systematic reviews and meta-analyses - nghd - … · systematic reviews and meta-analyses fasiha...
TRANSCRIPT
V
aigdpas
CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2013;11:483–490
SYSTEMATIC REVIEWS AND META-ANALYSESFasiha Kanwal, Section Editor
Proton Pump Inhibitor Use and the Risk of Small Intestinal BacterialOvergrowth: A Meta-analysis
WAI–KIT LO*,‡ and WALTER W. CHAN*
*Division of Gastroenterology, Hepatology and Endoscopy, Brigham and Women’s Hospital, Harvard Medical School; and ‡Department of Gastroenterology, BostonA Healthcare System, Boston, Massachusetts
BACKGROUND & AIMS: Use of proton pump inhibitors (PPIs) could predispose individuals to small intestinalbacterial overgrowth (SIBO) by altering the intraluminal environment and bacterial flora.There is controversy regarding the risk of SIBO among PPI users because of conflictingresults from prior studies. A systematic review and meta-analysis were performed toevaluate the association between PPI use and SIBO, using objective clinical outcomemeasures.
METHODS: Clinical studies comparing SIBO risk among adult users of PPIs vs nonusers were identifiedin MEDLINE/PubMed, EMBASE, the Cochrane Central Register of Controlled Trials, andthe National Institutes of Health Clinical Trials databases through July 2012. Two reviewersindependently extracted data on study characteristics and outcomes. The primary metame-ter was the odds ratio (OR) of SIBO among PPI users vs nonusers. Subgroup analyses wereperformed to examine the influence of study characteristics, such as SIBO diagnosticmodality, on study outcome.
RESULTS: Eleven studies (n � 3134) met inclusion criteria. The pooled OR of SIBO in PPI users vsnonusers was 2.282 (95% confidence interval [CI], 1.238 – 4.205). No significant single largestudy or temporal effect was seen. Subgroup analysis revealed an association between SIBOand PPI use in studies that used duodenal or jejunal aspirate cultures to diagnose SIBO(OR, 7.587; 95% CI, 1.805–31.894), but no relationship was found between SIBO and PPI usein studies that used the glucose hydrogen breath test (OR, 1.93; 95% CI, 0.69 –5.42). Funnelplot analysis identified 4 outlying studies, indicating the possible presence of publicationbias.
CONCLUSIONS: PPI use statistically was associated with SIBO risk, but only when the diagnosis was madeby a highly accurate test (duodenal or jejunal aspirate culture). Differences in study resultscould arise from the use of different tests to diagnose SIBO.
Keywords: Hypochlorhydria; Drug; Small Bowel Flora; Reflux; Side Effect.
Proton pump inhibitors (PPIs) are frequently prescribed byboth primary care physicians and gastroenterologists for a
wide range of indications. They represent one of the highestclasses of medications in sales and prescription in the UnitedStates, both in the inpatient and outpatient settings.1–3 Food
nd Drug Administration–approved indications for PPI usenclude treatment of gastroduodenal ulcer, erosive esophagitis,astroesophageal reflux disease, gastric hypersecretory syn-romes, and part of the treatment regimen for Helicobacterylori. Its use has also been recommended in the hospital settings prophylaxis against gastrointestinal bleeding resulting fromtress ulcer formation in critically ill patients.4
Since the release of the first PPI in 1989, an increasing
number of studies have evaluated the long-term side effects ofPPI use, including small intestinal bacterial overgrowth (SIBO).SIBO is a condition in which increased bacterial load in thesmall bowel results in excessive fermentation and inflamma-tion, leading to a variety of clinical complaints including bloat-ing and diarrhea. The clinical symptoms are often accompaniedby weight loss and nutritional malabsorption, which may resultfrom cellular injury caused by enterotoxin production and
Abbreviations used in this paper: CI, confidence interval; GHBT,glucose hydrogen breath test; OR, odds ratio; PPI, proton pump inhib-itor; SIBO, small intestinal bacterial overgrowth.
© 2013 by the AGA Institute1542-3565/$36.00
http://dx.doi.org/10.1016/j.cgh.2012.12.011
hsolt
adv
ossTsawp.pcbaTuwtg
484 LO AND CHAN CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 11, No. 5
bacterial adhesion.5 Chronic acid suppression and the resultantypochlorhydria associated with PPI use have been hypothe-ized to alter the intraluminal environment to promote growthf the bacterial flora in the small intestine. Several reports have
inked hypochlorhydria to increased gastric and duodenal bac-erial colonization,6 – 8 which have been shown to predispose
patients to SIBO development.9 Gastric hypochlorhydria haslso been implicated as a contributing factor in the risk ofeveloping SIBO in elderly10,11 and human immunodeficiencyirus12 patients.
In 1994, 2 initial reports were published examining thepossible association between PPI use and risk of SIBO.13,14 The2 studies used varying methodologies and design, includingdifferent diagnostic testing and criteria for SIBO. Conflictingoutcomes regarding the risk of SIBO associated with PPI useresulted from these 2 initial reports. Since then, a number ofcohort and case-controlled studies have been implemented inan attempt to clarify the magnitude and directionality of therisk,8,15–22 with varying results. The relationship between PPI useand SIBO, therefore, remains controversial and confusing forboth patients and clinicians. With the widespread use andavailability of over-the-counter PPI, a better assessment of itsside-effect profile and link to other conditions, such as SIBO, isessential.
A previously published review discussed the possible role ofPPIs as confounders in the association between SIBO and irri-table bowel syndrome.23 The objective of this meta-analysis wasto review available published peer-reviewed evidence to evaluateobjectively the relationship between PPI use and diagnosis ofSIBO.
MethodsData Sources and SearchesAn electronic search was performed using MEDLINE/
PubMed (1950 to July 2012), EMBASE (1980 to July 2012), theCochrane Central Register of Controlled Trials, and the USNational Institutes of Health Clinical Trials databases for case-control studies, cohort studies, or clinical trials of PPI use andSIBO published through July 2012. The following search termswere used as both keywords and medical subject heading terms,where applicable: proton pump inhibitor, PPI, omeprazole, lan-soprazole, pantoprazole, esomeprazole, rabeprazole, dexlanso-prazole, malabsorption syndromes, bacterial overgrowth. Nolanguage restrictions were applied.
Study SelectionThe authors independently reviewed the results of the
search strategy. The titles, abstracts, and keywords of identifiedarticles were screened for relevance and the reference lists ofthese articles were examined individually for additional perti-nent studies. To be included in this meta-analysis, studies wererequired: (1) to be randomized controlled trials, case-controlstudies, or cohort studies of SIBO risk in general adult (age �18y) PPI users vs nonusers; (2) to diagnose SIBO by one of thecurrently available clinical tests, namely small-bowel aspirateculture or breath testing; and (3) to include a non-PPI controlgroup. To minimize heterogeneity, studies that were publishedin abstract form were not included. Studies that focused exclu-sively on a specific subpopulation of patients at increased risk
for SIBO, such as patients with scleroderma or postabdominalsurgery, were excluded because some of these conditions may beeffect modifiers, thus skewing the association between PPI useand SIBO.
Data Extraction and Quality AssessmentTwo independent reviewers extracted data and assessed
the quality of the selected studies with no discrepancies. Thefollowing clinical data were identified a priori and extractedfrom each study, where possible: total number of subjects withand without SIBO off PPI; total number of subjects with andwithout SIBO on PPI; adjusted odds ratio (OR) of SIBO in PPIusers vs nonusers; geographic region of study; testing modalityto diagnose SIBO; cut-off values for hydrogen breath testing;number of subjects; mean age of subjects; and type, dose, andduration of PPI exposure. Formal methodologic quality wasassessed using the Newcastle–Ottawa Quality Assessment Scalefor nonrandomized studies.24 All included studies were rated 6to 8 of a maximum score of 9, and were of sufficient quality forinclusion in the meta-analysis.
Data Synthesis and Statistical AnalysisThe primary metameter for this study was the OR for
SIBO among PPI users vs nonusers. To quantify the variationacross studies caused by heterogeneity, the Cochran Q and I2
statistics were calculated, with a P value less than .05 for the Qstatistic or I2 greater than 20% considered significant for heter-
geneity. Potential sources for clinical heterogeneity amongtudies were specified a priori, including testing modality andtudy design, for which subgroup analyses were performed.hese subgroups were analyzed further by stratification to as-
ess for possible interactions between these factors. Because ofnticipated variability in patient population and study design,e used the DerSimonian–Laird random effects model forooled analysis with significance accepted at a P value less than
05. A one-study–removed analysis was performed on theooled data to examine the effect of single large studies. Aumulative analysis was also conducted to explore time trendsy publication year. The possibility of publication bias wasssessed with a funnel plot and adjusted with Duval andweedie25 trim-and-fill method. All analyses were conductedsing Comprehensive Meta-Analysis, version 2 (Biostat, Engle-ood, NJ). The meta-analysis was performed in accordance with
he Meta-analysis Of Observational Studies in Epidemiologyuidelines.26
ResultsStudy SelectionThe search strategy yielded 305 articles, of which 12
full-text studies were retrieved. The excluded 293 articles in-cluded duplicate citations, non–peer-reviewed abstracts, animalstudies, case reports or series, editorials, letters, reviews, andpractice guidelines. Of the 12 studies identified, 1 was excludedfor focusing on a specific subpopulation of patients with in-creased baseline risk of SIBO development27; however, the studywas small and had only 10 subjects, so its inclusion would havehad only a minimal impact on the overall pooled results. Theresulting 11 publications represented 11 discrete study cohortsthat were included in the meta-analysis. Figure 1 shows a flow
chart documenting the study selection process.r
bpafws
ble
1.
Evid
ence
Tabl
e
Stu
dyYe
arof
publ
icat
ion
Cou
ntry
Stu
dyde
sign
SIB
Odi
agno
stic
test
Bre
ath
test
cut-o
ffva
lue
Con
trol
type
Num
ber
ofsu
bjec
tsM
ean
age,
y
det
al12
19
94
Sw
itzer
land
Coh
ort
Aspi
rate
NA
Sep
arat
eco
ntro
l40
53.4
liset
al1
31
99
4Th
eN
ethe
rland
sC
ase-
cont
rol
14C
-gly
coch
olat
ebr
eath
test
NA
Sel
f-con
trol
40
58.9
iset
al1
51
99
6S
outh
Afric
aC
ase-
cont
rol
Aspi
rate
NA
Sel
f-con
trol
20
42.7
rens
etal
16
19
96
Sw
itzer
land
Coh
ort
Aspi
rate
NA
Sep
arat
eco
ntro
l47
42
tchi
nson
and
Loga
n17
19
97
Uni
ted
Kin
gdom
Coh
ort
GH
BT
H2
�20
Sep
arat
eco
ntro
l50
78.5
eira
etal
81
99
8U
nite
dK
ingd
omC
ase-
cont
rol
Aspi
rate
NA
Sel
f-con
trol
24
76
and
Pim
ente
l18
20
10
Uni
ted
Sta
tes
Coh
ort
LHB
TH
2�
20
Sep
arat
eco
ntro
l555
44.6
bard
oet
al1
92
01
0Ita
lyC
ohor
tG
HB
TH
2�
10
Sep
arat
eco
ntro
l450
37
ung
etal
20
20
11
Uni
ted
Sta
tes
Coh
ort
Aspi
rate
NA
Sep
arat
eco
ntro
l675
53
pare
etal
22
20
11
Italy
Cas
e-co
ntro
lG
HB
TH
2�
12
Sel
f-con
trol
42
36
uapl
iet
al2
32
01
2U
nite
dS
tate
sC
ohor
tG
HB
TH
2�
20
orC
H4
�15
Sep
arat
eco
ntro
l1191
60.9
T,la
ctul
ose
hydr
ogen
brea
thte
st.
May 2013 PPI USE AND SIBO 485
In all, 11 study cohorts were identified that reported SIBOrisk in PPI users vs nonusers (n � 3134) (Table 1). Nine studieswere prospective cohort studies,8,13–19,21 and 2 were retrospectiveeviews.20,22 Seven studies13,16 –20,22 used a separate cohort of PPI
nonusers for comparison, whereas the remaining 4 stud-ies8,14,15,21 compared outcomes in the same cohort before andafter PPI exposure. Five studies used duodenal/jejunal aspirateculture obtained via small-bowel intubation for the diagnosis ofSIBO,8,13,15,17,20 whereas the remaining 6 studies used a varietyof breath testing, including the 14C-glycocholate breath test,14
glucose hydrogen breath test (GHBT),17,19,21,22 or lactulose hy-drogen breath test.18 One study7 used the lactulose hydrogen
reath test as an additional confirmatory test in a minority ofatients (n � 6) after diagnosis of SIBO by duodenal/jejunalspirate, and a second study15 used 14C-D-xylose breath testingor diagnosis of SIBO in addition to jejunal aspirate; these dataere omitted from analysis to prevent double counting of
ubjects. Another study22 compared test results of 4 differentcut-off thresholds for SIBO diagnosis in GHBT. Data from thecut-off threshold of either H2 greater than 20 or CH4 greaterthan 15 was used for this meta-analysis because this reflects themost commonly accepted practice in clinical SIBO testing cur-rently. The mean age of subjects across all studies was 52.4, withwomen comprising 64.3% of the total study population. Allstudies had a Newcastle–Ottawa score between 6 and 8, andwere of acceptable methodologic quality for inclusion in themeta-analysis.
Meta-analysisAcross all studies, a statistically significant level of het-
erogeneity was observed (Q test, 61.2; P � .001; I2, 83.7), whichwas likely attributable to the various testing modalities andcut-off ranges used in the diagnosis of SIBO. In the analysis ofthe primary metameter, 6 of the 11 included studies showed anincreased risk of SIBO with PPI use, as calculated from theextracted numeric data. The overall meta-analysis revealed a
Figure 1. Flow diagram depicting the search and selection process.
statistically significant increase in risk of SIBO after applying
Ta FrieNe
Lew
Tho
Hu
Per
Law
Lom
Cho
Com
Rat
LHB
ttSs(
euwa1g
stwdtPgSsts3sar
486 LO AND CHAN CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 11, No. 5
the random-effects model to account for the observed hetero-geneity (OR, 2.282; 95% confidence interval [CI], 1.238 – 4.205)(Figure 2).
Additional analyses were run to exclude the role of a singlelarge study effect, assess the consistency of cumulative out-comes over time, and evaluate potential publication bias (Figure3). The one-study-removed analysis showed similar pooled out-comes without significant deviation from the overall result afterremoval of individual studies, suggesting no significant effectfrom a single large study (Figure 3A). Cumulative analysisshowed temporal progression, demonstrating a statistically sig-nificant risk of SIBO as studies were added over time (Figure3B). Potential publication bias was evaluated with a funnel plot,which showed 4 outlying studies with high standard error atthe right margin of the plot (Figure 3C). The Duval andTweedie25 trim-and-fill method was applied, resulting in 4 stud-ies being imputed with an adjusted OR of 1.440 (95% CI,0.778 –2.667), showing no relationship between risk of SIBOand PPI use after accounting for potential publication bias.
A subgroup analysis was performed to examine the effects ofSIBO testing modality, one of the presumed sources of heter-ogeneity, on the pooled results (Figure 4A). In the subgroupincluding only studies using duodenal/jejunal aspirate for SIBOdiagnosis, an increase in risk for SIBO among PPI users wasnoted (OR, 7.587; 95% CI, 1.805–31.894). Among studies usingGHBT for SIBO diagnosis, no relationship between PPI use andSIBO was seen (OR, 1.93; 95% CI, 0.69 –5.42). Further divisionof the breath-test subgroup was not possible given the presenceof only one study using the lactulose hydrogen breath test, andone study using the 14C-glycocholate breath test. Finally, whenhe Duval and Tweedie25 trim-and-fill method was applied tohe subgroup of studies using duodenal/jejunal aspirate forIBO diagnosis to account for potential publication bias, 2tudies were imputed with a resultant adjusted OR of 3.77695% CI, 1.144 –12.464).
A second subgroup analysis was performed to assess theffects of study and control group design (Figure 4B). Studiessing a self-control or pretest-posttest design, in which subjectsere tested for SIBO before and after PPI exposure, showed anssociation between SIBO and PPI use (OR, 9.607; 95% CI,.646 –56.089). On the other hand, studies using a separate
roup of PPI nonusers as controls showed no statistically sig- anificant risk for SIBO among PPI users (OR, 1.658; 95% CI,0.869 –3.163). Further analysis of the subgroups by stratifica-tion showed no change in the effect of SIBO testing modalityon outcome when controlling for study design. However, whencontrolling for SIBO testing modality, no difference in SIBOrisk among PPI users was seen by study design. These analysessuggest that the effect of test modality on outcome was inde-pendent of study design, but not vice versa.
Additional investigation of the effects of dose, duration, andtype of PPI exposure on SIBO diagnosis could not be assessedas a result of insufficient data. Clinical indications for SIBOtesting, such as irritable bowel syndrome, were not specified inevery study, and therefore could not be included in furtheranalysis.
DiscussionFor the primary metameter of interest, risk of SIBO
with PPI use, 11 studies were examined and evidence of in-creased risk was detected using random-effects pooled meta-analysis. Although there was significant heterogeneity in resultsamong studies, 6 of 11 included studies reported a statisticallysignificant increase in SIBO risk with PPI use. When potentialpublication bias was accounted for using the Duval andTweedie25 trim-and-fill method, the adjusted OR was no longertatistically significant. Subgroup analysis showed an associa-ion between PPI use and SIBO when the diagnosis was madeith duodenal/jejunal aspirate culture, the current gold stan-ard. This association persisted even after adjusting for poten-ial publication bias. In contrast, there was no relationship withPI use when SIBO was diagnosed by GHBT. Although sub-roup analysis by study type showed a limited association withIBO and PPI use for both self- and separately controlledtudies, there appeared to be some interaction between studyype and testing modality. A review of the studies (Table 1) alsohowed that of the 5 studies using aspirate for SIBO diagnosis,
featured separate controls, whereas of the 7 studies witheparate controls, only 3 were aspirate studies. In this case,ssociation as a result of study type may account for part of theelationship between SIBO and PPI use seen with studies using
Figure 2. Meta-analysis ofSIBO risk with PPI use.
spirate culture as the testing modality.
iPm
cr
mafStFS
May 2013 PPI USE AND SIBO 487
This meta-analysis explored the relationship between PPI useand the risk of SIBO, an area that continues to remain contro-versial. Given the large patient population analyzed in thisstudy, we believe that it contributes significantly to scientificknowledge and clinical care of patients taking PPIs. When ahighly accurate diagnostic test (duodenal/jejunal aspirate cul-ture) was used, PPI use appeared to increase the risk of SIBO.
Figure 3. (A) One-study-re-oved analysis of SIBO risk to
ssess the single large study ef-ect. (B) Cumulative analysis ofIBO risk to assess the consis-
ency of results over time. (C)unnel plot of standard error byIBO risk.
Along with studies linking other adverse events to chronic acid S
suppression, such as pneumonia,28 bone loss,29 and entericnfections,30 these results highlight the potential toxicity ofPIs, the need for judicious application, and the importance ofedication reduction when reaching treatment goals.The disparate findings between duodenal/jejunal aspirate
ulture and GHBT studies underscore the need for a clinicallyelevant and accurate outcome measure for the diagnosis of
IBO. The current gold standard is duodenal/jejunal aspirateaactgt
f
mttSop
A
488 LO AND CHAN CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 11, No. 5
culture, which involves an invasive procedure such as endos-copy or fluoroscopy to intubate the small bowel to obtainduodenal/jejunal fluid. SIBO is diagnosed when bacterialcounts exceed 105 colony forming units per milliliter of fluid
spirate. Although the sensitivity and specificity of the testpproach 100%,31 practical considerations such as patient dis-omfort and expense resulted in the introduction of breathesting for SIBO diagnosis. Breath testing relies on the in-estion of a test dose of carbohydrate, which can then lead tohe release of a radioactive isotope (14C-D-xylose or 14C-
glycocholate breath tests) or hydrogen species (glucose orlactulose hydrogen breath tests) when digested by a higherload of bacterial flora in SIBO patients. Although diagnosticbreath tests for SIBO are simpler to administer, less invasive,and less costly than duodenal/jejunal aspirate culture, theyare comparatively less sensitive and specific31,32 (Table 2).Potential contributors to the low sensitivities of breath test-ing include increased conversion of hydrogen to methane bycertain gut microbes,33 recent ingestion of high-carbohydrateoods that can prolong hydrogen secretion, hyperventilation
Table 2. Sensitivity and Specificity of SIBO TestingModalities
Testing modality Sensitivity Specificity
spirate culture 100% 100%14C–D-xylose breath test 14%–95% 100%14C-glycocholate breath test 33% 76%GHBT 6%–93% 78%–100%
Lactulose hydrogen breath test 6%–68% 44%–70%from recent exercise or pulmonary disease, increased loads oforal bacteria, short-gut syndrome, and low anaerobic bacte-rial load in the colon.34 Compounding the situation are the
ultiple thresholds of isotope, hydrogen, and methane de-ection that have been used to diagnose SIBO by breathesting. A systematic review of diagnostic testing options forIBO published in 2008 showed no clear standard or validityf any of the testing modalities across 71 peer-reviewedublications.35 The relatively poor performance and lack of
standardization of breath tests therefore have led many toquestion their validity and utility in the clinical diagnosis ofSIBO.
In this meta-analysis, the absence of a clear associationbetween PPI use and SIBO in the pooled analysis of studiesusing GHBT as a diagnostic modality may have resulted frommisclassification of SIBO diagnosis owing to its low sensitivityand variable specificity, leading to inaccurate and inconsistentoutcomes. One possible contributor to the observed inconsis-tency was the lack of a universally accepted cut-off threshold formany of the breath tests used to diagnose SIBO, even within asingle testing modality. Such inconsistency likely detracts fromthe usability of GHBT and other diagnostic breath tests forSIBO diagnosis. Our data also highlighted the possibility thatGHBT may be inadequate to accurately diagnose SIBO, espe-cially as an outcome measure in research studies. Althoughduodenal/jejunal aspirate culture remains the gold standard fordiagnosis of SIBO, its cost and invasiveness would likely pre-vent its widespread use in the clinical setting and for epidemi-ologic study. Therefore, further refinement of current breath
Figure 4. Subgroup analysesof SIBO risk with PPI use. (A)Subgroup analysis of SIBO riskwith PPI use based on SIBOtesting modality (duodenal/jeju-nal aspirate culture vs GHBT). (B)Subgroup analysis based onstudy design (pretest-posttest/self-control vs separate control).
testing modalities or development of a new, accurate, low-cost,
boptpaStrijbwiswfma
penriaaalivsOabsmpc
rooiisdi
May 2013 PPI USE AND SIBO 489
and noninvasive diagnostic test for SIBO is urgently needed forboth clinical and investigative use.
The second subgroup analysis, which examined the possiblerole of study design, showed a statistically significant associa-tion between PPI use and SIBO in studies comparing the samecohort of patients before and after PPI exposure. On the otherhand, studies comparing SIBO diagnosis in separate cohorts ofpatients with and without PPI exposure showed no association.These results may reflect the underlying pathophysiology andpatient-specific nature of SIBO. Prior studies of duodenal/jejunal aspirate cultures showed that SIBO may be caused byincreased levels of regular flora or bacterial species that have notyet been identified.36,37 Indeed, the present findings support the
elief that the pathophysiology and exact bacterial compositionf SIBO may be unique for each individual, thus leading tootential inconsistent outcomes when different groups of pa-ients are compared against each other. In addition, althoughrevious studies have identified several characteristics such asge, anatomy, and motility disorders that may predispose toIBO,38 there are likely many other currently unknown factorshat contribute to the alteration in intraluminal bacterial envi-onment associated with SIBO. These undiscovered factors mayntroduce confounding to studies featuring 2 cohorts of sub-ects. Even in the case-control studies included, subjects coulde matched only by demographics or known factors associatedith SIBO, which may not be sufficient owing to the complex-
ties of SIBO pathophysiology. Therefore, subgroup analysis oftudies featuring subjects who served as their own controls, inhich the intestinal flora was most comparable and unknown
actors were least likely to contribute to the results, showed aore consistent and significant association between PPI use
nd the development of SIBO.One limitation of this study was the detection of possible
ublication bias. The funnel plot (Figure 4) suggests the pres-nce of 4 positive, outlying studies that were not balanced byegative studies. Additional investigations including referenceeview did not reveal any further peer-reviewed studies fornclusion. Although this may represent publication bias, it maylso reflect a truly significant positive relationship between PPInd SIBO. Given the controversy in the field, the clinical utilitynd impact of even a negative study makes publication bias lessikely. However, as stated earlier, the subgroup analysis of stud-es using duodenal/jejunal aspirate testing to detect SIBO re-ealed an association between PPI use and SIBO that remainedignificant even after adjusting for potential publication bias.ther expected limitations of meta-analyses, including lack of
ccess to primary data, and covariates of interest that could note accounted for individually, such as age of subjects andpecific medical diagnoses, should be considered in the assess-
ent of these findings. In addition, although they were stronglyositive, the wide CIs detected in the subgroup analyses requireare in further interpretation of the results.
Finally, caution should be taken in the translation of theseesults to clinical practice. Prior studies have shown significantverlap between symptoms associated with SIBO and manyther functional disorders of the gastrointestinal tract, such as
rritable bowel syndrome.39 – 41 However, results from SIBO test-ng have been inconsistent among patients with irritable bowelyndrome or functional gastrointestinal symptoms.42– 44 In ad-ition, variable response to antibiotic treatment for SIBO both
n terms of symptoms and follow-up testing further highlights 1
the complexity in the pathophysiology and epidemiology ofgastrointestinal symptoms, functional and motility disorders,and SIBO. Therefore, a clear causal relationship between posi-tive SIBO testing (including both duodenal/jejunal aspirateculture and breath testing) and gastrointestinal symptoms hasnot been firmly established. Because SIBO testing results, ratherthan clinical symptoms, have been the primary measured out-comes of this meta-analysis and most studies to date on therelationship between PPI and SIBO, it remains to be determinedwhether the observed increase in SIBO among PPI users actu-ally correlates with significant clinical outcomes such as symp-toms, malabsorption, or mortality.
In conclusion, this meta-analysis showed a statistically sig-nificant association between PPI use and SIBO, only when thediagnosis was made by a highly accurate testing modality (du-odenal/jejunal aspirate culture). Studies using GHBT to diag-nose SIBO did not result in a significant outcome. Prior con-flicting outcomes in the published literature may have resultedfrom inconsistencies in the diagnostic methods and cut-offvalues used in those studies. Both clinicians and patientsshould be judicious in the use of PPI and consider dose-taper-ing whenever possible. In addition, a high level of suspicionshould be raised when evaluating PPI users presenting withsymptoms or signs suggestive of SIBO. Further studies areneeded to refine the diagnostic modalities for SIBO, character-ize the underlying pathophysiology of SIBO associated withacid suppression, and assess the impact of this relationship onclinical outcomes.
References
1. IMS Health. IMS Health reports US prescription sales grew 5.1 percentin 2009, to $300.3 billion. Press release. IMS health. Available at:http://www.imshealth.com/portal/site/imshealth/menuitem.a46c6d4df3db4b3d88f611019418c22a/?vgnextoid�d690a27e9d5b7210VgnVCM100000ed152ca2RCRD&vgnextchannel�41a67900b55a5110VgnVCM10000071812ca2RCRD&vgnextfmt�default. Acces-sed April 1, 2010.
2. Bartholow M. Top 200 prescription drugs of 2009. PharmacyTimes, 2010. Available at: http://www.pharmacytimes.com/issue/pharmacy/2010/May2010/RxFocusTopDrugs-0510.
3. Yachimski PS, Farrell EA, Hunt DP, et al. Proton pump inhibitorsfor prophylaxis of nosocomial upper gastrointestinal tract bleed-ing: effect of standardized guidelines on prescribing practice.Arch Intern Med 2010;170:779–783.
4. Cook D, Heyland D, Griffith L, et al. Risk factors for clinicallyimportant upper gastrointestinal bleeding in patients requiringmechanical ventilation. Canadian Critical Care Trials Group. CritCare Med 1999;27:2812–2817.
5. Kirsch M. Bacterial overgrowth. Am J Gastroenterol 1990;85:231–237.
6. Williams C. Occurrence and significance of gastric colonizationduring acid-inhibitory therapy. Best Pract Res Clin Gastroenterol2001;15:511–521.
7. Theisen J, Nehra D, Citron D, et al. Suppression of gastric acidsecretion in patients with gastroesophageal reflux disease re-sults in gastric bacterial overgrowth and deconjugation of bileacids. J Gastrointest Surg 2000;4:50–54.
8. Pereira SP, Gainsborough N, Dowling RH. Drug-induced hypochlo-rhydria causes high duodenal bacterial counts in the elderly.Aliment Pharmacol Ther 1998;12:99–104.
9. Rana SV, Bhardwaj SB. Small intestinal bacterial overgrowth.Scand J Gastroenterol 2008;43:1030–1037.
0. Hoffmann JC, Zeitz M. Small bowel disease in the elderly: diar-
490 LO AND CHAN CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 11, No. 5
rhoea and malabsorption. Best Pract Res Clin Gastroenterol2002;16:17–36.
11. Husebye E, Skar V, Høverstad T, et al. Fasting hypochlorhydriawith gram positive gastric flora is highly prevalent in healthy oldpeople. Gut 1992;33:1331–1337.
12. Welage LS, Carver PL, Revankar S, et al. Alterations in gastricacidity in patients infected with human immunodeficiency virus.Clin Infect Dis 1995;21:1431–1438.
13. Fried M, Siegrist H, Freil R, et al. Duodenal bacterial overgrowthduring treatment in outpatients with omeprazole. Gut 1994;35:23–26.
14. Nelis GF, Engelage AH, Samson G. Does long-term inhibition ofgastric acid secretion with omeprazole lead to small intestinalbacterial overgrowth? Neth J Med 1994;45:93–100.
15. Lewis SJ, Franco S, Young G, et al. Altered bowel function andduodenal bacterial overgrowth in patients treated with omepra-zole. Aliment Pharmacol Ther 1996;10:557–561.
16. Thorens J, Froehlich F, Schwizer W, et al. Bacterial overgrowthduring treatment with omeprazole compared with cimetidine: aprospective randomised double blind study. Gut 1996;39:54–59.
17. Hutchinson S, Logan R. The effect of long-term omeprazole onthe glucose-hydrogen breath test in elderly patients. Age Ageing1997;26:87–89.
18. Law D, Pimentel M. Proton pump inhibitor therapy does not affecthydrogen production on lactulose breath test in subjects withIBS. Dig Dis Sci 2010;55:2302–2308.
19. Lombardo L, Foti M, Ruggia O, et al. Increased incidence of smallintestinal bacterial overgrowth during proton pump inhibitor ther-apy. Clin Gastroenterol Hepatol 2010;8:504–508.
20. Choung RS, Ruff KC, Malhotra A, et al. Clinical predictors of smallintestinal bacterial overgrowth by duodenal aspirate culture. Ali-ment Pharmacol Ther 2011;33:1059–1067.
21. Compare D, Pica L, Rocco A, et al. Effects of long-term PPItreatment on producing bowel symptoms and SIBO. Eur J ClinInvest 2011;41:380–386.
22. Ratuapli SK, Ellington TG, O’Neill MT, et al. Proton pump inhibitortherapy use does not predispose to small intestinal bacterialovergrowth. Am J Gastroenterol 2012;107:730–735.
23. Spiegel BM, Chey WD, Chang L. Bacterial overgrowth and irritablebowel syndrome: unifying hypothesis or a spurious consequenceof proton pump inhibitors? Am J Gastroenterol 2008;103:2972–2976.
24. Wells GA, Shea B, O’Connell D, et al. The Newcastle-Ottawascale (NOS) for assessing the quality of nonrandomized studiesin meta-analyses. Ottawa Hospital Research Institute. Availableat: http://www.ohri.ca/programs/clinical_epidemiology/oxford.htm.
25. Duval S, Tweedie R. Trim and fill: a simple funnel-plot-basedmethod of testing and adjusting for publication bias in meta-analysis. Biometrics 2000;56:455–463.
26. Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observa-tional studies in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies in Epidemiology (MOOSE)group. JAMA 2000;283:2008–2012.
27. Gough A, Andrews D, Bacon PA, et al. Evidence of omeprazole-induced small bowel bacterial overgrowth in patients with sclero-derma. Br J Rheumatol 1995;34:976–977.
28. Eurich DT, Sadowski CA, Simpson SH, et al. Recurrent commu-nity-acquired pneumonia in patients starting acid-suppressing
drugs. Am J Med 2010;123:47–53.29. Targownik LE, Lix LM, Metge CJ, et al. Use of proton pumpinhibitors and risk of osteoporosis-related fractures. CMAJ 2008;179:319–326.
30. Leonard J, Marshall JK, Moayyedi P. Systematic review of the riskof enteric infection in patients taking acid suppression. Am JGastroenterol 2007;102:2047–2056.
31. Hawkey CJ, Bosch K, Richter JE, et al, editors. Textbook ofclinical gastroenterology and hepatology. Hoboken, NJ: Wiley-Blackwell, 2012:308.
32. Donald IP, Kitchingmam G, Donald F, et al. The diagnosis of smallbowel bacterial overgrowth in elderly patients. J Am Geriatr Soc1992;40:692–696.
33. Strocchi A, Furne J, Ellis C, et al. Methanogens outcompetesulphate reducing bacteria for H2 in the human colon. Gut 1994;35:1098–1101.
34. Romagnuolo J, Schiller D, Bailey RJ. Using breath tests wisely ina gastroenterology practice: an evidence-based review of indica-tions and pitfalls in interpretation. Am J Gastroenterol 2002;97:1113–1126.
35. Khohsini R, Dai SC, Lezcano S, et al. A systematic review ofdiagnostic tests for small intestinal bacterial overgrowth. Dig DisSci 2008;53:1443–1454.
36. Sullivan A, Törnblom H, Lindberg G, et al. The micro-flora of thesmall bowel in health and disease. Anaerobe 2003;9:11–14.
37. Bouhnik Y, Alain S, Attar A, et al. Bacterial populations contam-inating the upper gut in patients with small intestinal bacterialovergrowth syndrome. Am J Gastroenterol 1999;94:1327–1331.
38. Dukowicz AC, Lacy BE, Levine GM. Small intestinal bacterialovergrowth: a comprehensive review. Gastroenterol Hepatol2007;3:112–122.
39. Pyleris E, Giamarellos-Bourboulis EJ, Tzivras D, et al. The preva-lence of overgrowth by aerobic bacteria in the small intestine bysmall bowel culture: relationship with irritable bowel syndrome.Dig Dis Sci 2012;57:1321–1329.
40. Sachdeva S, Rawat AK, Reddy RS, et al. Small intestinalbacterial overgrowth (SIBO) in irritable bowel syndrome: fre-quency and predictors. J Gastroenterol Hepatol 2011;26(Suppl 3):135–138.
41. Reddymasu SC, Sostarich S, McCallum RW. Small intestinalbacterial overgrowth in irritable bowel syndrome: are there anypredictors? BMC Gastroenterol 2010;10:23.
42. Ford AC, Spiegel BM, Talley NJ, et al. Small intestinal bacterialovergrowth in irritable bowel syndrome: systematic review andmeta-analysis. Clin Gastroenterol Hepatol 2009;7:1279–1286.
43. Shah ED, Basseri RJ, Chong K, et al. Abnormal breath testing inIBS: a meta-analysis. Dig Dis Sci 2010;55:2441–2449.
44. Posserud I, Stotzer PO, Björnsson ES, et al. Small intestinalbacterial overgrowth in patients with irritable bowel syndrome.Gut 2007;56:802–808.
Reprint requestsAddress requests for reprints to: Walter W. Chan, MD, MPH, Division
of Gastroenterology, Brigham and Women’s Hospital, 75 FrancisStreet, Boston, Massachusetts 02115. e-mail: [email protected];fax: (617) 525-0338.
Conflicts of interest
The authors disclose no conflicts.