Content uploaded by Murray Ehrinpreis
Author content
All content in this area was uploaded by Murray Ehrinpreis on Aug 10, 2015
Content may be subject to copyright.
© 2012 by the American College of Gastroenterology The American Journal of GASTROENTEROLOGY
nature publishing group 1001
REVIEW
CLINICAL AND SYSTEMATIC REVIEWS
INTRODUCTION
Clostridium di cile is a common spore-forming bacillus that
can cause gastrointestinal illness, ranging in severity from mild
diarrhea to fulminant colitis and even death. It is the most com-
mon infectious cause of health care – associated diarrhea in devel-
oped countries ( 1 ).
e recent epidemiological changes in C. di cile -associated
diarrhea (CDAD) are notable for an increasing incidence, more
virulent strains, and the identi cation of new “ at risk ” populations
other than traditionally recognized groups such as the elderly and
patients with previous illnesses ( 1 – 3 ). Although antibiotic use
remains the dominant risk factor for CDAD, other documented
risk factors include advancing age, severe underlying illness, hos-
pitalization, the use of naso-gastric tubes, anti-neoplastic chemo-
therapy, and immunosuppressants ( 4 – 6 ).
As the incidence and the severity of the disease increase, an
expenditure of $ 3 billion dollars is estimated to be the annual costs
associated with the treatment of CDAD ( 1,3,7 – 9 ). e prevalence
Clostridium diffi cile -Associated Diarrhea and Proton
Pump Inhibitor Therapy: A Meta-Analysis
S a i l a j a h J a n a r t h a n a n , M D 1 , I v o D i t a h , M D , M P h i l 1 , D o u g l a s G . A d l e r , M D 2 a n d M u r r a y N . E h r i n p r e i s , M D 1
OBJECTIVES: Clostridium diffi cile -associated diarrhea (CDAD) is a major cause of morbidity and increasing
health-care costs among hospitalized patients. Although exposure to antibiotics remains the most
documented risk factor for CDAD, attention has recently been directed toward a plausible link with
proton pump inhibitors (PPIs). However, the results of studies on the association between CDAD
and PPIs remain controversial. We have conducted a meta-analysis to summarize the association
between PPIs and CDAD among hospitalized patients.
METHODS: A systematic search of published literature on studies that investigated the association between
PPIs and CDAD from 1990 to 2010 was conducted on Medline and PubMed. The identifi ed
articles were reviewed for additional references. The most adjusted risk estimates were extracted
by two authors and summarized using random effects meta-analysis. We also conducted a
subgroup analysis by study design. Publication bias was evaluated using the Begg and Egger
tests. A sensitivity analysis using the Duval and Tweedie “ trim-and-fi ll ” method has also been
performed.
RESULTS: Twenty-three studies including close to 300,000 patients met the inclusion criteria. There was
a 65 % (summary risk estimate 1.69 with a 95 % confi dence interval (CI) from 1.395 to 1.974;
P < 0.000) increase in the incidence of CDAD among patients on PPIs. By study design, whether
case – control study (17) or cohort study (6), there was still a signifi cant increase in the incidence of
CDAD among PPI users. The risk estimates were 2.31 (95 % CI from 1.72 to 3.10; P < 0.001) and
1.48 (95 % CI from 1.25 to 1.75; P < 0.001) for cohort and case – control studies, respectively.
CONCLUSION: There is suffi cient evidence to suggest that PPIs increase the incidence of CDAD. Our meta-analysis
shows a 65 % increase in the incidence of CDAD among PPI users. We recommend that the routine
use of PPIs for gastric ulcer prophylaxis should be more prudent. Establishing a guideline for the use
of PPI may help in the future with the judicious use of PPIs. Further studies, preferably prospective,
are needed to fully explore the association between PPIs and CDAD.
Am J Gastroenterol 2012; 107:1001 – 1010; doi: 10.1038/ajg.2012.179; published online 19 June 2012
1 Division of Gastroenterology, Department of Internal Medicine, Wayne State University School of Medicine, Harper University Hospital , Detroit , Michigan ,
USA ;
2 Division of Gastroenterology and Hepatology, Department of Internal Medicine, University of Utah School of Medicine , Salt Lake City , Utah , USA .
Correspondence: Douglas G. Adler, MD , Director of Therapeutic Endoscopy, Division of Gastroenterology and Hepatology, Department of Internal Medicine,
University of Utah School of Medicine , 30N 1900E 4R118, Salt Lake City , Utah 84132 , USA . E-mail: douglas.adler@hsc.utah.edu
Received 29 March 2011; accepted 6 December 2011
see related editorial on page 1020
The American Journal of GASTROENTEROLOGY VOLUME 107 | JULY 2012 www.amjgastro.com
1002
REVIEW
Janarthanan et al.
of CDAD has increased steadily, not only in hospitals and nursing
homes, but also in the outpatient setting ( 4,7 ).
Recently, proton pump inhibitors (PPIs) have come under
intense scrutiny because of a possible association between these
agents and the development of CDAD. PPIs are the mainstay of
therapy in acid-related disorders, including gastroesophageal
re ux disease and peptic ulcer disease. ey are among the most
commonly prescribed medications in the outpatient and inpa-
tient settings worldwide ( 10 ). PPIs are generally thought to have a
safe side-e ect pro le and this has led to widespread use by
clinicians ( 10 ).
Gastric acid is important in eliminating ingested pathogens
from the digestive tract. It is thus biologically plausible that sup-
pressing gastric acidity may result in an increased load of patho-
genic microbes in the gastrointestinal tract. Both human and
animal studies have shown that increased gastric acidity is e ec-
tive in killing C. di cile and neutralizing its toxin. It has also been
shown that there were signi cant di erences in epithelial damage,
edema, and neutrophil in ltration in colons when PPIs were
used as opposed to when they were not used ( 11,12 ). Given the
millions of individuals on PPIs, even a slight increase in the risk
of CDAD conferred by these drugs could have major public health
implications.
Several studies have examined the association between PPI use
and risk of CDAD, with con icting results. A systematic review
done in 2007 has shown a positive association with acid suppres-
sion and CDAD ( 13 ). e study has focused on identifying the
risk of enteric infection in patients taking acid suppression. ey
showed a stronger association to CDAD with PPIs compared with
H2 receptor blocker therapy.
In this manuscript we present a meta-analysis to evaluate the
magnitude and the direction of the association between gastric
acid suppression with PPIs and CDAD.
METHODS
Search strategy
Two reviewers (SJ and ID) independently and in duplicate con-
ducted a systematic literature search on MEDLINE and PubMed
databases from 1990 through December 2010. Search terms
included “ proton pump inhibitor, ” “ acid suppressive therapy, ”
individual PPI generic names, combined with “ Clostridium dif-
cile infection ” , “ diarrhea ” , “ colitis ” “ gastrointestinal infection ” ,
and “ pseudomembranous colitis. ” e title and abstract of eligible
studies were then reviewed to exclude any that were irrelevant to
the research question. British spelling terms were also searched to
increase search yield. is process included electronic searching
of supplementary abstracts published in various gastroenterology
and other journals. e titles and abstracts of studies identi ed
in the search were reviewed to exclude any that were irrelevant to
the research question. e full text of the remaining articles was
read and their bibliographic lists reviewed for additional publi-
cations on the subject. No language restriction was used in the
search lter. We did not include data presented only as abstracts
at conferences.
Study references and citations were collected in Endnote so ware
application version 8.0 ( omson Reuters, New York, NY). A data
collection form was designed in Microso Access 2007 (Microso ,
Redmond, WA). Abstracted data from each study included: rst
author ’ s last name, year of publication, country of the population
studied, study design, age range, number of participants, duration
of PPI use, e ect estimates (odds ratio (ORs) or risk ratios (RRs)),
and 95 % con dence intervals (CIs) of PPI exposure ( Table 1 ). e
disease exposure was de ned as acute diarrhea with toxin con r-
mation. In studies reporting several e ect estimates, we extracted
that with the greatest degree of control for potential confounders.
Inter-extractor discrepancies were resolved by referring to the
original article.
Study quality assessment
Our study quality assessment included screening for studies
according to the Meta-Analysis of Observational Studies in Epi-
demiology (MOOSE) criteria. Studies were statistically adjusted
for confounding factors by the primary researchers. e criteria
for inclusion consisted of observational studies (both case – con-
trol and cohort), adult populations who received PPI therapy for
at least 3 months prior to having acute onset diarrhea with labora-
tory C. di cile con rmation, and who were adjusted for antibi-
otic use. ese moderate quality studies were included in the nal
analysis. e low-quality studies with poorly de ned population,
such as C. di cile diagnosis on registry with no laboratory con-
rmation and not having been adjusted for antibiotic use, were
excluded. No study was high quality as none were randomized
controlled clinical trials.
Statistical analysis
We included in this meta-analysis the studies reporting di erent
risk estimates: case – control studies (ORs), and cohort studies (RRs).
In practice, these measures of e ect would be expected to yield
very similar RR estimates, given that the absolute risk of CDAD
is low ( 14 ). However, all ORs in this study were converted to RRs
using the formula RR = OR / (1 − OR) + (Po × OR), where Po is the
risk of the outcome / event in the control group ( 15 ).
All studies were initially analyzed together and were subse-
quently grouped on the basis of study design (cohort vs. case –
control). is was done to examine consistency of results, and in
an attempt to explain the presence of heterogeneity between all the
studies. We used the DerSimonian-Laird method, which assumes
a random-e ects model to calculate the pooled e ect estimate
( 16 ). In the presence of heterogeneity, the random-e ects model is
recommended by the Cochrane collaboration, because its assump-
tions account for the presence of variability among studies ( 17 ).
erefore, reported estimates are from a random-e ects model.
We used Forest plots to summarize the results. Publication bias
was assessed using the Begg ’ s ( 18 ), and Egger ’ s ( 19 ) regression
asymmetry test. Furthermore, Duval and Tweedie ’ s nonparametric
“ trim-and- ll ” rank-based technique, which formalizes the use of
funnel plots and estimates, and adjusts for the number of missing
studies, as well as the outcomes of the missing studies, was per-
formed ( 20 ). A comprehensive meta-analysis was done to depict
© 2012 by the American College of Gastroenterology The American Journal of GASTROENTEROLOGY
1003
REVIEW
C. difficile -Associated Diarrhea and PPI Therapy
Table 1 . The studies and their patient population, sample sizes, and study types
Author Year Country Study design
Study
type
Age
range
Duration of
PPI use Sample RR / OR 95 % CI
Inpatient /
community
Kim et al.
( 30 )
2010 South Korea Retrospective analysis of
records of 125 patients
admitted to hospital in
Seoul, South Korea, from
January 2006 to December
2007
Case –
control
53 – 81 Not
specifi ed
125 2.36 1.79; 3.11 Inpatient
Howell
et al. ( 31 )
2010 USA Pharmaco – epidemiologic
cohort study, performing a
secondary analysis of data
collected prospectively on
101,796 patients during a
period of 5 years in Boston,
Massachusetts
Cohort All age
groups
Not
specifi ed
101,796 1.74 1.39; 2.18 Inpatient
Linsky et al.
( 10 )
2010 USA Retrospective cohort study
using administrative data-
bases of the New England
Veterans Health care system
(1,166 patients involved,
527 received PPI and 639
did not)
Retro-
spective
cohort
All age
groups
< 3 Months 1,166 1.42 1.11; 1.82 Inpatient
and
community
Turco et al.
( 29 )
2010 Italy Pediatric hospital-based.
Overall, 68 cases of CDAD
was paired with 68 control
in Naples, Italy
Case –
control
1.1 – 17.8 Not
specifi ed
136 4.5 1.4; 14.4 Inpatient
Bajaj et al.
( 32 )
2010 USA Review of two data sets:
National inpatient sample
2005, and record of patients
admitted to Froedtert
Hospital in Wisconsin from
January 2002 to December
2006, using a case – control
design (60,194 cases vs.
82,065 controls)
Case –
control
46 – 89 Not
specifi ed
83,230 37.6 6.22;
227.8
Community
Debast
et al. ( 33 )
2009 The Nether-
lands
Retrospective analysis
of records of patients
admitted to a St Jansdal
Hospital in Harderwirjk,
the Netherlands, during
an epidemic of CDAD
(45 cases vs. 90 controls)
Case –
control
>18 Not
specifi ed
135 1.6
0.8; 3.4 Inpatient
Dalton
et al. ( 25 )
2009 Canada Retrospective cohort study
conducted at two hospitals
in Canada, over a period
of 37 months (84 / 66 CDAD
cases (exposure + / − )
vs. 5,688 / 8,882 controls
(exposure + / − ) were
involved)
Cohort
study
51 – 85 < 2 Months 20,490 1.96 1.42; 2.72 Inpatient
Aseeri
et al. ( 34 )
2008 USA Retrospective analysis of
patients admitted to Wesley
Medical Center Wichita,
Kansas, who developed
CDAD during hospitalization
(94 cases matched to 94
controls; 61 cases and 36
controls had received PPI)
Case –
control
All age
groups
Not speci-
fi ed
188 3.6 1.7; 8.3 Inpatient
Dubberke
et al. ( 26 )
2007 USA Retrospective cohort study
of 36,086 patients admitted
to Barnes-Jewish Hospital,
Missouri, from 1 January
2003 to 31 December 2003
Cohort
study
All age
groups
Not speci-
fi ed
36,086 1.6 1.3; 2.1 Inpatient
The American Journal of GASTROENTEROLOGY VOLUME 107 | JULY 2012 www.amjgastro.com
1004
REVIEW
Janarthanan et al.
Table 1 . Continued
Author Year Country Study design
Study
type
Age
range
Duration of
PPI use Sample RR / OR 95 % CI
Inpatient /
community
Baxter
et al. ( 35 )
2008 USA Analysis of data extracted
from records of hospital-ac-
quired CDAD from 1999 to
2005 at Kaiser Permanente
in North Carolina (1,142
cases matched with 3,351
controls)
Case –
control
All age
groups
Not
specifi ed
4,493 1.23 1.03; 1.48 Inpatient
Cadle et al.
( 36 )
2007 USA Data for study obtained from
electronic patient medical
records from Michael E.
DeBakey Veterans Affairs
Medical Center. Total 140
patients were involved: 138
men and 2 women
Case –
control
52 – 82 Not
specifi ed
140 4.17 1.66;
10.38
Inpatient
Beaulieu
et al. ( 52 )
2007
Canada Hospital-based cohort study
including patients in Montreal
between 14 March 2002 and
31 May 2004 (827 record
where eligible for analysis)
Cohort
study
>65 Not
specifi ed
827 0.9 0.59;1.38 Inpatient
Jayatilaka
et al. ( 37 )
2009 USA Data obtained from compu-
ter-based admissions
from 1 January 2001 to
31 December 2005 at
St Joseph’s Medical Center
in New Jersey
Case –
control
>18 Not
specifi ed
130 2.75 1.68; 4.52 Inpatient
Akhtar
et al. ( 38 )
2007 UK Retrospective analysis of
records of African-American
and Hispanics admitted to
hospital over a 9 year period
with the diagnosis of diarrhea
(274 cases vs. 169 controls)
Case –
control
19 – 101 Not
specifi ed
12,90 2 1.6; 2.6 Inpatient
Lowe et al.
( 44 )
2006 Canada A population-based, nested
case – control study of linked
health-care databases in
Ontario, Canada, from 1 April
2002 to 31 March 2005
Case –
control
>65 90 Days
to 1 year
13,692 0.9 0.8; 1.1 Community
Dial et al.
( 39
)
2006 UK Review of data from the
United Kingdom’s General
Practice Research Database
(GPRD) from 1 January 1994
to 31 December 2004 (317
cases vs. 3,167 controls)
Case –
control
34 – 84 Not
specifi ed
3,484 3.5 2.3; 5.2 Community
Dial et al.
( 40 )
2005 UK Conducted two population-
based case – control studies
using the UK GPRD (1,233
cases matched with 12,330
controls)
Case –
control
All age
groups
Not
specifi ed
13,563 2.9 2.4; 2.4 Community
Dial et al.
( 24 )
2004 Canada Hospital-based cohort
study, including patients in
Montreal, over a period of
9 months plus case – control
study in another hospital
setting. Cohort study involved
1,187 inpatients, while
case – control study involved
94 cases matched to 94
controls
Case –
control
and
cohort
com-
bined
62 – 88 Not
specifi ed
1,187 2.1 1.2; 24 Inpatient
Yearsley
et al. ( 28 )
2006 UK Prospective case – control
study of consecutive hospital
inpatients in South Wales, UK
(155 cases vs. 153 controls)
Case
–
control
>18 Not
specifi ed
308 2.03 1.21; 3.41 Inpatient
© 2012 by the American College of Gastroenterology The American Journal of GASTROENTEROLOGY
1005
REVIEW
C. difficile -Associated Diarrhea and PPI Therapy
CDAD by matching or adjustment. e publication dates of the
studies included in the meta-analysis ranged from 1990 to 2010.
e characteristics of the studies are shown in Tab le 1 .
Meta-analysis of all studies combined
PPI exposure was associated with a statistically signi cant
increase in the risk of CDAD as shown by the random-e ects
model: Overall RR is 1.69 (1.40 – 1.97) with P < 0.001. ere was
evidence of heterogeneity between the studies ( Table 2 ). e I 2
was 91.93 suggesting evidence of heterogeneity. e random-
e ects meta-analysis was therefore chosen to minimize the e ects
of heterogeneity. e P values for the Begg ’ s and Egger ’ s tests were
low, suggesting a low probability of publications bias ( P = 0.08 and
0.16, respectively).
the e ect of large and small sample – sized studies. Funnel plots
were also visually inspected.
To assess statistical heterogeneity between studies, we used the
Cochrane ’ s Q -statistic and the I 2 -statistic. An I 2 value of > 50 % or
a P value < 0.05 for the Q -statistic was taken to indicate signi cant
heterogeneity ( 21,22 ). All statistical tests were two-tailed, and a
probability level of < 0.05 was considered signi cant. Results are
presented in accordance with the guidelines proposed by MOOSE
( 23 ). All analyses were done using STATA 11.1 (STATA, College
Station, TX) statistical so ware and Comprehensive meta-analysis
so ware version 2 (Biostat, Englewood, NJ).
RESULTS
Search results
e search identi ed 49 potentially relevant studies on the sub-
ject area. Overall, 9 were excluded because they were review
articles; 12 focused primarily on pathophysiology; 2 others
were excluded because they were either letters or comments
on articles / case reports on the association between PPIs and
CDAD risk. ree studies were excluded because they were
position statements / recommendations by regulatory bodies,
and nally two were excluded as they were assessed to be
low in quality. Twenty-three studies were included in the nal
analysis.
Figure 1 shows a ow-sheet of the studies found, reasons for
exclusion, included studies and their classi cation by study design.
Nine eligible studies were cohort studies and 17 were case – control
studies. We ranked 23 studies as moderate quality. ese studies
involved a total of 288,620 participants. All the studies evaluated
exposure to any PPI for at least the prior 3 months to CDAD
episode. ey controlled for majority of known risk factors for
Table 1 . Continued
Author Year Country Study design
Study
type
Age
range
Duration of
PPI use Sample RR / OR 95 % CI
Inpatient /
community
Pepin et al.
( 27 )
2005 Canada Review of data obtained
from 7,421 episodes of care
corresponding to 5,619
hospitalized individuals
Cohort >18 3 – 12
Months
5,619 1 0.79; 1.28 Inpatient
Cunning-
ham et al.
( 41 )
2003 UK Retrospective study on
patients admitted to
Plymouth Hospital NHS
Trust, UK (58 cases vs.
31 controls)
Case –
control
All age
groups
>2 Months 89 2.5 1.5; 4.2 Inpatient
Shah et al.
( 42 )
2000 UK Retrospective study on
patients admitted to internal
and elderly care wards of
Moriston Hospital in South
Whales (126 cases vs. 126
controls)
Case –
control
>65 Not
specifi ed
251 0.86 0.47; 1.6 Inpatient
Kazakova
et al. ( 43 )
2006 USA Retrospective study of hospi-
talized patients and patients
in long-term care facilities
Case –
control
All age
groups
Not
specifi ed
195 5.02 1.3; 19.36 Inpatient
CDAD, Clostridium diffi cile -associated diarrhea; CI, confi dence interval; OR, odds ratio; PPI, proton pump inhibitor; RR, risk ratio.
49 Potentially relevant studies were identified
23 Studies included in the analysis
10 Articles on pathophysiology
were excluded
9 Were excluded as they were
reviews
2 Studies excluded: letters,
comments, or News reports
3 Studies-position statements
/recommendations excluded
2 Were excluded by inclusion
criteria
6 Cohort studies 17 Case–control studies
Figure 1 . Search and outcome of potential studies.
The American Journal of GASTROENTEROLOGY VOLUME 107 | JULY 2012 www.amjgastro.com
1006
REVIEW
Janarthanan et al.
calculated to be RR of 1.65 (1.38 – 1.98). e test of heterogene-
ity was nonsigni cant among the cohort studies and case – con-
trol studies in random-e ects analysis. is sensitivity analysis by
study design failed to explain the heterogeneity observed in the
overall analysis.
Publication bias . e funnel plot appears asymmetric ( Figure 3 ),
suggesting publication bias — although the Begg and Egger tests
were not statistically signi cant ( P = 0.08 and 0.16, respectively).
erefore, we performed a sensitivity analysis by using the trim-
and- ll method ( 23 ), which conservatively imputes hypothetical
negative unpublished studies to mirror the positive studies that
cause funnel plot asymmetry. e imputed studies produced a
symmetrical funnel plot. e pooled analysis incorporating the
hypothetical studies continued to show a statistically signi cant
association between CDAD and PPIs (with adjusted RR = 1.26, 95 %
CI 1.03, 2.241; P = 0.025). Additionally, by study design, the “ trim
and ll ” analysis method from Duval and Tweedie ( Figure 4a )
showed an RR of 1.37 (95 % CI 1.15, 1.63; P < 0.001) for the case –
control and 2.28 (95 % CI 1.51, 3.44; P < 0.001) for the cohort
studies. ere was also no evidence of publication bias by study
design using the Begg and Egger tests.
e funnel plot is shown in Figure 3 . e visual assessment
shows some asymmetry. To further complete the analysis of publi-
cation bias, Duval and Tweedie ’ s trim-and- ll ( Figure 4a ) method
was performed along with a cumulative analysis ( Figure 4b ). e
adjusted point estimate remained very close to the nal overall risk
estimate in the trim-and- ll method and the limits overlapped.
e cumulative analysis gave a very close overall RR of 1.39 (lower
limit 1.33 to upper limit 1.44).
Subgroup analyses
We also attempted to examine consistency across varying study
designs with di erent potential biases. We strati ed the data
into subgroups on the basis of the study design. e association
remained signi cant across study designs. Cohort studies ( 10,24 –
27 ) in a random-e ect model showed a RR of 1.66 (1.23 – 2.24)
with P < 0.001 ( Figure 2) . e case – control studies ( 28 – 43 ) were
Group by
Subgroup within study
Study name Subgroup within study Statistics for each study Risk ratio and 95% CI
Risk Lower Upper
ratio limit limit Z-value P-value
Case–control Ahktar et al. Case–control 1.600 1.380 1.855 6.228 0.000
Case–control Aseeri et al. Case–control 1.810 1.340 2.445 3.868 0.000
Case–control Bajaj et al. Case–control 2.720 2.200 3.363 9.243 0.000
Case–control Candle et al. Case–control 1.390 1.170 1.651 3.746 0.000
Case–control Cunningham et al. Case–control 1.950 1.370 2.776 3.708 0.000
Case–control Debast et al. Case–control 1.330 0.860 2.057 1.282 0.200
Case–control Dial et al., 2006 Case–control 1.460 1.110 1.920 2.706 0.007
Case–control Dial et al. Case–control 3.110 2.160 4.478 6.101 0.000
Jayatilaka et al. Case–control 1.270 1.120 1.440 3.727 0.000
Case–control Kim et al. Case–control 1.750 1.310 2.338 3.788 0.000
Case–control Lowe et al. Case–control 0.920 0.840 1.008 –1.796 0.072
Case–control Pepin et al. Case–control 1.000 0.800 1.250 0.000 1.000
Case–control Shah et al. Case–control 0.910 0.580 1.428 –0.410 0.682
Case–control Kazakova et al. Case–control 2.590 1.412 4.751 3.075 0.002
Case–control Yearsley et al. Case–control 1.610 1.150 2.254 2.774 0.006
Case–control Beaulieu et al. Case–control 0.960 0.780 1.182 –0.385 0.700
Case–control
Case–control
Baxter et al. Case–control 1.140 1.020 1.274 2.309 0.021
0.000
1.478
0.000
0.000
0.000
0.000
0.005
0.010
0.000
0.000
Dalton et al.
Dial et al., 2004
Dubberke et al.
Howell et al.
Linsky et al.
Turco et al.
Cohort
Cohort
Cohort
Cohort
Cohort
Cohort
Cohort
Overall
Cohort
Cohort
Cohort
Cohort
Cohort
Cohort
0.01 0.1 1 10 100
Favors A Favors B
Meta-analysis
1.960
2.900
3.800
1.740
1.420
4.500
2.309
1.648
1.249
1.420
2.400
3.000
1.390
1.110
1.440
1.718
1.424
1.749
2.705
3.504
4.813
2.178
1.817
14.063
3.103
1.908
4.547
4.092
11.027
11.069
4.834
2.790
2.587
5.550
6.698
Case–control
Figure 2 . Random-effects meta-analysis forest plot of all studies on the association between proton pump inhibitor therapy and risk of Clostridium diffi cile -
associated diarrhea. Study-specifi c effect sizes are shown as black squares, with the size of the square being inversely proportional to the study-specifi c
effect size variance. Horizontal lines represent 95 % confi dence intervals (CIs) for the study-specifi c effect size. The pooled risk ratio is shown as diamond.
The middle of the diamond corresponds to the risk ratio, and the width represents the 95 % CI. The vertical dashed line provides a visual comparison of the
pooled risk ratio with the corresponding study-specifi c effect size. Subgroup analysis results according to study type is also shown.
Table 2 . Heterogeneity analysis
Q d.f. P I
2
272.636 22 0.000 91.931
© 2012 by the American College of Gastroenterology The American Journal of GASTROENTEROLOGY
1007
REVIEW
C. difficile -Associated Diarrhea and PPI Therapy
Cumulative analysis ( Figure 4b ) was done entering the larger
studies at the top and adding the smaller studies at the bottom.
e larger studies are assumed to be unbiased, but the smaller
studies may tend to overestimate the e ect size. e e ect size was
not increased when smaller studies were added, showing that it
was not related to study sample size. Although the e ect size was
somewhat reduced, the association was maintained (RR = 1.39;
lower limit 1.33 to upper limit 1.44).
DISCUSSION
is study is the rst meta-analysis performed to assess the
speci c association between CDAD and PPIs. Similar analysis
in a systematic review has been undertaken previously by
Leonard et al. ( 13 ). ey found that there is an increased risk
–2.0 –1.5 –1.0 –0.5 0.0 0.5 1.0 1.5 2.0
0.0
0.1
0.2
0.3
0.4
0.5
0.6
Standard error
Log-risk ratio
Funnel plot of standard error by log-risk ratio
Figure 3 . Funnel plot with pseudo 95 % confi dence limits.
–2.0 –1.5 –1.0 –0.5 0.0 0.5 1.0 1.5 2.0
0.0
a
b
0.1
0.2
0.3
0.4
0.5
0.6
Standard error
Log-risk ratio
Funnel plot of standard error by log-risk ratio
Study name Cumulative statistics Cumulative risk ratio (95% CI)
Lower Upper
Point limit limit Z-value P-value
Bajaj et al. 2.720 2.200 3.363 9.243 0.000
Howell et al. 2.179 1.406 3.376 3.487 0.000
Dubberke et al. 2.617 1.698 4.035 4.357 0.000
Dalton et al. 2.448 1.724 3.477 5.002 0.000
Lowe et al. 1.998 1.092 3.656 2.245 0.025
Dial et al., 2005 2.145 1.232 3.735 2.697 0.007
Pepin et al. 1.920 1.197 3.081 2.705 0.007
Baxter et al. 1.789 1.247 2.566 3.159 0.002
Dial et al., 2006 1.748 1.258 2.430 3.323 0.001
Akhtar et al. 1.730 1.295 2.310 3.710 0.000
Dial et al., 2004 1.817 1.354 2.437 3.980 0.000
Linsky et al. 1.779 1.353 2.339 4.126 0.000
Beaulieu et al. 1.695 1.309 2.196 4.003 0.000
Yearsley et al. 1.689 1.320 2.161 4.169 0.000
Shah et al. 1.630 1.284 2.069 4.017 0.000
kazakova et al. 1.666 1.320 2.103 4.292 0.000
Aseeri et al. 1.674 1.338 2.093 4.513 0.000
Cadle et al. 1.655 1.343 2.038 4.737 0.000
Debast et al. 1.638 1.338 2.004 4.788 0.000
Jayatilaka et al. 1.614 1.341 1.942 5.072 0.000
Kim et al. 1.620 1.354 1.937 5.283 0.000
Cunningham et al. 1.633 1.371 1.944 5.506 0.000
Turco et al. 1.659 1.395 1.974 5.713 0.000
1.659 1.395 1.974 5.713 0.000
0.01 0.1 1 10 100
Favors A Favors B
Meta-analysis
Figure 4 . Detailed sensitivity analyses. ( a ) Funnel plot of all studies with pseudo 95 % confi dence interval (CI), including the hypothetical studies using the
“ trim-and-fi ll ” method. ( b ) Cumulative analysis of studies with decreasing sample sizes. Horizontal lines represent 95 % CIs for the cumulative effect size
of the added studies. The pooled risk ratio is shown as a diamond. The middle of the diamond corresponds to the risk ratio and the width represents
the 95 % CI.
The American Journal of GASTROENTEROLOGY VOLUME 107 | JULY 2012 www.amjgastro.com
1008
REVIEW
Janarthanan et al.
ies did not specify duration of PPI therapy. is did not give an
option to asses for the e ect of PPI therapy duration on CDAD.
As with all observational studies, it is possible that the association
of CDAD with PPIs could be a result of uncontrolled confound-
ing. A meta-analysis is not able to eliminate bias due to confound-
ing, if these were not properly controlled in the primary studies.
Although the studies included appear to be well designed, most
of the studies used data from computerized databases, which were
collected for other purposes and may be lacking in detail informa-
tion on potential confounders (e.g., medication use, comorbidi-
ties, chemotherapy, naso-gastric tubes, and so on) ( 4 – 6 ). Patients
with comorbidities (especially the elderly) are more likely to be
prescribed acid-suppressive therapy, and these patients are more
likely to have CDAD ( 45 ). Some of these factors, and their interac-
tion with established risk factors were not accounted for in some
of these studies and perhaps, could partially explain the moderate
increased risk in CDAD observed. ird, our results are likely to
be a ected by some degree of misclassi cation of exposure. e
de nition of PPI exposure was somewhat comprehensive in this
meta-analysis. Only two of the studies reported information on
medication compliance among study participants. Although it is
di cult to tell how many people were actually taking the pills as
prescribed, evidence from other studies has shown that withdrawal
rates from PPIs are less than 1 to 2 % ( 46 ). If misclassi cation of
exposure occurred in the primary studies, it is likely to have only
led to an underestimation of the true size of the e ect of PPIs on
CDAD.
ere are no randomized placebo-controlled clinical trials to
solidify the causality of this e ect. Ethically, it does not seem feasi-
ble to either withhold PPI therapy for those in whom it is indicated,
and likewise unethical to administer PPI to those in whom it is not
needed. As randomized controlled clinical trials being undertaken
seem unlikely, high-quality, prospective trails are probably the best
we could hope for. erefore, determining causality will continue
to be a challenge.
e heterogeneity between studies was not explained by the
subgroup analysis. e presence of heterogeneity renders the pool-
ing of OR / RRs data somewhat controversial. Considering these,
the summary static here should be interpreted with caution. e
association is also shown in xed-e ect meta-analysis. (1.39; 95 %
CI 1.33 – 1.44). erefore, the overall impression of the pooled data
showing a signi cant association between PPIs and CDAD could
be appreciated.
Studies that delved into the pathophysiology of the CDAD pos-
tulated that the acid suppression with PPI in uences the C. di -
cile infection ( 40 ). is was attributed to the survival of vegetative
spores of C. di cile in gastric contents obtained from patients
receiving PPIs, where the stomach alkaline status was increased
to a pH>5.0 ( 47 ). is is the likely explanation, as there is evi-
dence, that acid suppression by H2 receptor blockers also causes
an increase in CDAD ( 48 ). It has been suggested that bile salts
may have a role in stimulating the conversion of C. di cile spores
to the vegetative phase (47,49), and bacterial overgrowth mark-
edly increases the luminal concentration of unconjugated bile
acids. Bacterial overgrowth of the upper gut has been described
of C. di cile infection in people taking antisecretary therapy.
e paper concentrated on enteric infections as a whole. We
are trying to establish the association speci cally with PPI and
CDAD.
Our meta-analysis showed a statistically signi cant increase
in the incidence of CDAD among patients on PPI therapy. Most
of the results of the individual studies were consistent with the
overall results. e overall risk estimate was 1.654 (1.415 – 1.933)
with a 95 % CI ( P < 0.001). e study designs, case – control, and
cohort showed a signi cant increase in the overall results as well.
e increase in risk was stronger among the cohort studies (less
prone to bias). ere was, however, evidence of heterogeneity,
which could not be explained by study design. ere was no
evidence of publication bias. Additionally, a sensitivity analysis
using Duval and Tweedie “ trim-and- ll ” method continued to
show a signi cant increase in the CDAD risk with PPI use.
We attempted to classify the studies as prospective and retro-
spective to add weight to our conclusion. We were able to identify
only two prospective case – control studies (one by Howell et al.
( 28 ) and another by Yearsley et al. (31 ). We were unable to ana-
lyze this impact against all the rest of the retrospective studies,
as we had only 2 studies against 21. Our search revealed fewer
studies that did not show an increased association with CDAD
and PPI use ( 42,44 ). ere were three studies that showed an
increased association in an outpatient setting, but the study by
Lowe et al. ( 44 ) suggests that there is no increase in association
between outpatient use of PPI and CDAD ( 32,39,40,44 ). How-
ever, owing to the fewer study numbers ( 3 ), a subgroup analysis
could not be performed as the interpretation would not be valid.
is is an area that needs further research. e dose and duration
of therapy was not de ned in many of the studies. erefore, we
could not analyze the association with dose increase and the pres-
ence of CDAD.
Our study has several limitations. First, when a meta-analysis
is based on published data, consideration of selection bias is criti-
cal. Identi cation and selection of studies for our analysis did not
exclude any articles based on methodological characteristics or any
subjective quality criteria. e Begg and Egger tests suggest that
it is reasonable to assume that there was no relationship between
study size and being published among all included studies. is
was reinforced by the cumulative analysis done with increasing
study sample size. is again did not change the e ect size signi -
cantly a er the rst few studies ( Figure 4b ).
e results we obtained from the funnel plot analysis suggested
some evidence of publication bias. Because of this, we undertook a
sensitivity analysis using the trim-and- ll method ( 20 ). e trim-
and- ll sensitivity analysis did not change direction of the results,
although the magnitude of the association was somewhat weak-
ened, indicating that the association is not an artifact of unpub-
lished negative studies. We are con dent that publication of large
studies with signi cant ndings on the association between PPI
and CDAD is unlikely to fully explain our results. Nevertheless,
that possibility is not fully excluded by these methods.
Our exclusion criteria did not di erentiate between initial,
acute infection and recurrent infection ( 30 ). Most included stud-
© 2012 by the American College of Gastroenterology The American Journal of GASTROENTEROLOGY
1009
REVIEW
C. difficile -Associated Diarrhea and PPI Therapy
in PPI users ( 12,50 – 52 ). ere are suggestions of mucosal edema,
increased myeloperoxidase activity, epithelial damage, and neu-
trophil in ltrate in colons with PPI usage according to the studies
in mice ( 12,53 ). is also possibly has a role in the pathogenesis of
CDAD in the presence of PPI.
ere have been concerns arising from the over the counter use
of PPIs and recommendations have been made toward proper
administration of PPIs and on-demand and intermittent therapy
as opposed to continuous therapy ( 54 ). is may provide the solu-
tion in the reduction of CDAD incidence and prevalence.
CONCLUSION
PPI use increases the risk for CDAD signi cantly. Future studies
should focus on evaluating the dosage and duration of PPI use
and the risk of CDAD and its recurrence. Further high-quality,
prospective trials focusing on duration of therapy and dose of
therapy are further warranted to show the strength of this asso-
ciation.
CONFLICT OF INTEREST
Guarantor of the manuscript: Sailajah Janarthanan, MD.
Speci c author contributions: Sailajah Janarthanan: conception
of the idea, collecting data, statistical analysis, and writing the
manuscript; Ivo Ditah: data collection, statistical analysis, writing
the manuscript; Douglas G. Adler and Murray N. Ehrinpreis: writing
and editing the manuscript.
Financial support: None.
Potential competing interests: N o n e .
REFERENCES
1 . B l o s s o m D B , M c D o n a l d L C . e challenges posed by reemerging
Clostridium di cile infection . Clin Infect Dis 2007 ; 45 : 222 – 7 .
2 . K e l l y C P . P l a y i n g h o s t t o t h e d i cult Clostridium . Clin Gastroenterol
Hepatol 2007 ; 5 : 912 – 4 .
3 . K y n e L , H a m e l M B , P o l a v a r a m R et al. Health care costs and mortality
associated with nosocomial diarrhea due to Clostridium di cile . Clin Infect
Dis 2002 ; 34 : 346 – 53 .
4 . P o u t a n e n S M , S i m o r A E . Clostridium di cile -associated diarrhea in adults .
CMAJ 2004 ; 171 : 51 – 8 .
5 . A n a n d A , G l a t t A E . Clostridium di cile infection associated with anti-
neoplastic chemotherapy: a review . Clin Infect Dis 1993 ; 17 : 109 – 13 .
6 . B i g n a r d i G E . R i s k f a c t o r s f o r Clostridium di cile infection . J Hosp Infect
1998 ; 40 : 1 – 15 .
7 . McFarland LV . Update on the changing epidemiology of Clostridium dif-
cile -associated disease . Nat Clin Pract Gastroenterol Hepatol 2008 ; 5 : 40 – 8 .
8 . S o n g X , B a r t l e t t J G , S p e c k K et al. Rising economic impact of clostridium
di cile -associated disease in adult hospitalized patient population . Infect
Control Hosp Epidemiol 2008 ; 29 : 823 – 8 .
9 . O ’ Brien JA , Lahue BJ , Caro JJ et al. e emerging infectious challenge of
clostridium di cile -associated disease in Massachusetts hospitals: clinical
and economic consequences . Infect Control Hosp Epidemiol 2007 ; 28 :
1219 – 27 .
10 . Linsky A , Gupta K , Lawler EV et al. Proton pump inhibitors and risk for
recurrent Clostridium di cile infection . Arch Intern Med 2010 ; 170 : 772 – 8 .
1 1 . G u r i a n L , W a r d T T , K a t o n R M . P o s s i b l e f o o d b o r n e t r a n s m i s s i o n i n a c a s e
of pseudomembranous colitis due to Clostridium di cile : in uence of gas-
trointestinal secretions on Clostridium di cile infection . Gastroenterology
1982 ; 83 : 465 – 9 .
1 2 . K a u r S , V a i s h n a v i C , P r a s a d K K et al. Comparative role of antibiotic and
proton pump inhibitor in experimental Clostridium di cile infection in
mice . Microbiol Immunol 2007 ; 51 : 1209 – 14 .
1 3 . L e o n a r d J , M a r s h a l l J K , M o a y y e d i P . S y s t e m a t i c r e v i e w o f t h e r i s k o f
enteric infection in patients taking acid suppression . Am J Gastroenterol
2007 ; 102 : 2047 – 56 .
14 . Greenland S . Quantitative methods in the review of epidemiologic litera-
ture . Epidemiol Rev 1987 ; 9 : 1 – 30 .
1 5 . Z h a n g J , Y u K F . W h a t ’ s t h e r e l a t i v e r i s k ? A m e t h o d o f c o r r e c t i n g t h e o d d s
ratio in cohort studies of common outcomes . JAMA 1998 ; 280 : 1690 – 1 .
16 . DerSimonian R , Laird N . Meta-analysis in clinical trials . Controlled Clin
Trials 1986 ; 7 : 177 – 88 .
17 . Petitti DB . Meta-analysis and endocrinology . Endocrinol Metabol Clin
North America 1997 ; 26 : 31 – 44 .
18 . Begg CB , Mazumdar M . Operating characteristics of a rank correlation test
for publication bias . Biometrics 1994 ; 50 : 1088 – 101 .
19 . Egger M , Davey Smith G , Schneider M et al. Bias in meta-analysis detected
by a simple, graphical test . Br Med J 1997 ; 315 : 629 – 34 .
2 0 . D u v a l S , T w e e d i e R . T r i m a n d ll: a simple funnel-plot-based method of
testing and adjusting for publication bias in meta-analysis . Biometrics
2000 ; 56 : 455 – 63 .
21 . Deeks JJ . Issues in the selection of a summary statistic for meta-analysis of
clinical trials with binary outcomes . Stat Med 2002 ; 21 : 1575 – 600 .
2 2 . H i g g i n s J P , o m p s o n S G , D e e k s J J et al. Measuring inconsistency in meta-
analyses . Br Med J 2003 ; 327 : 557 – 60 .
2 3 . S t r o u p D F , B e r l i n J A , M o r t o n S C et al. Meta-analysis of observational stud-
ies in epidemiology: a proposal for reporting. Meta-analysis Of Observa-
tional Studies in Epidemiology (MOOSE) group . JAMA 2000 ; 283 : 2008 – 12 .
24 . Dial S , Alrasadi K , Manoukian C et al. R i s k o f Clostridium di cile diarrhea
among hospital inpatients prescribed proton pump inhibitors: cohort and
case-control studies . CMAJ 2004 ; 171 : 33 – 8 .
2 5 . D a l t o n B R , L y e - M a c c a n n e l l T , H e n d e r s o n E A et al. Proton pump inhibi-
tors increase signi cantly the risk of Clostridium di cile infection in a
low-endemicity, non-outbreak hospital setting . Aliment Pharmacol er
2009 ; 29 : 626 – 34 .
2 6 . D u b b e r k e E R , R e s k e K A , Ya n Y et al. Clostridium di cile -associated disease
in a setting of endemicity: identi cation of novel risk factors . Clin Infect
Dis 2007 ; 45 : 1543 – 9 .
2 7 . P e p i n J , A l a r y M E , V a l i q u e t t e L et al. Increasing risk of relapse a er treat-
ment of Clostridium di cile colitis in Quebec, Canada . Clin Infect Dis
2005 ; 40 : 1591 – 7 .
2 8 . Y e a r s l e y K A , G i l b y L J , R a m a d a s A V et al. Proton pump inhibitor therapy is
a risk factor for Clostridium di cile -associated diarrhoea . Aliment Pharma-
col er 2006 ; 24 : 613 – 9 .
2 9 . T u r c o R , M a r t i n e l l i M , M i e l e E et al. Proton pump inhibitors as a risk fac-
tor for paediatric Clostridium di cile infection . Aliment Pharmacol er
2010 ; 31 : 754 – 9 .
3 0 . K i m J W , L e e K L , J e o n g J B et al. Proton pump inhibitors as a risk factor for
recurrence of Clostridium-di cile -associated diarrhea . World J Gastroen-
terol 2010 ; 16 : 3573 – 7 .
3 1 . H o w e l l M D , N o v a c k V , G r g u r i c h P et al. Iatrogenic gastric acid suppression
and the risk of nosocomial Clostridium di cile infection . Arch Intern Med
2010 ; 170 : 784 – 90 .
3 2 . B a j a j J S , A n a n t h a k r i s h n a n A N , H a f e e z u l l a h M et al. Clostridium di cile is
associated with poor outcomes in patients with cirrhosis: a national and
tertiary center perspective . Am J Gastroenterol 2010 ; 105 : 106 – 13 .
3 3 . D e b a s t S B , v a n L e e n g o e d L A , G o o r h u i s A et al. Clostridium di cile PCR
ribotype 078 toxinotype V found in diarrhoeal pigs identical to isolates
from a ected humans . Environ Microbiol 2009 ; 11 : 505 – 11 .
3 4 . A s e e r i M , S c h r o e d e r T , K r a m e r J et al. Gastric acid suppression by proton
pump inhibitors as a risk factor for Clostridium di cile -associated diarrhea
in hospitalized patients . Am J Gastroenterol 2008 ; 103 : 2308 – 13 .
3 5 . B a x t e r R , R a y G T , F i r e m a n B H . C a s e - c o n t r o l s t u d y o f a n t i b i o t i c u s e a n d
subsequent Clostridium di cile -associated diarrhea in hospitalized patients .
Infect Control Hosp Epidemiol 2008 ; 29 : 44 – 50 .
3 6 . C a d l e R M , M a n s o u r i M D , L o g a n N et al. Association of proton-pump
inhibitors with outcomes in Clostridium di cile colitis . Am J Health Sys
Pharm 2007 ; 64 : 2359 – 63 .
37 . Jayatilaka S , Shakov R , Eddi R et al. Clostridium di cile infection in an
urban medical center: ve-year analysis of infection rates among adult
admissions and association with the use of proton pump inhibitors .
Ann Clin Lab Sci 2007 ; 37 : 241 – 7 .
38 . Akhtar AJ , Shaheen M . Increasing incidence of Clostridium di cile -associ-
ated diarrhea in African-American and Hispanic patients: association
with the use of proton pump inhibitor therapy . J Natl Med Assoc 2007 ; 99 :
500 – 4 .
The American Journal of GASTROENTEROLOGY VOLUME 107 | JULY 2012 www.amjgastro.com
1010
REVIEW
Janarthanan et al.
4 7 . J u m p R L , P u l t z M J , D o n s k e y C J . V e g e t a t i v e Clostridium di cile survives
in room air on moist surfaces and in gastric contents with reduced acidity:
a potential mechanism to explain the association between proton pump
inhibitors and C. di cile -associated diarrhea? Antimicrob Agents
Chemother 2007 ; 51 : 2883 – 7 .
4 8 . K h a t a m i S S , M u k u n d a B , R a v a k h a h K . C o i n f e c t i o n w i t h G i a r d i a l a m b l i a a n d
Clostridium di cile a er use of ranitidine . Am J Med Sci 2004 ; 327 : 91 – 3 .
4 9 . W i l s o n K H , S h e a g r e n J N , F r e t e r R . P o p u l a t i o n d y n a m i c s o f i n g e s t e d
Clostridium di cile in the gastrointestinal tract of the Syrian hamster .
J Infect Dis 1985 ; 151 : 355 – 61 .
5 0 . L e w i s S J , F r a n c o S , Yo u n g G et al. Altered bowel function and duodenal
bacterial overgrowth in patients treated with omeprazole . Aliment
Pharmacol er 1996 ; 10 : 557 – 61 .
51 . Williams C , McColl KE . Review article: proton pump inhibitors and
bacterial overgrowth . Aliment Pharmacol er 2006 ; 23 : 3 – 10 .
5 2 . eisen J , Nehra D , Citron D et al. Suppression of gastric acid secretion
in patients with gastroesophageal re ux disease results in gastric bacterial
overgrowth and deconjugation of bile acids . J Gastrointest Surg 2000 ; 4 : 50 – 4 .
53 . Paredes-Sabja D , Setlow P , Sarker MR . Germination of spores of
Bacillales and Clostridiales species: mechanisms and proteins involved .
Trends Microbiol 2011 ; 19 : 85 – 94 .
54 . Inadomi JM , Fendrick AM . PPI use in the OTC era: who to treat, with
what, and for how long? Clin Gastroenterol Hepatol 2005 ; 3 : 208 – 15 .
39 . Dial S , Delaney JA , Schneider V et al. Proton pump inhibitor use and
risk of community-acquired Clostridium di cile -associated disease
de ned by prescription for oral vancomycin therapy . CMAJ 2006 ; 175 :
745 – 8 .
4 0 . D i a l S , D e l a n e y J A , B a r k u n A N et al. Use of gastric acid-suppressive agents
and the risk of community-acquired Clostridium di cile -associated disease .
JAMA 2005 ; 294 : 2989 – 95 .
41 . Cunningham R , Dale B , Undy B et al. Proton pump inhibitors as a risk
factor for Clostridium di cile diarrhoea . J Hosp Infect 2003 ; 54 : 243 – 5 .
4 2 . S h a h S , L e w i s A , L e o p o l d D et al. Gastric acid suppression does not
promote clostridial diarrhoea in the elderly . QJM 2000 ; 93 : 175 – 81 .
4 3 . K a z a k o v a S V , W a r e K , B a u g h m a n B et al. A hospital outbreak of diarrhea
due to an emerging epidemic strain of Clostridium di cile . A r c h I n t e r n
Med 2006 ; 166 : 2518 – 24 .
4 4 . L o w e D O , M a m d a n i M M , K o p p A et al. Proton pump inhibitors and
hospitalization for Clostridium di cile -associated disease: a population-
based study . Clin Infect Dis 2006 ; 43 : 1272 – 6 .
45 . Beaulieu M , Williamson D , Pichette G et al. R i s k o f Clostridium di cile -
associated disease among patients receiving proton-pump inhibitors in
a Quebec medical intensive care unit . Infect Control Hosp Epidemiol
2007 ; 28 : 1305 – 7 .
4 6 . o m s o n A B , S a u v e M D , K a s s a m N et al. Safety of the long-term use of
proton pump inhibitors . World J Gastroenterol 2010 ; 16 : 2323 – 30 .
