CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2007;5:178 –185
Effects of Proton-Pump Inhibitors on Functional Dyspepsia:
A Meta-analysis of Randomized Placebo-Controlled Trials
WEI HONG WANG,*,‡ JIA QING HUANG,‡,§ GE FAN ZHENG,‡ HARRY H. X. XIA,‡ WAI MAN WONG,‡ XIN GUANG LIU,*
JOHAN KARLBERG,§ and BENJAMIN C. Y. WONG‡
*Department of Gastroenterology, Peking University First Hospital, Beijing, China; the ‡Department of Medicine and the §Clinical Trials Centre, Faculty of Medicine,
University of Hong Kong, Hong Kong, China
See CME exam on page 140.
Background & Aims: The aim of this study was to
assess systematically the efficacy of proton pump inhib-
itors (PPIs) in the treatment of functional dyspepsia
compared with placebo and to determine if any differ-
ence in the response exists between symptom subgroups
of functional dyspepsia. Methods: A literature search
was performed through September 2005 in PubMed, Med-
line, Embase, CINAHL, and Cochrane databases to include
randomized, double-blind, placebo-controlled trials evalu-
ating the efficacy of PPIs for the treatment of functional
dyspepsia. Relative risk (RR) and relative risk reduction
(RRR) and 95% confidence intervals (CI) were calculated
under a random-effects model. Results: Seven studies
with a total of 3725 patients were identified. PPIs were
found to be more effective than placebo for reducing
symptoms in patients with functional dyspepsia (RRR,
10.3%; 95% CI, 2.7%–17.3%). The estimated number
needed to treat is 14.6 (95% CI, 8.7–57.1). When stratified
analyses were performed, a significant difference in the
efficacy was observed only in patients with ulcer-like
(RRR, 12.8%; 95% CI, 7.2%–18.1%) and reflux-like dyspep-
sia (RRR, 19.7%; 95% CI, 1.8%–34.3%), but not in those
with dysmotility-like (RRR, 5.1%; 95% CI, ⴚ10.9% to
18.7%) and unspecified dyspepsia (RRR, ⴚ8.0%; 95% CI,
ⴚ23.7% to 5.6%). The effect of H pylori on the efficacy of
PPIs remains unclear. Significant heterogeneity among
studies was found for the overall analysis, dysmotility-
like dyspepsia, H pylori–negative subgroup, and different
dose subgroups. Conclusions: PPIs are more effective
than placebo for the management of patients with ulcer-
like and reflux-like functional dyspepsia.
F unctional, or nonulcer, dyspepsia (NUD) is defined as
persistent or recurrent pain or discomfort centered in the
upper abdomen for more than 12 weeks in the preceding 12
months with no evidence of an organic disease that is likely to
explain the symptoms.1 NUD has a significant adverse impact
on patients’ quality of life and economic consequences.2,3 How-
ever, because the pathophysiology of functional dyspepsia is
understood poorly, symptom control is currently the main-
stream of management. Eradication of Helicobacter pylori and
acid suppression are used commonly in the treatment of func-
tional dyspepsia, although conflicting results regarding their
efficacy have been reported in the literature.4 –11 Most studies
have shown that proton pump inhibitors (PPIs) are significantly
more effective than placebo for treating patients with func-
tional dyspepsia.4 –10 However, in a recent large, well-designed,
randomized study conducted in Hong Kong, Wong et al11
reported that lansoprazole 15 mg or 30 mg once daily given for
4 weeks showed similar efficacy to placebo for treating patients
with functional dyspepsia. Thus, it is not clear to the practicing
physicians whether PPIs should be recommended for use in
controlling symptoms in patients with functional dyspepsia.
Functional dyspepsia can be divided into 4 symptom groups:
ulcer-like, dysmotility-like, reflux-like, and unspecified dyspep-
sia.12 Because patients with reflux-like dyspepsia are likely to
have gastroesophageal reflux disease, they therefore are no
longer considered as functional dyspepsia according to the
Rome II criteria.1 Symptom overlap is very common in patients
with functional dyspepsia12,13 and the efficacy of treatments
targeting specific symptoms has not been well studied. Al-
though our comprehensive search identified 2 published sys-
tematic reviews evaluating the effect of PPIs in NUD,10,14 they
suffered several serious methodologic flaws such as an incom-
plete literature search, inclusion of duplicate publications, and
no stratified analysis by dose of PPIs or different types of
symptoms.10,14 A more recent, well-conducted, systematic re-
view shows that PPIs have a small but statistically significant
effect on symptoms of NUD.15 However, this article included a
study16 that was part of a multinational trial4 and missed a
study that met the inclusion criteria.17
Therefore, we performed this meta-analysis to evaluate the
efficacy of PPIs for treating patients with functional dyspepsia
in comparison with placebo. We also assessed if there was any
difference in the response to PPI treatment between patients in
different symptom subgroups.
Materials and Methods
Literature Search
A comprehensive computerized literature search was
performed using the PubMed, Medline, Embase, CINAHL,
Cochrane Controlled Trials Register databases for clinical
Abbreviations used in this paper: CI, confidence interval; NNT, num-
ber needed to treat; NUD, nonulcer dyspepsia; PPI, proton pump
inhibitor; RR, relative risk; RRR, relative risk reduction.
© 2007 by the AGA Institute
1542-3565/07/$32.00
doi:10.1016/j.cgh.2006.09.012
February 2007
trials published through September 2005 without any lan-
guage restriction. The following search terms as either MeSH
terms or text words were used in various combinations:
non-ulcer, functional, dyspepsia, NUD, in combination with
proton pump inhibitors, PPI, and the name of each respec-
tive drug (omeprazole, lansoprazole, pantoprazole, rabapra-
zole, and esomeprazole). The search was limited to clinical
trials only. Meeting abstracts were searched from CD-ROMs
of major gastrointestinal meetings (American Digestive Dis-
ease Week, American College of Gastroenterology, World
Congress of Gastroenterology, and United European Gastro-
enterology Week) held in the past 8 years using the earlier-
described search terms. The titles and abstracts of all poten-
tially relevant studies were screened for their relevance before
retrieval of full articles. Full articles also were scrutinized for
relevance if the title and abstract were ambiguous. A fully
recursive search also was performed from the reference lists
of all retrieved articles to ensure a complete and comprehen-
sive search of the published literature. All searches were
conducted independently by 3 reviewers (W.H.W., G.F.Z., and
J.Q.H.) and their search results were combined.
Study Selection
Studies that met the following criteria were included:
(1) randomized, double-blind, placebo-controlled trials evaluat-
ing the efficacy of PPIs for treating patients with functional
dyspepsia; (2) functional dyspepsia defined as persistent or
recurrent dyspepsia with no evidence of organic disease to
explain the patient’s symptoms; (3) studies performed in an
adult population; and (4) studies that reported information
about the number of patients treated and the number of pa-
tients with improvement.
Studies in which patients had gastroesophageal reflux dis-
ease, peptic ulcer disease, nonsteroidal anti-inflammatory drug–
related gastropathy, irritable bowel syndrome, biliary tract dis-
ease, or pancreatic disease were excluded. We also excluded
duplicate publications or studies that did not provide raw data,
or that included patients who received a combination of treat-
ments.
Data Extraction
Data were extracted from each study by 2 reviewers
(W.H.W. and G.F.Z.) independently and entered into a comput-
erized database. Any differences were resolved by discussion to
reach consensus between the authors. The information re-
trieved covered study design, publication type, patient char-
acteristics (age, sex, history of dyspepsia, dyspepsia symp-
toms, H pylori status), diagnostic criteria, treatment regimen
(drug, dose, duration), outcome measures, and study out-
comes. Data on the number of patients showing improve-
ment and the status of H pylori infection before and after
treatment were extracted according to intention-to-treat cri-
teria. If possible, data also were extracted by subgroups of
symptoms (ie, ulcer-like, reflux-like, dysmotility-like, and un-
specified dyspepsia).
Quality Assessment
Study quality was assessed by a series of validity criteria,
including study design, level of blinding, method of random-
ization, diagnostic criteria of dyspepsia, baseline characteristics,
PPIS ON FUNCTIONAL DYSPEPSIA 179
patient compliance, and outcome measures, and analysis by
intention to treat criteria. Any discrepancies in quality assess-
ment were resolved by consensus among the authors. No qual-
ity score was assigned to any study to avoid possible introduc-
tion of subjectivity by the authors. Instead, we used these
validity criteria to rank the studies (Table 1).
Statistical Analysis
For the meta-analysis the terms no symptoms, no symp-
toms for further management, and improvement on gastrointestinal
symptom rating scale after the completion of treatment were
defined as treatment success. Relative risk (RR) and relative risk
reduction (RRR) ([1 ⫺ RR] ⫻ 100%) together with 95% confi-
dence intervals (CIs) were calculated to estimate the efficacy of
intervention using raw data from selected studies and then
statistically combined under a random-effects model.18 The
number needed to treat (NNT) also was calculated by eval-
uating the mean proportion of patients with continued dys-
pepsia in the placebo groups of the trials (P2). The NNT ⫽
1 ⁄ 共P2 ⫻ 关1 ⫺ RR兴兲 and the 95% CIs were obtained from the
95% CI of the RR.
Meta-regression was performed to explore the influence
of the following factors on treatment effects: (1) the dose of
PPI, (2) the proportion of patients with predominant symp-
toms, (3) the proportion of patients with H pylori infection, and
(4) the duration of treatment. A standard dose of PPI included
omeprazole 20 mg/day and lansoprazole 30 mg/day, and low-
dose PPI was defined as omeprazole 10 mg/day and lansopra-
zole 15 mg/day. The data for different dyspepsia subgroups also
were pooled to evaluate if there was any difference in the
response to PPI treatment between patients in different symp-
tom subgroups.
Statistical heterogeneity among studies was assessed using the
Q value calculated from the Mentel–Haenszel method. In the
presence of statistical heterogeneity, we searched for the sources of
any possible clinically important heterogeneity (ie, methodologic
or biological heterogeneity). We did not simply exclude outliers on
the basis of statistical tests of heterogeneity. Furthermore, to
exclude any possible influence of a single study, we performed a
sensitivity analysis to evaluate whether the exclusion of any single
study substantially altered the magnitude or statistical result of
the summary estimate. The potential effect of any publication bias
was assessed using a funnel plot of log RR vs precision of individ-
ual studies as suggested by Egger et al.19
All analyses were performed using Comprehensive Meta-
analysis software (version 1.0.25, Biostat, Englewood, NJ).
Results
Literature Search and Study Selection
We identified a total of 716 citations with the comput-
erized search and 398 meeting abstracts from the CD-ROMs of
major gastrointestinal meetings. A manual search of the ref-
erence list of the retrieved articles yielded 1 additional study.
Screening of the title and abstract of the publications iden-
tified 31 potentially relevant studies for full article retrieval.
Of these, 7 studies met the inclusion criteria,4 – 6,11,17,20,21
and 24 studies subsequently were excluded for the following
reasons: 8 studies evaluated patients with uninvestigated
180 WANG ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 2

Table 1. Characteristics of Studies Selected for Meta-Analysis

Diagnostic No. of Definition of
Study method Inclusion criteria Exclusion criteria Treatment patients symptom relief

Wong et al11 Endoscopy Rome II criteria (at least Reflux symptoms LAN 30 mg, 453 Complete relief at
12 wk within the without epigastric 4 wk the last 3 days of
preceding 12 mo), symptoms, RE, PU, LAN 15 mg, treatment
symptoms in 2 weeks erosive change in 4 wk
before endoscopy stomach or
duodenum, alcohol,
history of PU and
GERD, NSAID,
severe illness, IBS,
constipation
Talley et al4 Endoscopy ⬍5 At least 1 of the 3 days Reflux symptoms OME 20 mg, 1248 Complete relief at
gastric erosion before randomization, without epigastric 4 wk the last 3 days of
1-mo history of symptoms, RE, PU, OME 10 mg, treatment
dyspeptic symptoms, erosive change in 4 wk
symptoms occur on stomach/duodenum,
⬎25% of days during upper-GI surgery,
that month alcohol, history of
PU and GERD, IBS,
severe illness,
alarm symptom,
constipation
Peura et al20 Endoscopy ⬍5 Symptoms for at least Predominant LAN 30 mg, 921 Complete relief
gastric erosion 30% of the days symptoms indicating 8 wk during the 3 days
during the GERD or IBS, PU, LAN 15 mg, before visit
pretreatment, 3 ⬎5 gastric erosions 8 wk
months history of
dyspeptic symptoms
Bolling-Sternevald Endoscopy ⬍5 At least 3 days in the Predominant OME 20 mg, 196 Complete relief at
et al5 gastric erosion run-in week symptoms symptoms indicating 2 wk the last days of
24-h pH for at least 1 month GERD or IBS, history treatment
monitoring of pu and GERD,
upper-GI surgery,
alarm symptom,
NSAID
Blum et al6 Endoscopy ⬍10 At least 3 days in the Reflux symptoms OME 20 mg, 598 No symptom
gastric erosion run-in week symptoms without concomitant 2 wk for further
Sonography, CT for at least 1 month epigastric OME 10 mg, management
symptoms, IBS, RE, 2 wk
PU, alcohol
abdominal surgery,
alarm symptom
Gerson and Endoscopy Rome II criteria (at least RE, predominant OME 20 mg, 40 Improvement on
Triadafilopoulos17 12 wk within the GERD symptoms, 4 wk GSRS score
preceding 12 mo), malignancy, NSAID between baseline
H pylori negative, lack and 4 weeks
of symptom relief on
H2-blocker
Hengels21 Endoscopy Symptoms for at least Unknown LAN 15 mg, 269 No symptoms
sonography 1 week 2 wk during the last 5
days of treatment

NOTE. Studies were ranked according to their study quality.

RE, reflux esophagitis; PU, peptic ulcer; LAN, lansoprazole; GERD, gastroesophageal reflux disease; NSAID, nonsteroidal anti-inflammatory drug;
IBS, irritable bowl syndrome; OME, omeprazole; GI, gastrointestinal; GSRS, gastrointestinal symptom rating.

dyspepsia without endoscopy,22–29 7 studies had no placebo-
controlled group,30 –36 1 study was a follow-up study of
patients receiving treatment for dyspepsia,37 1 study had no
raw data,38 6 studies were duplicate publications,39 – 44 and 1
study16 reported results that were used by a multinational
trial.4 When duplicate publications were found, we only
included the article that provided the most comprehensive
information.
Study Characteristics and Quality
Of the 7 studies, 6 were published as full arti-
cles,4 – 6,11,17,20 and 1 was in abstract form.21 Four studies
February 2007
Table 2. Statistical Results Comparing PPIs With Placebo in Subgroups of Functional Dyspepsia Under Random-Effects
Model
No. of No. of
Groups studies patients RR; 95% CI
Ulcer-like 3 1405 0.87; 0.82–0.93
Reflux-like 3 380 0.80; 0.66–0.98
Dysmotility-like 3 583 0.95; 0.81–1.11
Dysmotility-likea 2 339 0.84; 0.69–1.02
Unspecified 4 289 0.08; 0.94–1.24
aExcluding one heterogeneous study.11
compared omeprazole4 – 6,17 with placebo, and 3 compared
lansoprazole11,20,21 with placebo. Six studies recruited pa-
tients with symptoms for at least 1 or 3 months,4 – 6,11,17,20
and 1 study did not provide information about patients’
previous history of functional dyspepsia.21 The Rome II
criteria were used to select patients by 2 studies.11,17 In all
studies, endoscopy was performed before the enrollment of
patients. Patients with a history of peptic ulcer disease were
excluded in 5 studies.4 – 6,11,20 Gastroesophageal reflux dis-
ease was considered an exclusion criterion in 6 stud-
ies.4 – 6,11,17,20 However, this information was not available
from 1 abstract (Table 1).21 All studies provided information
about H pylori assessment.4 – 6,11,17,20,21 Detailed information
on subgroups of symptoms also was provided by 4 stud-
ies.4,11,17,20
All studies were performed in a double-blind fashion.
Computer-generated randomization was reported in 4 stud-
ies,4,5,11,20 and this information was not given in the remaining
3 (Table 1).6,17,20,21 Demographic information was reported in
6 studies.4 – 6,11,17,20 The mean patient age was 43.1 ⫾ 2.1 years
(⫾SEM) and 37% were male. The duration of medical therapy in
the studies varied among 2 weeks,5,6,21 4 weeks,4,11,17 and 8
weeks (Table 2).20 With respect to the outcomes, no symptoms
during the last 3 days of treatment was considered as complete
relief by 3 studies,4,11,20 whereas a different definition of symp-
tom relief was used in the remaining 4 studies.5,6,17,21
Overall Analysis
There were a total of 7 studies consisting of 3725
patients (PPI, 2387; placebo, 1338). There was a modest but
statistically significant difference in symptom relief in patients
receiving PPIs (40.3%) compared with those given placebo
(32.7%), yielding a RRR of 10.3% (95% CI, 2.7%–17.3%, test for
heterogeneity, Q ⫽ 31.46, P ⫽ .0005). The estimated NNT was
14.6 patients (95% CI, 8.7–57.1 patients). The individual and
summary RRs and 95% CIs are shown in Figure 1.
To avoid possible bias of the analysis caused by the inclusion
of an abstract,21 we performed a sensitivity analysis by exclud-
ing this study, yielding a RRR of 8.8% (95% CI, 1.2%–15.9%; test
for heterogeneity, Q ⫽ 27.16; P ⫽ .001). No significant differ-
ence in RRR was observed between the 2 analyses (P ⫽ .97).
Test for Heterogeneity and Sensitivity Analyses
There was significant heterogeneity among the 7 in-
cluded studies for the overall analysis (Q ⫽ 31.46, P ⫽ .0005).
Meta-regression did not find any relationship between the es-
PPIS ON FUNCTIONAL DYSPEPSIA 181
Test of
heterogeneity
RRR (%); 95% CI NNT; 95% CI Q value P value
12.8; 7.2–18.1 10.9; 7.6–19.5 1.746 .88
19.7; 1.8–34.3 6.8; 3.9–74.1 9.26 .10
5.1; ⫺10.9 to 18.7 27.1; ⫺12.7 to 7.4 13.47 .04
15.8; ⫺2.2 to 30.6 8.5; ⫺68.0 to 4.4 5.30 .15
⫺8.0; ⫺23.7 to 5.6 ⫺18.8; ⫺6.3 to 26.9 2.75 .84
timated RRR of dyspeptic symptoms and the proportion of
patients with predominant reflux-like symptoms (P ⫽ .101) and
H pylori infection (P ⫽ .146), and the dose (P ⫽ .340) and
duration of PPI treatment (P ⫽ .255). Therefore, any of these
prespecified factors could explain the heterogeneity between
trials.
No further heterogeneity (Q ⫽ 11.06, P ⫽ .20) was found
after exclusion of the study by Wong et al,11 suggesting that the
heterogeneity was caused by this study. Looking at the inclu-
sion criteria, it can be seen that this large study included
Chinese patients, who are known to have a relatively low prev-
alence of gastroesophageal reflux disease.45 No additional con-
founding factors such as study design, level of blinding, com-
pliance of patient, and definition of outcomes were identified.
Sensitivity analysis by excluding the data from the study by
Wong et al11 showed no difference in the overall analysis (RRR,
14.7%; 95% CI, 9.5%–19.5%).
Dose Effect
A low dose of PPIs was compared with placebo in 5 studies
involving 2418 patients,4,6,11,20,21 whereas standard-dose PPIs were
used in 6 studies with a total of 2392 patients.4 – 6,11,17,20 There
was a clear trend that both low-dose and standard-dose PPIs
were superior to placebo for the treatment of functional dys-
pepsia (Figure 1). Meta-regression did not suggest that the dose
of PPI influenced the estimated RRR of dyspeptic symptoms
(P ⫽ .340). The individual and summary RRs and 95% CIs from
the individual trials and combined estimates of low- and stan-
dard-dose PPIs are shown in Figure 1.
Helicobacter pylori Infection
A total of 1394 H pylori–positive patients from 6 studies
were available for analysis.4 – 6,11,20,21 There was a significant
reduction in dyspeptic symptoms in patients receiving PPIs
compared with those given placebo, with a RRR of 13.3% (95%
CI, 5.0%–20.9%; test for heterogeneity, Q ⫽ 14.45; P ⫽ .11). The
NNT was estimated at 11.4 patients (95% CI, 7.2–30.2 patients).
A total of 2284 H pylori–negative patients from 7 studies
were analyzed.4 – 6,11,17,20,21 There was a trend toward PPIs being
more effective than placebo for symptom improvement, al-
though the difference was not statistically significant (RRR,
5.7%; 95% CI, ⫺5.3% to 15.5%). Meta-regression did not indicate
that the effect of PPIs on dyspeptic symptoms was associated
with H pylori infection (P ⫽ .146).
182 WANG ET AL
Dyspepsia Subgroup
Four of the 7 studies provided raw data on the number
of patients with ulcer-like, reflux-like, dysmotility-like, and un-
specified dyspepsia according to the predominance of symp-
toms.4,11,17,20 One study was excluded from the analysis of
ulcer-like, reflux-like, and dysmotility-like subgroups because of
a small number of patients (nⱕ2).17 PPIs were found to be more
effective than placebo for reducing symptoms in patients with
ulcer-like and reflux-like functional dyspepsia, with a RRR of
12.8% (95% CI, 7.2%–18.1%; test for heterogeneity, Q ⫽ 1.75; P
⫽ .88) and 19.7% (95% CI, 1.8%–34.3%; test for heterogeneity, Q
⫽ 9.26; P ⫽ .10), respectively. The NNT with PPIs to improve 1
case of dyspepsia was 10.9 (95% CI, 7.6 –19.5) and 6.8 (95% CI,
3.9 –74.1), respectively. The RRRs for dysmotility-like and un-
specified subgroups were 5.1% (95% CI, ⫺10.9% to 18.7%; test
for heterogeneity, Q ⫽ 13.47; P ⫽ .04) and ⫺8.0% (95% CI,
⫺23.7% to 5.6%; test for heterogeneity, Q ⫽ 2.75; P ⫽ .84),
respectively, indicating that PPIs were not superior to placebo
for the treatment of dysmotility-like and unspecified dyspepsia.
Publication Bias
The funnel plots obtained by plotting the log RRs vs
precision (1/SE) of individual studies was symmetric for overall
and subgroup analyses, suggesting that there was no significant
publication bias in this literature (data not shown).
Discussion
Our present meta-analysis shows that PPIs were signif-
icantly more effective than placebo for treating patients with
CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 2
Figure 1. Individual and summary
relative risks and 95% CIs from 7
studies comparing PPIs with placebo
for the treatment of functional dyspep-
sia.
functional dyspepsia, with a 10.3% reduction in the RR of
dyspeptic symptoms and an NNT of 14.6 patients. This finding
was consistent across different doses of PPIs and the patients’
status of H pylori infection.
Previous studies have shown that there was an extensive
overlap among patients with different subgroups of functional
dyspepsia.12,13 We found 4 studies that provided information
on symptom subgroups.4,11,17,20 The combined analysis of
these studies showed that, in patients with reflux-like and
ulcer-like dyspepsia, the relative risk reduction of dyspepsia was
significantly greater in patients receiving PPIs than those
treated with placebo. However, no significant difference in the
efficacy was observed between PPIs and placebo in patients with
dysmotility-like and unspecified dyspepsia. These results have
provided some evidence that gastric acid may play a role in the
pathogenesis of ulcer-like or reflux-like functional dyspepsia
and the results generated from the subgroup analysis of symp-
toms could have potentially clinical use for predicting patients’
response to the treatment.
It is possible that patients with reflux-predominant dys-
pepsia are actually those with gastroesophageal reflux disease
because there is a considerable overlap in symptoms between
functional dyspepsia and nonerosive or endoscopic-negative
reflux disease.46,47 Although meta-regression found no sig-
nificant relationship between the RRR of dyspeptic symp-
toms and the proportion of patients with reflux-predomi-
nant dyspepsia, there was a trend that the reported benefit of
PPIs increased with the proportion of patients with reflux-
like dyspepsia in the individual studies.4,11,17,20 Therefore, it
February 2007
is conceivable that inclusion of patients with reflux-predom-
inant dyspepsia might increase the difference in symptom
relief between patients treated with PPIs and placebo. This
might be the rationale why, in the Rome II criteria, reflux-like
dyspepsia was excluded.1
There was statistically significant heterogeneity between
studies. Although meta-regression did not suggest that the
proportion of patients with predominant reflux-like symptoms
and H pylori infection, and the dose and duration of PPI treat-
ment, could explain the heterogeneity between trials, these
factors still may be important modifiers of the efficacy of PPI in
patients with functional dyspepsia. Based on the statistical
heterogeneity identified in the overall analysis, we found that
the study by Wong et al11 was responsible for all the heteroge-
neities identified. Further scrutiny revealed that this was a
considerably large trial involving 453 patients, and Wong et al11
did not find any significant difference in symptom relief be-
tween patients receiving PPIs and those given placebo regardless
of symptom subgroups. Three specific differences were found
between this study and the others. First, this study was per-
formed in Chinese patients, who are known to have a relatively
low prevalence of gastroesophageal reflux disease.45 Second, the
study population was a carefully selected cohort by a group of
gastroenterologists,11 resulting in the exclusion of most pa-
tients with gastroesophageal reflux disease. In this study, ap-
proximately 4% of patients reported acid reflux symptoms as
their predominant symptom and the majority of patients had
dysmotility-like symptoms (54%) as their predominant symp-
tom.11 Furthermore, a previous study revealed that patients
seen by gastroenterologists responded better to placebo than
those seeing general practitioners.4 In this study, a relatively
higher placebo response rate also was found. Third, the Rome II
criteria were used in this study to select patients.11 In the
analysis of the effect of PPIs for the treatment of functional
dyspepsia, the main concern is the admixture of patients with
nonerosive gastroesophageal reflux disease because patients
with nonerosive gastroesophageal reflux disease are known to
respond to PPIs.48 –50 Therefore, the observed efficacy of PPIs in
functional dyspepsia actually may be the result of treating
patients with nonerosive gastroesophageal reflux disease.16,51
However, this may not be the case in the study by Wong et al11
because of the adoption of the Rome II criteria for patient
recruitment.
The role of H pylori infection in functional dyspepsia has
long been a subject of controversy. Two meta-analyses have
come to opposite conclusions regarding whether H pylori erad-
ication leads to sustained symptom improvement in H pylori–
infected patients with functional dyspepsia.52,53 A recent sys-
tematic review concluded that H pylori eradication benefits
functional dyspepsia patients, but the effect size is likely to be
small. The reason for the differences between studies may relate
to the overall numbers of patients included in the analysis.54
Furthermore, there is evidence to show that a subset of patients
with functional dyspepsia who are infected with H pylori have
abnormal gastric acid secretion,55 suggesting a possible role of
gastric acid in functional dyspepsia. Virulent strains of H pylori
are associated with a higher prevalence of dyspeptic symptoms,
especially in patients with epigastric pain.56 Therefore, it is
reasonable to presume that this group of patients may respond
better to PPI treatment than H pylori–negative patients with
functional dyspepsia. However, in our meta-regression, we did
PPIS ON FUNCTIONAL DYSPEPSIA 183
not find that the variation in H pylori status was associated with
the variation in the efficacy of PPI treatment. At least part of the
reason for this may be because there were fewer studies in-
cluded in the meta-regression. Although meta-regression is an
extension to subgroup analyses that allows the effect of cate-
goric characteristics to be investigated, and in principle allows
the effects of multiple factors to be investigated simultaneously,
it is believed that at least 10 studies in a meta-analysis should
be available for each characteristic modeled to produce reliable
results.57 Thus, no firm conclusions can be made because of
concerns over the small sample size included in this study. The
question of whether PPI treatment is more effective in H pylori–
positive functional dyspepsia patients than those without the
infection warrants further investigation in future randomized
controlled trials.
As with meta-analysis in general, there were several limita-
tions to our meta-analysis.58,59 First, to avoid publication bias,
a phenomenon in which studies with positive results are more
likely accepted for publication than those with negative results,
we included both published studies and meeting abstracts in
our search strategy. However, a potential publication bias could
not be ruled out completely because 5 of the 7 studies included
in the analysis had positive results.4 – 6,20,21 Because the funnel
plots were symmetric for both the overall and subgroup anal-
yses, any possible publication bias might not be significant if it
exists. Second, the lack of consensus on the diagnosis of func-
tional dyspepsia and the use of different outcome measure-
ments of treatment has hampered us from comparing any
difference in the efficacy of treatment between studies.60,61 A
meta-analysis of studies with different inclusion criteria and
different definitions of symptom relief may produce biases. As
was indicated recently, the quality of trials included in a meta-
analysis may have an impact on the efficacy estimates of treat-
ment in functional dyspepsia.54 Third, different treatment du-
rations were used in the included studies ranging from 2 to 8
weeks, and no follow-up data were reported after the treatment
was stopped. Therefore, it was not clear whether symptoms
would recur once the treatment was stopped. Because func-
tional dyspepsia is a chronic relapsing disorder, future studies
are needed to evaluate if the benefit of PPIs can last after
treatment is stopped, or if intermittent or on-demand use of
PPIs could be recommended.
In conclusion, by using the technique of meta-analysis, we
found that PPIs were significantly more effective for controlling
dyspeptic symptoms than placebo, especially in patients with
ulcer-like and reflux-like dyspepsia. The estimated NNT for
treatment of functional dyspepsia overall is approximately 15.
The role of H pylori on the efficacy of PPIs in treating functional
dyspepsia remains to be determined.
References
1. Talley NJ, Stanghellini V, Heading RC, et al. Functional gastrodu-
odenal disorders. Gut 1999;45(Suppl 2):ii37–ii42.
2. Moayyedi P, Mason J. Clinical and economic consequences of
dyspepsia in the community. Gut 2002;50(Suppl 4):iv10 –iv12.
3. Talley NJ, Weaver AL, Zinsmeister AR. Impact of functional dys-
pepsia on quality of life. Dig Dis Sci 1995;40:584 –589.
4. Talley NJ, Meineche-Schmidt V, Pare P, et al. Efficacy of omepra-
zole in functional dyspepsia: double-blind, randomized, placebo-
controlled trials (the Bond and Opera studies). Aliment Pharmacol
Ther 1998;12:1055–1065.
5. Bolling-Sternevald E, Lauritsen K, Aalykke C, et al. Effect of
184 WANG ET AL
profound acid suppression in functional dyspepsia: a
double-blind, randomized, placebo-controlled trial. Scand J Gas-
troenterol 2002;37:1395–1402.
6. Blum AL, Arnold R, Stolte M, et al. Short course acid suppressive
treatment for patients with functional dyspepsia: results depend
on Helicobacter pylori status. The Frosch Study Group. Gut 2000;
47:473– 480.
7. Finney JS, Kinnersley N, Hughes M, et al. Meta-analysis of anti-
secretory and gastrokinetic compounds in functional dyspepsia.
J Clin Gastroenterol 1998;26:312–320.
8. Allescher HD, Böckenhoff A, Knapp G, et al. Treatment of non-
ulcer dyspepsia: a meta-analysis of placebo-controlled prospec-
tive studies. Scand J Gastroenterol 2001;36:934 –941.
9. Soo S, Moayyedi P, Deeks J, et al. Pharmacological interventions
for non-ulcer dyspepsia. Cochrane Database Syst Rev 2000;2:
CD001960.
10. Shiau JY, Shukla VK, Dube C. The efficacy of proton pump inhib-
itors in adults with functional dyspepsia. Technology report no.
22. Ottawa: Canadian Coordinating Office for Health Technol-
ogy Assessment (CCOHTA), 2002. Available at: http://www.
ccohta.ca. Accessed June 20, 2003.
11. Wong WM, Wong BC, Hung WK, et al. Double blind, randomised,
placebo controlled study of four weeks of lansoprazole for the
treatment of functional dyspepsia in Chinese patients. Gut 2002;
51:502–506.
12. Talley NJ, Zinsmeister AR, Schleck CD, et al. Dyspepsia and
dyspepsia subgroups: a population-based study. Gastroenterol-
ogy 1992;102:1259 –1268.
13. Grainger SL, Klass HJ, Rake MO, et al. Prevalence of dyspepsia:
the epidemiology of overlapping symptoms. Postgrad Med J
1994;70:154 –161.
14. Moayyedi P, Soo S, Deeks J, et al. Pharmacological interventions
for non-ulcer dyspepsia. Cochrane Database Syst Rev 2003;1:
CD001960.
15. Moayyedi P, Delaney BC, Vakil N, et al. The efficacy of proton
pump inhibitors in nonulcer dyspepsia: a systematic review and
economic analysis. Gastroenterology 2004;127:1329 –1337.
16. Farup PG, Hovde O, Torp R, et al. Patients with functional dys-
pepsia responding to omeprazole have a characteristic gastro-
oesophageal reflux pattern. Scand J Gastroenterol 1999;34:
575–579.
17. Gerson LB, Triadafilopoulos G. A prospective study of oesopha-
geal 24-h ambulatory pH monitoring in patients with functional
dyspepsia. Dig Liver Dis 2005;37:87–91.
18. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control
Clin Trials 1986;7:177–188.
19. Egger M, Davey Smith G, Schneider M, et al. Bias in meta-
analysis detected by a simple, graphical test. BMJ 1997;315:
629 – 634.
20. Peura DA, Kovacs TOG, Metz DC, et al. Lansoprazole in the
treatment of functional dyspepsia: two double-blind, randomized,
placebo-controlled trials. Am J Med 2004;116:740 –748.
21. Hengels KJ. Therapeutic efficacy of 15 mg lansoprazole mane in
269 patients suffering from non-ulcer dyspepsia (NUD): a multi-
centre, randomized, double-blind study. Gut 1998;43:A89.
22. Laheij RJ, Severens JL, Van de Lisdonk EH, et al. Randomized
controlled trial of omeprazole or endoscopy in patients with per-
sistent dyspepsia: a cost-effectiveness analysis. Aliment Phar-
macol Ther 1998;12:1249 –1256.
23. Lewin-van-den-Broek NT, Numans ME, Buskens E, et al. A ran-
domised controlled trial of four management strategies for dys-
pepsia: relationships between symptom subgroups and strategy
outcome. Br J Gen Pract 2001;51:619 – 624.
24. Meineche-Schmidt V, Christensen E. Which dyspepsia patients
will benefit from omeprazole treatment? Analysis of a Danish
multicenter trial. Am J Gastroenterol 2000;95:2777–2783.
25. Rabeneck L, Souchek J, Wristers K, et al. A double blind, ran-
CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 2
domized, placebo-controlled trial of proton pump inhibitor therapy
in patients with uninvestigated dyspepsia. Am J Gastroenterol
2002;97:3045–3051.
26. Meineche-Schmidt V. The symptom relieving effect of omeprazole
40 mg daily is not significantly better than 20 mg in patients with
acid related disease from general practice. Gastroenterology
2001;120(Suppl 1):A730.
27. Meineche-Schmidt V. Classification of dyspepsia and response
to treatment with proton-pump inhibitors. Aliment Pharmacol Ther
2004;20:1171–1179.
28. Meineche-Schmidt V. Empiric treatment with high and standard
dose of omeprazole in general practice: two-week randomized
placebo-controlled trial and 12-month follow-up of health-care
consumption. Am J Gastroenterol 2004;99:1050 –1058.
29. Veldhuyzen van Zanten SJ, Chiba N, Armstrong D, et al. A ran-
domized trial comparing omeprazole, ranitidine, cisapride, or
placebo in helicobacter pylori negative, primary care patients with
dyspepsia: the CADET-HN Study. Am J Gastroenterol 2005;100:
1477–1488.
30. Bytzer P, Hansen JM, Rune S, et al. Identifying responders to acid
suppression in dyspepsia using a random starting day trial. Ali-
ment Pharmacol Ther 2000;14:1485–1494.
31. Goves J, Oldring JK, Kerr D, et al. First line treatment with
omeprazole provides an effective and superior alternative strat-
egy in the management of dyspepsia compared to antacid/algi-
nate liquid: a multicentre study in general practice. Aliment Phar-
macol Ther 1998;12:147–157.
32. Jones R, Crouch SL. Low-dose lansoprazole provides greater
relief of heartburn and epigastric pain than low-dose omeprazole
in patients with acid-related dyspepsia. Aliment Pharmacol Ther
1999;13:413– 419.
33. Jones RH, Baxter G. Lansoprazole 30 mg daily versus ranitidine
150 mg b.d. in the treatment of acid-related dyspepsia in general
practice. Aliment Pharmacol Ther 1997;11:541–546.
34. Mason I, Millar LJ, Sheikh RR, et al. The management of acid-
related dyspepsia in general practice: a comparison of an ome-
prazole versus an antacid-alginate/ranitidine management strat-
egy. Compete Research Group [corrected]. Aliment Pharmacol
Ther 1998;12:263–271.
35. Madsen LG, Bytzer P. Reproducibility of a symptom response to
omeprazole therapy in functional dyspepsia evaluated by a ran-
dom-starting-day trial design. Aliment Pharmacol Ther 2004;20:
365–372.
36. Madsen LG, Wallin L, Bytzer P. Identifying response to acid
suppressive therapy in functional dyspepsia using a random
starting day trial—is gastro-oesophageal reflux important? Ali-
ment Pharmacol Ther 2004;20:423– 430.
37. Meineche-Schmidt V, Talley NJ, Pap A, et al. Impact of functional
dyspepsia on quality of life and health care consumption after
cessation of antisecretory treatment. A multicentre 3-month fol-
low-up study. Scand J Gastroenterol 1999;34:566 –574.
38. Talley N, Vakil N, Lauritsen K, et al. An acid suppression trial with
esomeprazole identifies patients with non-ulcer dyspepsia who
will respond to prolonged acid suppression: a randomized con-
trolled trial of 1589 endoscoped patients. R.0066, World Con-
gress of Gastroenterology, Sep 10 –14, 2005, Montreal, Canada.
39. Singh A, Newton W. Which patients with ulcer- or reflux-like dys-
pepsia will respond favorably to omeprazole? J Fam Pract 2001;
50:365.
40. Talley NJ, Lauritsen K. The potential role of acid suppression in
functional dyspepsia: the BOND, OPERA, PILOT, and ENCORE
studies. Gut 2002;50(Suppl 4):iv36 –iv41.
41. Lauritsen K, Aslykke C, Havelund T, et al. Effect of omeprazole in
functional dyspepsia: a double-blind, randomized, placebo-con-
trolled study. Gastroenterology 1996;110(Suppl 4):A702.
42. Peura DA, Kovacs TO, Metz D, et al. Low-dose lansoprazole:
February 2007
effective for non-ulcer dyspepsia (NUD). Gastroenterology
2000;118(Suppl 2):A439.
43. Pilmer L, Kovacs TOG, Metz DC, et al. Low-dose lansoprazole:
effective for non-ulcer dyspepsia (NUD). Gut 2001;49(suppl 3):
A1873.
44. Bolling-Sternevald E, Lauritsen K, Talley NJ, et al. Is it possible to
predict treatment response to a proton pump inhibitor in func-
tional dyspepsia? Aliment Pharmacol Ther 2003;18:117–124.
45. Goh KL, Chang CS, Fock KM, et al. Gastro-oesophageal reflux
disease in Asia. J Gastroenterol Hapatol 2000;15:230 –238.
46. Quigley EM. Non-erosive reflux disease: part of the spectrum of
gastro-oesophageal reflux disease, a component of functional
dyspepsia, or both? Eur J Gastroenterol Hepatol 2001;13(Suppl
1):S13–S18.
47. Mantinez SD, Malagon ZB, Garewal HS, et al. Non-erosive reflux
disease (NERD)—acid reflux and symptom patterns. Aliment
Pharmacol Ther 2003;17:537–545.
48. Richter JE, Campbell DR, Kahrilas PJ, et al. Lansoprazole com-
pared with ranitidine for the treatment of nonerosive gastroesoph-
ageal reflux disease. Arch Intern Med 2000;160:1803–1809.
49. van P, Numans ME, Bonis PA, et al. Short-term treatment with
proton pump inhibitors, H2-receptor antagonists and prokinetics
for gastro-oesophageal reflux disease-like symptoms and endos-
copy negative reflux disease. Cochrane Database Syst Rev 2001;
4:CD002095.
50. Armstrong D, Pare P, Pericak D, et al. Symptom relief in gastro-
esophageal reflux disease: a randomized, controlled comparison
of pantoprazole and nizatidine in a mixed patient population with
erosive esophagitis or endoscopy-negative reflux disease. Am J
Gastroenterol 2001;96:2849 –2857.
51. Small PK, Loudon MA, Waldron B, et al. Importance of reflux
symptoms in functional dyspepsia. Gut 1995;36:189 –192.
52. Laine L, Schoenfeld P, Fennerty MB. Therapy for Helicobacter
pylori in patients with nonulcer dyspepsia. A meta-analysis of
randomized, controlled trials. Ann Intern Med 2001;134:361–
369.
53. Moayyedi P, Soo S, Deeks J, et al. Eradication of Helicobacter
pylori for non-ulcer dyspepsia. Cochrane Database Syst Rev
2001;1:CD002096.
54. Abraham NS, Moayyedi P, Daniels B, et al. Systematic review: the
methodological quality of trials affects estimates of treatment
efficacy in functional (non-ulcer) dyspepsia. Aliment Pharmacol
Ther 2004;19:631– 641.
PPIS ON FUNCTIONAL DYSPEPSIA 185
55. El-Omar E, Penman I, Ardill JES. A substantial proportion of
non-ulcer dyspepsia patients have the same abnormality of acid
secretion as duodenal ulcer patients. Gut 1995;36:534 –538.
56. Treiber G, Schwabe M, Ammon S, et al. Dyspeptic symptoms
associated with Helicobacter pylori infection are influenced by
strain and host specific factors. Aliment Pharmacol Ther 2004;
19:219 –231.
57. Deeks J, Higgins J, Altman D on behalf of the Cochrane Statistical
Methods Group. Analysing and presenting results. In: Higgins
JPT, Green S, eds. Cochrane Handbook for Systematic Reviews
of Interventions 4.2.4. [updated March 2005]. Chichester, UK:
John Wiley & Sons, Ltd.
58. Egger M, Davey Smith G. Meta-analysis bias in location and
selection of studies. BMJ 1998;316:61– 66.
59. Davey Smith G, Egger M. Meta-analysis: unresolved issues and
future developments. BMJ 1998;316:221–225.
60. Veldhuyzen van Zanten SJO, Cleary C, Talley NJ, et al. Drug
treatment of functional dyspepsia: a systematic analysis of trials
methodology with recommendations for design of future trials.
Am J Gastroenterol 1996;91:660 – 673.
61. Westbrook JI, McIntosh JH, Talley NJ. The impact of dyspepsia
definition on prevalence estimates: considerations for future re-
searchers. Scand J Gastroenterol 2000;35:227–233.
Address requests for reprints to: Benjamin C. Y. Wong, Department
of Medicine, University of Hong Kong, Hong Kong, China. e-mail:
bcywong@hku.hk; fax: (852) 2872-5828.
Supported by the Gastroenterological Research Fund, University of
Hong Kong, Hong Kong. W.H.W. is the recipient of a Visiting Profes-
sorship under the Croucher Foundation Chinese Visitorship Scheme to
the University of Hong Kong, Hong Kong.
The authors would like to thank Dr L. B. Gerson (VA Palo Alto Health
Care System, Palo Alto, California), Dr D. A. Peura (University of
Virginia, Charlottesville Virginia), Dr N. J. Talley (Centre for Enteric
Neurosciences and Translational Epidemiological Research, Division
of Gastroenterology and Hepatology, Mayo Clinic, Rochester, Minne-
sota), and Dr E. Bolling-Sternevald (Family Medicine Stockholm, Karo-
linska Institute, Sweden; and AstraZeneca Clinical R&D, Mälndal, Swe-
den) for their generosity in providing us with the raw data of their
studies, and Dr P. G. Farup (Department of Medicine, Gjøvik Country
Hospital, Gjøvik, Norway) for providing other information from their
studies.