Design, Setting, and Participants Incident acid-suppressive drug users with at least 1 year of valid database
history were identified from the Integrated Primary Care Information database
between January 1, 1995, and December 31, 2002. Incidence rates for pneumonia
were calculated for unexposed and exposed individuals. To reduce confounding
by indication, a case-control analysis was conducted nested in a cohort of
incident users of acid-suppressive drugs. Cases were all individuals with
incident pneumonia during or after stopping use of acid-suppressive drugs.
Up to 10 controls were matched to each case for practice, year of birth, sex,
and index date. Conditional logistic regression was used to compare the risk
of community-acquired pneumonia between use of proton pump inhibitors (PPIs)
and H2-receptor antagonists.
Main Outcome Measure Community-acquired pneumonia defined as certain (proven by radiography
or sputum culture) or probable (clinical symptoms consistent with pneumonia).
Results The study population comprised 364 683 individuals who developed
5551 first occurrences of pneumonia during follow-up. The incidence rates
of pneumonia in non–acid-suppressive drug users and acid-suppressive
drug users were 0.6 and 2.45 per 100 person-years, respectively. The adjusted
relative risk for pneumonia among persons currently using PPIs compared with
those who stopped using PPIs was 1.89 (95% confidence interval, 1.36-2.62).
Current users of H2-receptor antagonists had a 1.63-fold increased
risk of pneumonia (95% confidence interval, 1.07-2.48) compared with those
who stopped use. For current PPI users, a significant positive dose-response
relationship was observed. For H2-receptor antagonist users, the
variation in dose was restricted.
Conclusion Current use of gastric acid–suppressive therapy was associated
with an increased risk of community-acquired pneumonia.
Gastrointestinal symptoms are common: annually, 20% to 40% of the general
population has at least 1 episode of dyspepsia or gastroesophageal reflux
disease, and 5% consult a general practitioner for these complaints.1,2 The most effective treatment strategy
for these symptoms in primary care is reduction of gastric acid secretion,
which can be achieved by using H2-receptor antagonists (H2RAs) or proton pump inhibitors (PPIs).3 Currently,
a common approach in western countries is to prescribe acid-suppressive drugs
for upper gastrointestinal tract symptoms without suspicion of a malignancy
and to refer nonresponders for gastrointestinal endoscopy.4 The
consequence of this policy is that acid-suppressive drugs are among the most
frequently prescribed drugs, whereas their use is not without risk.
H2-receptor antagonists and PPIs increase susceptibility
to infections by increasing gastric pH.5,6 Intragastric
acidity constitutes a major nonspecific defense mechanism of the stomach to
ingested pathogens. In normal gastric juice with a pH below 4, most pathogens
are promptly killed, whereas they survive in hypochlorhydric to achlorhydric
circumstances. For the effective management of upper gastrointestinal tract
symptoms, the intragastric pH should, however, be maintained above 4 for at
least 18 hours. Treatment with acid-suppressive drugs may therefore lead to
an insufficient elimination, or even increased colonization, of ingested pathogens.7- 12 There
is some evidence that acid-suppressive therapy facilitates nosocomial infections.13- 15
To our knowledge, there are no large-scale studies of the association
between use of acid-suppressive drugs and the risk of infections. We examined
the association between the use of gastric acid–suppressive drugs and
community-acquired pneumonia in a population-based cohort study.
All data were retrieved from the Integrated Primary Care Information
(IPCI) project, a general practice research database containing data from
electronic patient records of a group of about 150 general practitioners in
the Netherlands. Details of the database have been described.16,17 Briefly,
the database contains the complete medical records of approximately 500 000
patients. The electronic records contain coded and anonymous data on patient
demographics, reasons for visit (in free text), diagnoses (using the International
Classification for Primary Care18 and free
text) from general practitioners and specialists, referrals, laboratory findings,
hospitalizations, and drug prescriptions, including their indications and
dosage regimen. To maximize completeness of the data, general practitioners
participating in the IPCI project are not allowed to maintain a system of
paper-based records aside from the electronic medical records. The system
complies with European Union guidelines on the use of medical data for medical
research and has been proven valid for pharmacoepidemiologic research.19 The Scientific and Ethical Advisory Board of the
IPCI project approved the study.
The study started on January 1, 1995, and ended on December 31, 2002.
The source population comprised all individuals with at least 1 year of valid
database history, which means that the general practitioner supplied standard
data for at least 1 year and the patient was registered for 1 year with the
general practitioner. We required this preenrollment period to be able to
characterize the patient and verify previous use of drugs and a history of
pneumonia. All individuals were followed up from the moment they had 1 year
of valid history until one of the following events: pneumonia, death, moving
out of the practice area, or end of the study period, whichever came first.
To retain patients without recent occurrence of pneumonia, we excluded all
individuals who had a diagnosis of pneumonia in the preenrollment period.
From the source population, a study cohort of incident (no use in the
year before enrollment) acid-suppressive drug users (those with at least 1
H2RA or PPI prescription) was identified. From this exposure cohort,
we excluded all individuals who received acid-suppressive drugs in combination
with antibiotics to eradicate Helicobacter pylori infection.
The duration of use of individual acid-suppressive drugs was calculated from
the prescribed quantity and prescribed dosing regimen. Person-time of exposure
was accumulated during follow-up time for calculation of incidence rates.
Individuals who never used acid-suppressive drugs before or during the study
period were considered unexposed; follow-up time was accumulated for calculation
of incidence rates.
The first occurrence of pneumonia for each individual was identified
through searches on diagnoses and free-text indicators of pneumonia. The medical
records of all potential cases were reviewed manually to classify the pneumonia
as certain (proven by thorax radiography or microbiological culture), probable
(clinical symptoms consistent with pneumonia but no objective evidence), possible,
or no pneumonia. Definitions of pneumonia vary widely. Some require only the
presence of infiltrates on chest radiography, whereas others require only
certain symptoms or signs. In our analysis, we included pneumonia proven by
chest radiography or sputum culture (certain) or presence of respiratory symptoms
(probable). Cases with certain or probable pneumonia caused by aspiration
(n=5), obstruction (n=6), or nosocomial infection (n=13) were excluded. The
date of first pneumonia was defined as the index date.
To reduce confounding by indication, a nested case-control analysis
was conducted within the cohort of persons who used acid suppressants during
the study period. For each case of pneumonia, we randomly selected up to 10
controls from the cohort, matched on sex, year of birth, and index date to
the case. Exposure to H2RAs and PPIs was classified separately
by time since last use. Drug use was defined as current if the prescription
length covered the index date and as past if the end of the last prescription
was before the index date. Past use was further categorized into recent past,
past, and distant past if the end of the last prescription was less than 30
days ago, between 30 and 180 days ago, and more than 180 days ago, respectively.
Because all persons had used an acid-suppressive drug during the study
period, we had no unexposed subjects. In analyses of the dose and duration
effects of H2RAs and PPIs separately, we restricted the cases and
controls to individuals who never used the other type of acid suppressant.
Among current users of H2RAs or PPIs, we studied the active compound
(cimetidine, famotidine, nizatidine, ranitidine, roxatidine, omeprazole, esomeprazole,
lansoprazole, pantoprazole, rabeprazole), the daily dosage (less than, equal
to, and more than the defined daily dose), and the duration of use (<14
days, 14-28 days, 28-42 days, and >42 days). The defined daily dose for the
PPIs omeprazole, esomeprazole, lansoprazole, pantoprazole, and rabeprazole
is 20, 20, 30, 40, and 20 mg, respectively. The defined daily dose for the
H2RAs cimetidine, famotidine, nizatidine, ranitidine, and roxatidine
is 800, 40, 300, 300, and 150 mg, respectively.19
As covariates in the case-control analysis, we considered age, sex,
and calendar time (matching factors), the indication for therapy (cancer,
peptic ulcer disease and gastroesophageal reflux disease with or without esophagitis,
functional dyspepsia, and endoscopically uninvestigated), diabetes mellitus,
heart failure, chronic obstructive lung disease (asthma and chronic obstructive
pulmonary disease), lung cancer, stomach cancer, number of physician visits
in the year before, use of antibiotics, and use of systemic immunosuppressive
agents (glucocorticoids, cyclosporine). In patients empirically treated with
gastric acid–suppressive drugs and endoscopically uninvestigated, the
indication for therapy was unknown and may have varied from relevant organic
disease (peptic ulcer disease and reflux esophagitis) to no organic abnormalities
(ie, functional dyspepsia). The indication for the use of acid-suppressive
drugs was obtained from the prescription records. For patients with more than
1 indication, the endoscopically verified diagnosis was used.
Crude incidence rates of pneumonia in unexposed and exposed individuals
were calculated by dividing the number of pneumonia cases by the corresponding
person-years. Relative risks of pneumonia (plus 95% confidence intervals [CIs])
during current and recent use of H2RAs and PPIs were estimated
with odds ratios (ORs) by using conditional logistic regression analysis adjusted
for all covariates that were univariately associated with pneumonia (P<.10). Stratified analyses were conducted to explore
the effect of season and age as effect modifiers. In addition, we stratified
for presence of cancer.
A sensitivity analysis that excluded all probable cases was conducted
to examine the effect of outcome misclassification. We then calculated the
population attributable risk percentage. The attributable risk percentage
is the percentage of exposed cases that can be attributed to exposure, which
is calculated [(OR−1)/OR]. The attributable risk percentage (obtained
from adjusted ORs) was subsequently multiplied with the incidence rate ratio
in PPI- and H2RA-exposed individuals to obtain the expected incidence
rate attributable to exposure. Because this rate is per 100 person-years and
we express number needed to harm in number of persons, we calculated how many
persons were necessary for 100 person-years of exposure by applying the mean
duration of exposure. All analyses were conducted in SPSS/PC, version 11 (SPSS
Inc, Chicago, Ill). The level of significance for all statistical tests was
The source population comprised 364 683 persons who had on average
2.7 years of follow-up and 5551 first occurrences of pneumonia (Table 1). Eighteen percent of the 5551 pneumonia occurrences were
confirmed by radiography or microbiological testing. During the study period,
19 459 individuals received a first prescription for acid-suppressive
drugs, 10 177 H2RAs and 12 337 PPIs (some individuals
used both). The mean duration of use for H2RAs was 2.8 months;
for PPIs, 5.0 months. Most persons had not undergone endoscopy and were empirically
treated for upper gastrointestinal tract symptoms.
In the exposed cohort, 185 persons developed a first pneumonia occurrence
during use of acid-suppressive drugs and 292 after stopping their use. The
incidence rate during use of PPIs was 2.5 per 100 person-years and during
use of H2RAs, 2.3 per 100 person-years compared with 0.6 for nonusers.
Patients using acid-suppressive drugs developed pneumonia 4.5 (95% CI, 3.8-5.1)
times more often compared with those who never used acid-suppressive drugs.
This association measure was not adjusted for potential confounders.
For the nested case-control analysis, 475 of the 477 patients who developed
pneumonia during or after stopping acid-suppressive drug use could be matched
to a total of 4690 controls (31% of the pneumonia cases were confirmed by
chest radiography or sputum test). Two cases could not be matched and were
therefore excluded from the analysis. Cases more often had diabetes mellitus,
heart failure, and pulmonary diseases; more frequently used immunosuppressants;
and more frequently had used antibiotics than did controls in the previous
year (Table 2). The indication for acid-suppressive
therapy was not associated with the risk of pneumonia.
To investigate the association between use of acid-suppressive drugs
overall and pneumonia, we first examined all acid-suppressive drugs together.
Current use of acid-suppressive drugs was associated with a small increase
in the risk of pneumonia (adjusted OR, 1.27; 95% CI, 1.06-1.54). To examine
the effects of current use of H2RAs and PPI separately, we considered
current combined use of both compounds in a separate group. Because the association
between pneumonia and past or distant past use of acid-suppressive drugs was
similar, we combined these 2 categories to have a more stable reference category.
The adjusted relative risk for pneumonia among persons currently using PPIs
compared with those who stopped using PPIs was 1.89 (95% CI, 1.36-2.62). Current
users of H2RAs had a 1.63-fold increased risk of pneumonia (95%
CI, 1.07-2.48) compared with those who stopped.
Although there was variation between individual PPIs and H2RAs,
numbers were small and the heterogeneity was not significant. A significant
dose response was observed in current users of PPIs. Persons using more than
1 defined daily dose had a 2.3-fold increased risk of pneumonia compared with
past use of acid suppressants (Table 3).
The dose response was not observed for H2RAs, but the variation
in dose of these drugs was limited. Among current users of PPIs or H2RAs, the risk seemed most pronounced among persons who started drug
use within the last 30 days (Table 3).
Excluding pneumonia without a laboratory confirmation led to higher
risk estimates for developing pneumonia: 2.2 (95% CI, 1.4-3.5) for PPIs and
1.7 (95% CI, 0.8-2.9) for H2RAs. There was no significant effect
modification by age, season, or presence of cancer.
The adjusted attributable risk percentage is 42% for PPIs and 37% for
H2RAs. Therefore, 1.05 pneumonia cases per 100 person-years of
PPI exposure can be attributed directly to the use of PPIs and 0.86 pneumonia
cases during 100 person-years of H2RA exposure. Because the average
duration of use was 0.23 years for H2RAs and 0.42 years for PPIs,
this roughly translates to 1 case of pneumonia per 226 patients treated with
PPIs and 1 case of pneumonia per 508 persons treated with H2RAs.
In this large cohort, current use of acid-suppressive drugs was associated
with an increased risk of community-acquired pneumonia. The increase in risk
was most pronounced for PPIs and showed a clear dose-response relationship,
which supports a real biological effect. The results are in agreement with
a previous small study that was conducted ad hoc in a clinical setting. This
study showed that subjects using acid-suppressive drugs more often reported
clinical manifestations of respiratory tract infections and complications
compared with those who did not use acid-suppressive drugs.15
Gastric acid is an important barrier against pathogen invasion through
the gastrointestinal tract. Although it is well known that a raised pH increases
bacterial and virus colonization, the clinical consequences of gastrointestinal
pathogen overgrowth have not been convincingly demonstrated.5,6 Reports
on the clinical consequences have been only anecdotal or nonclinical. Already
in 1934, Hurst20 suggested that bacillary and
amoebic dysentery occurred much more frequently in subjects with achlorhydria
or hypochlorhydria. More recent studies suggested that acid-suppressive drugs
might be responsible for the development of esophageal candidiasis and enteric
infections, although this was not solidly supported by clinical evidence.10,13,14,20- 22 Furthermore,
experimental evidence suggests that acid-suppressive drugs inhibit polymorphonuclear
neutrophil functions and cytotoxic T lymphocyte and natural killer cell activity,23- 26 which
might add to the increased susceptibility to infection because of these drugs.
Studies in mechanically ventilated patients support the results from
our study that the use of acid-suppressive drugs modifies the risk of pneumonia.
These studies showed that during mechanical ventilation, intestinal pathogens
colonize the oral space via the stomach.13,14 The
colonized secretions may gain access to the lower airways and cause lower
respiratory infections. Aspiration of gastric acid itself carries a risk of
developing chemical pneumonia.27 Backflow of
gastric acid and content into the esophagus because of incompetent barriers
at the gastroesophageal junction is a prevalent gastrointestinal disorder.
Reflux of gastric contents into the lower esophagus even occurs in the majority
of healthy individuals but usually does not result in clinical sequelae.
In this study, we were able to take advantage of the fact that in the
Dutch health care system, all medical information is prospectively collected
at general practices that cover the total population instead of subjects presenting
in a clinical setting. As a consequence, the data are generalizable to the
general population and are not prone to selection bias. Nevertheless, given
its observational nature, this study should be interpreted in the light of
its limitations. First, we cannot exclude that some misclassification of outcome
occurred. Such misclassification might be false negative or false positive.
False-negative misclassification by underestimation of pneumonia may have
occurred because of the exclusion of the possible cases (ie, cases for which
we had insufficient diagnostic information). False-positive misclassification
was possible in the group of probable pneumonia; most of the mild pneumonia
is dealt with in primary care and therefore is not confirmed by chest radiograph
or microbiological testing. Although we classified as probable pneumonia only
infections with all clinical symptoms of pneumonia, we may have included some
patients with bronchitis. When we excluded all probable cases from the analysis,
the detected associations became stronger. Diagnostic bias could have occurred
if patients who were taking acid-suppressive drugs had better diagnostic evaluation
than distant-past users; however, the percentage of certain cases confirmed
by radiograph or sputum was similar for current users (32%) and distant-past
Second, misclassification of exposure may have occurred because we used
outpatient prescription data and had no information about whether the prescription
was actually dispensed and taken. It is likely, however, that such exposure
misclassification was random and evenly distributed among cases and controls.
Third, one could suspect that pneumonia would be diagnosed especially
when the patient was treated by the general practitioner, the so-called diagnostic
classification bias. However, the results from our study showed that the risk
for pneumonia was increased in current drug users (pneumonia was reported
during prescription length) but also in patients for whom the last prescription
was at least 30 days old (recent past). These patients were probably no longer
being treated by the general practitioner, making classification bias less
Finally, protopathic bias (ie, pneumonia symptoms) may have resulted
in an acid-suppressive-drug prescription if patients received acid-suppressive
drugs for symptoms related to pneumonia. However, on exclusion of all persons
who began receiving acid-suppressive drugs within 1 week before the index
date, the relative risk estimates did not change, and therefore protopathic
bias can be excluded.
We observed a large difference between the unadjusted incidence rate
ratios and the adjusted ORs in the nested case-control study. The incidence
rate ratio should be interpreted with caution because it compares users of
acid-suppressive drugs (with more comorbidity) with nonusers, which may result
in substantial confounding by indication. In fact, we designed the nested
case-control study to control for this confounding effect. As confounding
factors, we included other respiratory illnesses, comorbidities, and drug
use that are strongly associated with pneumonia. Inclusion of these confounders
in the model did not change the estimates to a large extent, which implies
that confounding was minimal. Despite this outcome, we cannot exclude the
presence of uncontrolled confounders, which is a limitation inherent in all
The effectiveness of acid-suppressive drugs in the treatment of upper
gastrointestinal tract symptoms is excellent. Acid-suppressive drugs nevertheless
seem to have some significant drawbacks. Persons using acid-suppressive drugs
more often develop a community-acquired pneumonia compared with those who
do not use acid-suppressive drugs, which is in general not a problem because
the risk for developing pneumonia is low. The increased risk for pneumonia
is a problem for patients who are at increased risk for infection, especially
because community-acquired pneumonia is potentially dangerous.28 Groups
of persons who are at increased risk for infection and for whom pneumonia
is a major source of mortality have been identified.29 Pneumonia
is more pronounced in persons with asthma or chronic obstructive lung disease,
immunocompromised persons, children, and elderly persons. Elderly patients
are likely to incur severe infection, which is partly due to a decreased immune
response, including the natural reduction of gastric acid secretion after
age 60 years. To avoid the calculated excess pneumonia, patients with asthma
or chronic obstructive lung disease, immunocompromised persons, children,
and elderly persons should be treated with acid-suppressive drugs only when
necessary and with the lowest possible dose.
In conclusion, our results suggest that acid-suppressive drugs such
as H2RAs and PPIs are associated with an increased risk of community-acquired
pneumonia, probably because of reduction of gastric acid secretion, facilitating
Corresponding Author: Robert J. F. Laheij,
PhD, University Medical Center St. Radboud, Department of Gastroenterology,
PO Box 9101, 6500 HB, Nijmegen, the Netherlands (R.Laheij@mdl.umcn.nl).
Author Contributions: Dr Laheij had full access
to all of the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analysis.
Study concept and design: Laheij, Sturkenboom.
Acquisition of data: Laheij, Sturkenboom, Hassing.
Analysis and interpretation of data: Laheij,
Sturkenboom, Dieleman, Stricker, Jansen.
Drafting of the manuscript: Laheij, Sturkenboom,
Critical revision of the manuscript for important
intellectual content: Hassing, Dieleman.
Statistical analysis: Laheij, Sturkenboom,
Administrative, technical, or material support:
Study supervision: Jansen.