Systematic review: antacids, H2-receptor antagonists, prokinetics, bismuth and sucralfate therapy for non-ulcer dyspepsia

Authors


Professor P. Moayyedi, Gastroenterology Unit, City Hospital NHS Trust, Dudley Road, Winson Green, Birmingham B18 7QH, UK.
E-mail: p.moayyedi@bham.ac.uk

Summary

Background: Evidence for the effectiveness of antacids, histamine-2 receptor antagonists, bismuth salts, sucralfate and prokinetic therapy in non-ulcer dyspepsia is conflicting.

Aim: To conduct a systematic review evaluating these therapies in non-ulcer dyspepsia.

Methods: Electronic searches were performed using the Cochrane Controlled Trials Register, Medline, EMBASE, Cinahl and SIGLE until September 2002. Dyspepsia outcomes were dichotomized into cured/improved vs. same/worse.

Results: Prokinetics [14 trials, 1053 patients; relative risk reduction (RRR), 48%; 95% confidence interval (95% CI), 27–63%] and histamine-2 receptor antagonists (11 trials, 2164 patients; RRR, 22%; 95% CI, 7–35%) were significantly more effective than placebo. Bismuth salts (RRR, 40%; 95% CI, − 3% to 65%) were superior to placebo, but this was of marginal statistical significance. Antacids and sucralfate were not statistically significantly superior to placebo. A funnel plot suggested that the prokinetic and histamine-2 receptor antagonist results could be due to publication bias.

Conclusions: The meta-analyses suggest that histamine-2 receptor antagonists and prokinetics are superior to placebo. These data are difficult to interpret, however, as funnel plot asymmetry suggests that the magnitude of the effect could be due to publication bias or other heterogeneity-related issues.

Introduction

Patients with upper gastrointestinal symptoms are often referred for investigation, with 1% of the population having an endoscopy each year.1 This may reveal oesophagitis or a peptic ulcer for which there are effective treatments.2 Endoscopy is essentially normal in over 50% of cases, however, and these patients are often labelled as having functional or non-ulcer dyspepsia.3 This is the most common diagnosis for patients with dyspepsia in primary care, yet the treatment for this condition is uncertain. Proton pump inhibitor therapy may be effective in a small subset of patients.4 These drugs are expensive, however, and, given their modest efficacy, a cheaper drug is desirable. Antacids, histamine-2 receptor antagonists (H2RAs), prokinetics, bismuth and sucralfate are usually less expensive than proton pump inhibitor therapy, but randomized controlled trials have given conflicting results.5, 6 Systematic reviews have suggested that prokinetic drugs may be the most effective therapy for non-ulcer dyspepsia,7, 8 although the methodology of these reviews has been criticized.9, 10

Patients with a normal endoscopy can have gastro-oesophageal reflux disease, particularly if heartburn or acid reflux is the predominant symptom.11 Non-ulcer dyspepsia trials often include this patient group and the efficacy of any therapy may relate to the treatment of gastro-oesophageal reflux disease rather than non-ulcer dyspepsia. We have conducted a systematic review evaluating the efficacy of antacids, H2RAs, prokinetic drugs and mucosal protecting agents in the treatment of non-ulcer dyspepsia. We have also evaluated the outcome according to each upper gastrointestinal symptom to evaluate whether any treatment effect is due to the impact on gastro-oesophageal reflux disease rather than non-ulcer dyspepsia.

Methods

Search strategy

Parallel-group randomized controlled trials and the first period of cross-over randomized controlled trials fulfilling the inclusion criteria were eligible for the review. Trials were identified from the Cochrane Controlled Trials Register (September 2002), MEDLINE (1966 to September 2002), EMBASE (1988 to September 2002), CINAHL (1982 to September 2002) and the grey literature using SIGLE. Experts in the field of dyspepsia in 15 countries and pharmaceutical companies were contacted for any relevant unpublished materials. A recursive search of the bibliography of relevant papers was also conducted. General medical and major gastroenterology journals were routinely scanned over the previous year to ensure that the most recent studies were included. Papers were considered regardless of language and publication status, although abstracts were only included when further details were available from the authors. Details of the search strategy for this review are published in the Cochrane Library.12

Assessment of eligibility and trial quality

All randomized controlled trials evaluating adult patients with dyspepsia that fulfilled published definitions13, 14 were eligible to be included in the systematic review. Patients in the trials had to have negative findings at endoscopy or barium study. Findings of hiatus hernia, less than five gastric erosions or mild duodenitis were permitted. Studies with outcome assessment carried out in less than 1 week were excluded, as were trials that evaluated patients who were predominantly (> 20%) taking non-steroidal anti-inflammatory drugs. Trials evaluating patients with gastro-oesophageal reflux disease symptoms only were also excluded. Two investigators independently reviewed all identified papers according to these eligibility criteria. A third reviewer was involved when disagreements arose and a consensus view was taken.

The quality of the trials was also evaluated according to pre-defined criteria. The quality assessment focused on whether the methods used for randomization, concealment of allocation and masking of participants and investigators were stated. The use of intention-to-treat analyses and the completeness of follow-up were also recorded, and the use of validated dyspepsia questionnaires was noted. Trials described as randomized, but which did not state the method of randomization, were included.

Data extraction

A single investigator extracted data from eligible trials onto a standardized form, which was checked by a second investigator. Data from intention-to-treat analyses were used when provided, and outcomes were recorded for the final visit. Dyspepsia outcomes were recorded in categories which were dichotomized a priori to improved/resolved vs. same/worse dyspepsia symptoms. Thus, trials that included patients with at least moderate dyspepsia at entry were dichotomized into improved/resolved if patients had mild or no symptoms at the final visit and same/worse if patients had moderate or severe symptoms at the final visit. Standard mean differences were recorded for continuous dyspepsia scores if the data could not be dichotomized.

Data synthesis

The effect of a particular therapy in each trial was expressed as a relative risk, comparing the numbers remaining dyspeptic in the group allocated to that therapy compared with the group receiving alternative therapy or placebo. Trials comparing more than one dose of therapy were analysed by calculating the relative risk compared with alternative therapy/placebo for each dose separately. Relative risks were pooled using a fixed effect (Mantel–Haenszel) model, the appropriateness of which was assessed using a test of homogeneity. If significant heterogeneity (P < 0.2) existed between the studies, the relative risks were pooled using a random effects model calculated according to the method of Der Simonian and Laird. Reasons for heterogeneity in the results were explored and the following factors were considered a priori: (i) trials specifically excluding patients with reflux-predominant symptoms; (ii) the length of follow-up; (iii) the method used to exclude organic upper gastrointestinal disease; (iv) the use of validated dyspepsia questionnaires; (v) multi-centre vs. single-centre trials; (vi) dosage, drug name and duration of treatment.

Egger's test for funnel plot asymmetry was investigated to assess publication bias, quality-related problems or heterogeneity.

The outcome was re-expressed in terms of a relative risk reduction (RRR) [RRR(%) = 100(1 − RR)]. The number needed to treat (NNT) was calculated using the mean dyspepsia rate in the placebo group (BR) and the RRR according to the following formula:

image

All results were reported with 95% confidence intervals (95% CI). Statistical analyses were performed using STATA (version 6.0). The review was undertaken according to a protocol published in the Cochrane Library, and will be regularly updated as a Cochrane Review as more information becomes available.

Results

A total of 11 796 citations was obtained. One hundred and forty-eight trials were retrieved and 89 trials fulfilled our eligibility criteria. Data extraction was not possible in 45 trials and 44 papers were included in the systematic review (Table 1).5, 6, 15–56 Reasons for exclusion of the studies are available from the corresponding author. Each paper could contribute more than one trial to the review as some studies evaluated more than two interventions.

Table 1.  Summary of included trials
ReferenceParticipantsInterventionOutcome
  1. CBS, colloidal bismuth subcitrate; GERD, gastro-oesophageal reflux disease; IBS, irritable bowel syndrome; NSAID, non-steroidal anti-inflammatory drug; NUD, non-ulcer dyspepsia; OGD, oesophago-gastroduodenoscopy; RCT, randomized controlled trial.

Agorastos et al.26Greece. RCT. Single-blind trial. 36 patients with NUD, 17 on cisapride, 19 on placebo. 2 weeks placebo run-in period4 weeks. Cisapride 5 mg t.d.s. vs. placeboThe symptomatic response was better for cisapride
Al-Quorain et al.17Saudi Arabia. RCT. Double-blind placebo-controlled trial. 89 patients, 44 on cisapride, 45 on placebo. 3 sub-groups: ulcer-like, reflux-like, dysmotility-like. 2-week placebo run-in period. 91% completed trial4 weeks. Cisapride 5 mg t.d.s. vs. placeboCisapride was significantly superior to placebo in improving heartburn, post-prandial bloating, epigastric pain, early satiety, epigastric burning and nausea
Bekhti and Rutgeerts18Belgium. RCT. Double-blind placebo-controlled trial. 40 patients, 20 in each arm. Chronic dyspepsia and delayed gastric emptying tests. Radiological examination only. 15% radiological reflux. No dropouts4 weeks. Domperidone 10 mg t.d.s. vs. placeboGlobal evaluation was significantly in favour of domperidone. Few side-effects
Blum et al.36Switzerland. RCT. Double-blind placebo controlled trial. 792 patients. Normal endoscopy and no response to 1-week course of low-dose antacid2 weeks. Omeprazole 20 mg od vs. omeprazole 10 mg od vs. ranitidine 150 mg od vs. placeboOmeprazole more effective than ranitidine which, in turn, was more effective than placebo. Authors noted that differences were most marked in H. pylori-positive patients, with no statistically significant differences between interventions in the H. pylori-negative group
Carvalhinhos et al.47Belgium. RCT. Double-blind placebo-controlled trial. 203 patients, 99 on cisapride, 104 on ranitidine. 1 month dyspepsia. 93% completed trial8 weeks. Cisapride 30 mg daily vs. ranitidine 300 mg dailyBoth cisapride and ranitidine improved the individual symptoms of NUD. Cisapride was superior to ranitidine particularly on the combined evaluation of the response to treatment and the recurrence of symptoms
Casiraghi et al.15Italy. RCT. Double-blind placebo-controlled trial. 125 patients, 52 in each arm. 1 month predominantly epigastric pain or heartburn as primary symptom and a list of other secondary symptoms. Had OGD. 83% completed trial2 weeks. Cimetidine 200 mg q.d.s. vs. antacid 1 tablet q.d.s.Cimetidine was significantly more effective than antacids in reducing the number of pain or heartburn episodes. Confirms the efficacy of cimetidine in providing symptomatic relief in patients affected with NUD, particularly when pain is the major complaint
Champion et al.5Canada. RCT. Double-blind placebo-controlled trial. 123 patients with NUD, 42 on cisapride 10 mg, 41 on cisapride 20 mg, 40 on placebo. 2-week placebo run-in period. Had OGD. 78% completed trial6 weeks. 2 different doses of cisapride at 10 mg t.d.s. vs. 20 mg t.d.s. vs. placeboCisapride at both doses was not effective compared with placebo in improving symptoms in NUD patients. Side-effects profile comparable to that of placebo
Chung6Korea. RCT. Double-blind placebo-controlled trial. 29 patients with chronic dyspepsia, 14 on cisapride, 15 on placebo. 97% completed trial4 weeks. Cisapride 10 mg t.d.s. vs. placeboBloating and epigastric discomfort were significantly reduced compared with placebo. Good or excellent global response in 71.4%. No significant side-effects noted
Creytens51Belgium. RCT. Double-blind placebo-controlled cross-over trial. 32 patients with chronic dyspepsia, 16 in each arm. Included reflux symptoms. 2 week drug-free period. Had either OGD or barium studies. No dropouts3 weeks. Cisapride 4–8 mg t.d.s. (doubling the dose after 10 days if needed) vs. placeboCisapride significantly superior to placebo. Markedly improved nausea. Good response for epigastric discomfort, reflux symptoms. No adverse events noted
De Groot et al.25Dutch. RCT. Double-blind placebo-controlled trial. 121 patients, 61 on cisapride, 60 on placebo. Upper abdominal pain for at least 4 weeks. Normal endoscopy4 weeks. Cisapride 10 mg t.d.s. vs. placeboSuccess defined as patient rating treatment good or excellent on 4-point likert scale. Paper also stratified by H. pylori status — no difference
Delattre et al.28USA. RCT. Double-blind placebo-controlled trial. 414 patients, 63 on treatment, 62 on placebo. 3 months of predominantly epigastric pain. Gastroduodenitis in 22% of patients4 weeks. Cimetidine 200 mg q.d.s. vs. placeboThe benefit of cimetidine in providing symptomatic relief was significantly higher in patients who previously responded favourably to antacids
De Nutte et al.19Belgium. RCT. Double-blind placebo-controlled trial. 32 predominantly epigastric pain. 17 on cisapride and 15 on placebo. No dropouts. NSAIDs withdrawn at start of treatment in 4 patients. 2-week placebo run-in period. No dropouts4 weeks. Cisapride 5 mg t.d.s. vs. placeboCisapride was effective in reducing epigastric pain.
Deruyttere et al.53Belgium. RCT. Double-blind placebo-controlled trial. 56 patients with chronic functional dyspepsia. 26 on cisapride, 30 on placebo. 93% completed trial3 weeks. Cisapride 4 mg t.d.s. vs. placeboCisapride was particularly superior to placebo in the improvement of a cluster of symptoms typical of post-prandial discomfort, including early satiety and nausea. Side-effects were minimal
Francois and De Nutte20Belgium. RCT. Double-blind placebo-controlled trial. 36 with dyspepsia (3/12). 18 in each arm. 64% had either gastritis and/or bulbitis. 2 weeks drug withdrawal. 94.4% completed trial3 weeks. Cisapride 5 mg t.d.s. vs. placeboCisapride was significantly superior to placebo in relieving epigastric burning or pain, heartburn, regurgitation and abdominal distension
Fumagalli et al.48Switzerland. RCT. Double-blind placebo-controlled trial. 60 patients, 30 in each arm. 1 month dyspepsia, included reflux symptoms. 2 sub-groups: reflux-like, typical dyspepsia. Had OGD. 95% completed trial. 2-week follow-up4 weeks. Cisapride 10 mg t.d.s. vs. metoclopramide 10 mg t.d.s.Significant improvement in both groups (no significant inter-group difference). At 2-week follow-up, the response rate was significantly better with cisapride (73%) than with metoclopramide (47%)
Goh et al.43Malaysia. RCT. Double-blind placebo-controlled trial. 71 NUD ± H. pylori infection. 40 patients were H. pylori-positive, 21 on treatment and 19 on placebo. 31 were H. pylori-negative,17 on treatment and 14 on placebo. Non-erosive duodenitis included. 84.5% completed trial.4 weeks. CBS 2 tablets b.d. vs. placeboAll groups reported improvement in symptom scores. In the H. pylori-positive group, CBS-treated group recorded a significantly higher improvement than the other groups. 12 of 16 patients relapse 1 month after withdrawal of CBS
Gotthard et al.29Sweden. RCT. Double-blind placebo-controlled trial. 210 patients, 73 on cimetidine, 74 on antacid, 75 on placebo. 3/12 with dyspepsia of unknown origin. Acid output studies performed. 16% duodenitis6-week comparisons of 3 drugs. Cimetidine 400 mg b.d. vs. placebo vs. antacid 10 mL q.d.s.Cimetidine was superior to both placebo and antacid in relieving pain and nausea but not bloating
Gudjonsson et al.45Iceland. RCT. Double-blind placebo-controlled trial. Private practice recruitment. 104 patients, 56 on cisapride, 48 on placebo. 1 month of dyspepsia. Had OGD. 91% completed trial.3 weeks. Sucralfate 1 g q.d.s. vs. placeboNo significant difference between the individual symptom and global symptom scores between sucralfate and placebo
Hadi30Indonesia. RCT. Double-blind placebo-controlled trial. 52 total, 26 on ranitidine, 26 on placebo. Duration of dyspepsia unclear. Gastritis on all OGD. Dropout rate for placebo was 23% and for ranitidine was 4%4 weeks. Ranitidine 300 mg daily vs. placeboRanitidine was effective as a short-term treatment for patients with endoscopically proven gastritis
Halter et al.49Switzerland. RCT. Double-blind placebo-controlled trial. 127 patients, 68 on cisapride, 59 on domperidone. 1 month dyspepsia. 2 sub-groups: reflux group, dyspepsia group. Had OGD. 2-month follow-up. 88.9% completed trial.4 weeks. Cisapride 10 mg q.d.s. vs. domperidone 20 mg q.d.s.Cisapride was more effective than domperidone in the reflux group. Domperidone was more effective against nausea in the reflux group
Hannon21Belgium. RCT. Double-blind placebo-controlled cross-over trial. 22 patients with NUD, 11 patients in each arm. 2-week wash-out period. No dropouts3 weeks. Cisapride 5 mg t.d.s. vs. placeboCisapride was superior to placebo in relieving a cluster of dyspeptic symptoms, in particular epigastric burning and early satiety. Global therapeutic effect was good or excellent in treatment group (64%) compared with the control group (27%)
Hansen et al.22Denmark. Primary care recruitment. RCT. Double-blind placebo-controlled trial. 330 patients, 109 on cisapride, 111 on nizatidine, 110 on placebo. Mean duration of dyspeptic symptoms was 88 months. 4 sub-groups: ulcer-like (13%), reflux-like (23%), dysmotility-like (46%) and unclassified (18%). Included superficial erosions on OGD. 85% completed trial2 weeks. Cisapride 10 mg t.d.s. vs. nizatidine 300 mg nocte vs. placeboThe effects of a 2-week course of cisapride or nizatidine in patients recruited from primary care were not superior to those of placebo. Symptom sub-grouping was not predictive of response to treatment
Holtmann et al.55Germany. RCT. Double-blind placebo controlled trial. 185 private patients with normal endoscopy. Rome definition of dyspepsia — predominant reflux symptoms excluded8 weeks. Cisapride 10 mg t.d.s. or simethicone 105 mg t.d.s. vs. placeboSymptom scores were primary outcome, but standard deviations of the unadjusted scores were not given. Proportion of patients judging treatment to be very good was also given and this was used in the meta-analysis. No difference between cisapride and placebo for this measure, although there was a difference in mean scores in favour of cisapride
Jian et al.27France. RCT. Double-blind placebo-controlled trial. 28 chronic idiopathic dyspepsia suggestive of gastroparesis. 15 on cisapride, 13 on placebo. Had gastric emptying test (59% positive) and OGD. 82% completed trial6 weeks. Cisapride 10 mg t.d.s. vs. placeboCisapride significantly improved gastric emptying for liquids and individual dyspeptic scores. Best symptomatic response to cisapride was found in patients with gastroparesis
Kairaluoma et al.46Finland. RCT. Double-blind placebo-controlled trial. 151 patients, 79 on sucralfate, 72 on placebo. 3 months of dyspepsia. Had OGD. 6% had duodenitis. 86% completed trial4 weeks. Sucralfate 1 g t.d.s. vs. placeboSignificant difference between sucralfate and placebo in the global response. Best response achieved in patients with mild or moderate symptoms
Kang et al.38Singapore. RCT. Double-blind placebo-controlled trial. 73 total, 21 on treatment, 19 on placebo. Dyspepsia with food-related abdominal pain and H. pylori gastritis. 70% completed trial8 weeks. CBS 1 tablet q.d.s. vs. placeboCBS benefited those with gastritis but not those without
Kazi et al.39India. RCT. Double blind — not stated. 52 patients, 26 in each arm. Dyspepsia and H. pylori gastritis. No dropouts3 weeks. Bismuth salicylate 500 mg t.d.s. vs. placeboResolution of gastritis and improvement of symptoms were significantly better in patients in whom H. pylori was eradicated when compared with those in whom H. pylori persisted
Kelbaek et al.31Denmark. RCT. Double-blind placebo-controlled trial. 52 patients, 24 on cimetidine, 26 on placebo. 1 month of epigastric pain. Acid output studies performed. Had OGD. 14 patients who were symptom free at end of treatment had 3 months' follow-up. 96% completed trial3 weeks. Cimetidine 200 mg t.d.s. and 400 mg nocte vs. placeboCimetidine does not seem to be superior to placebo in NUD
Kellow et al.23Australia. RCT. Double-blind placebo-controlled trial. 61 patients with NUD, 30 on cisapride, 31 on placebo. 2-week placebo run-in period. 2 sub-groups: reflux-like and dysmotility. Gastric emptying tests performed. 91.8% completed trial4 weeks. Cisapride 10 mg t.d.s. vs. placeboMajor differences in the short-term efficacy of cisapride and placebo. Indications of beneficial effects of cisapride over placebo in those with reflux-like dyspepsia, and in those without gastroparesis
Kumar et al.44India. RCT. Double-blind placebo-controlled trial. 81 patients, 18 on CBS, 15 on placebo I, 15 on sucralfate, 15 on placebo II. NUD and H. pylori infection. 78% completed trial4 weeks. CBS 240 mg b.d. vs. placebo I vs. sucralfate 2 g b.d. vs. placebo IICBS is more effective than sucralfate in inducing endoscopic and histological healing of H. pylori-related gastritis amongst NUD patients
Lambert et al.40Australia. RCT. Double-blind placebo-controlled trial. 82 patients with NUD, some with C. pylori infection. 48 C. pylori-positive: 22 on treatment, 26 on placebo. 30 C. pylori-negative: 16 on treatment, 14 on placebo. 95% completed trial4 weeks. Bismuth subcitrate 4 tablets daily vs. placeboClearance of Campylobacter pylori and histological improvement was associated with significant decrease in symptoms. In C. pylori-positive patients, improvement occurred in both placebo and treatment groups
Loffeld et al.41The Netherlands. RCT. Double-blind placebo-controlled trial. ;57 patients, 26 on treatment, 24 on placebo. NUD with C. pylori gastritis. 88% completed trial.4 weeks. CBS 240 mg daily vs. placeboThere was improvement in the subjective complaints in both groups, except for nausea and meteorism that improved more in the CBS group. Significant reduction in C. pylori colonization and gastritis score
Misra et al.16India. Randomized, open, controlled clinical trial. 100 patients, 47 on ranitidine, 53 on sucralfate. 1 month of abdominal symptoms referable to the upper gastrointestinal tract. GERD, IBS were excluded. 87% completed trial4 weeks. Ranitidine 150 mg b.d. vs. sucralfate 1 g q.d.s.Global relief in symptoms was significantly more frequent in the sucralfate group than in the ranitidine group. Sucralfate is superior to ranitidine
Muller et al.35Germany. RCT. Randomized double-blind controlled trial. 652 patients with normal endoscopy. All upper gastrointestinal symptoms included in definition4 weeks. Ranitidine 150 mg b.d. vs. placeboComplete relief of global symptoms
Nesland and Berstad32Norway. RCT. Double-blind placebo-controlled trial. 100 patients, 44 on cimetidine. 6 months of predominantly ulcer-like pain and with erosive pre-pyloric changes. 46 on placebo. Acid output studies performed. 90% completed trial4 weeks. Cimetidine 400 mg b.d. vs. placeboPatients with NUD and erosive pre-pyloric changes who have epigastric pain/discomfort as a prominent symptom seem to profit from treatment with cimetidine
Olubuyide and Atoba37Nigeria. RCT. Double-blind placebo controlled trial. Recruited duodenal ulcer and NUD patients, but the two groups were analysed separately. 45 NUD patients — all upper astrointestinal symptoms included4 weeks. Ranitidine 300 mg od vs. placebo1. Acid output 2. Absence of dyspepsia symptoms 3. General Health Questionnaire (not reported)
Rosch24Germany. RCT. Double-blind placebo-controlled trial. 118 patients with NUD, 54 on cisapride, 55 on placebo. Duration of dyspepsia unclear. 2 sub-groups: reflux-like, dysmotility-like. Had OGD4 weeks. Cisapride 10 mg t.d.s. vs. placeboCisapride caused significant improvement compared with placebo with regard to the frequency and severity of symptoms
Saunders et al.33UK. Primary care recruitment. RCT. Double-blind placebo-controlled multi-centre trial. 251 patients with NUD, 115 on ranitidine, 136 on placebo. 88% completed trial. One-year follow-up, but the results included other peptic disease6 weeks Ranitidine 150 mg b.d. vs. placeboSignificantly more NUD patients became symptom free with ranitidine compared with placebo
Singal et al.34India. RCT. Double-blind placebo-controlled trial. 67 patients, 33 on cimetidine, 34 on placebo. 1 month of primary symptom of upper abdominal discomfort. IBS excluded4 weeks. Cimetidine 400 mg b.d. vs. placeboAbdominal pain and other secondary dyspeptic symptoms were relieved in higher proportions in the cimetidine-treated group, although the difference was not significant
Testoni et al.52Italy. RCT. Single-blind placebo-controlled trial. 20 patients, 10 in each arm. 6 months of dyspepsia and delayed antroduodenal motility on manometry. Had OGD. No dropouts15 days. Cisapride 10 mg q.d.s. vs. placeboSymptomatic improvement in both groups, hardly significant (P = 0.049). Cisapride improved inter-digestive antroduodenal motor activity
Vaira et al.42Italy. RCT. Double-blind placebo-controlled trial. 80 patients, 40 in each arm. H. pylori-associated NUD. Follow-up at 4 weeks post-treatment was 97.5%. H. pylori eradication rate was 54%4 weeks. CBS 240 mg b.d. vs. placeboCBS is effective treatment for H. pylori-associated NUD with improved gastric antral histological appearances, and has a beneficial effect on symptoms
Van Outryve et al.50Belgium. RCT, double-blind placebo-controlled trial. 147 patients, 77 on metoclopramide, 70 on domperidone. Had 2.8 months mean duration of chronic dyspepsia. Included reflux symptoms. Had OGD within 3 months before trial. 18% clinical diagnoses of GERDPhase I study. 2 weeks. Metoclopramide 10 mg t.d.s. vs. domperidone 10 mg t.d.s. Phase II study. Cisapride 10 mg t.d.s. vs. placeboMetoclopramide and domperidone produced comparable alleviation of epigastric symptoms. Global efficacy was good or excellent in 62% and 57% of patients, respectively. In refractory patients, cisapride produced significant efficacy compared with placebo
Wood et al.54UK. RCT. Double-blind placebo controlled trial. 11 patients with normal endoscopy. Predominant epigastric pain4 weeks. Cisapride 10 mg t.d.s. vs. placeboAbsence of epigastric pain (day and night)
Yeoh et al.56Singapore. RCT. Double-blind placebo-controlled trial. 104 patients with functional dyspepsia. 38 patients in each arm, consisting of one group with gastritis and one without. 2-week antacid run-in period. 73% completed trial4 weeks. Cisapride 10 mg t.d.s. vs. placebo.Cisapride produced a good or better global response in 58% with gastritis and 53% without gastritis compared with 47% and 52%, respectively, of patients on placebo. There was no significant difference between the groups

H2RA therapy

Comparison with placebo.  Eleven papers compared H2RAs with placebo, evaluating a total of 2164 patients.22, 28–37 Five trials evaluated cimetidine,28, 29, 31, 32, 34 five trials ranitidine30, 33, 35–37 and one trial nizatidine.22 There was statistically significant heterogeneity between the studies [test for heterogeneity chi-squared = 54.6 (d.f. = 10), P < 0.0001] and a random effects model was used. Overall improvement in dyspepsia as a dichotomous variable was available in all 11 studies, and this suggested a significant benefit of H2RAs over placebo (RRR, 22%; 95% CI, 7–35%) (Figure 1). Assuming a placebo response rate of 40%, eight (95% CI, 5–24) patients need to be treated with H2RAs to improve one case of dyspepsia that would not have been improved by placebo. Meta-regression evaluating the type of H2RA, dose of drug, duration of therapy, method of randomization, concealment of allocation and dropout rate did not find any factor that explained the heterogeneity between the studies. A funnel plot exhibited significant asymmetry with a slight preponderance of small trials showing a more marked treatment effect and larger trials showing smaller benefit over placebo (Figure 2).57 This suggests that the benefit of H2RA over placebo may be overestimated by this meta-analysis (Egger's test for funnel plot asymmetry:57 coefficient, − 1.56; 95% CI, − 1.16 to − 1.96).

Figure 1.

Meta-analysis of histamine-2 receptor antagonists (H2RAs) compared with placebo in non-ulcer dyspepsia patients. Random effects model. Heterogeneity chi-squared = 54.6 (degrees of freedom = 10), P < 0.0001.

Figure 2.

Funnel plot of histamine-2 receptor antagonists compared with placebo in non-ulcer dyspepsia patients. SE, standard error.

The quality of the trial reports was poor, with only three mentioning the method of randomization and concealment. Individual symptoms were recorded in a few trials. H2RAs had no statistically significant effect on heartburn (data from two trials evaluating 635 patients:21, 27 RRR, 8%; 95% CI, − 13% to 55%). Epigastric pain was improved by H2RAs (data from four trials evaluating 477 patients:22, 29, 31, 32 RRR, 18%; 95% CI, 0–32%), as was post-prandial fullness (data from two trials evaluating 635 patients:22, 28 RRR, 28%; 95% CI, 0–49%).

Comparison with other therapies.  Two trials assessing 423 patients compared H2RAs with prokinetic therapy.22, 47 There was statistically significant heterogeneity between the trials (Q = 7.9, d.f. = 1, P = 0.005), with one trial suggesting that prokinetic therapy was superior to H2RAs47 and the other suggesting each had a similar efficacy.22 Overall, there was no statistically significant difference between the two interventions (prokinetic therapy vs. H2RA: RRR, 46%; 95% CI, − 33% to 78%; random effects model).

One paper compared the efficacy of proton pump inhibitors with H2RAs in 783 patients.36 There was a trend for proton pump inhibitors to be more effective than H2RAs, but this did not reach statistical significance (RRR, 7%; 95% CI, − 2% to 14%; P = 0.13).

Antacids and H2RAs were compared in one trial evaluating 104 patients with non-ulcer dyspepsia.29 There was a trend in favour of antacids, but this did not reach statistical significance (RRR, 30%; 95% CI, − 23% to 80%). One open randomized controlled trial compared sucralfate with H2RA therapy, and suggested that patients allocated to antisecretory therapy were 2.7 times as likely to have moderate to severe dyspepsia (95% CI, 1.3–6.0).16

Prokinetic therapy

Comparison with placebo. Fourteen papers compared prokinetic therapy (all but one evaluating cisapride18) with placebo and reported a dichotomous global dyspepsia outcome.5, 6, 17–25, 54–56 A total of 1053 patients were evaluated and there was statistically significant heterogeneity between the studies [test for heterogeneity chi-squared = 76.57 (d.f. = 13), P < 0.0001]. A random effects model was used and patients treated with prokinetics were twice as likely to have an improvement in dyspepsia symptoms compared with patients receiving placebo (RRR, 48%; 95% CI, 27–63%) (Figure 3). Assuming a placebo response rate of 41%, four (95% CI, 3–6) patients require treatment with prokinetics to improve one case of non-ulcer dyspepsia that would not have been improved with placebo. Meta-regression evaluating the dose of drug, duration of therapy, method of randomization, concealment of allocation and dropout rate did not find any factor that explained the heterogeneity between the studies. A funnel plot exhibited significant asymmetry with all small trials showing a marked treatment effect and larger trials showing no benefit over placebo (Egger's test, P = 0.002) (Figure 4). This suggests that the significant effect of prokinetics in non-ulcer dyspepsia may be due to publication bias or differences in the quality of the studies. Three papers reported the dyspepsia score as a continuous outcome, and there was a trend for improvement in dyspepsia, but this was not statistically significant [overall standard mean difference (in favour of prokinetic) = 0.24; 95% CI, 0.6 to − 0.13].23, 26, 27

Figure 3.

Meta-analysis of prokinetic therapy compared with placebo in non-ulcer dyspepsia patients. Random effects model. Heterogeneity chi-squared = 76.57 (degrees of freedom = 13), P < 0.0001.

Figure 4.

Funnel plot of relative risk of remaining dyspeptic against sample size for prokinetic therapy trials. SE, standard error.

The prokinetic therapy trials tended to be small and poorly reported, with only two stating the method of randomization and concealment. Trials reported improvements in individual dyspepsia symptoms more often than with other therapies. There was a trend for heartburn to improve with prokinetic therapy (six trials assessing 413 patients:6, 17, 18, 22, 51, 52 RRR, 48%; 95% CI, − 30% to 79%). Prokinetic therapy was statistically significantly superior to placebo for improving epigastric pain (eight trials evaluating 636 patients:5, 6, 17, 19, 22, 24, 51, 52 RRR, 60%; 95% CI, 30–77%) and post-prandial fullness (five trials evaluating 351 patients:18, 22, 51–53 RRR, 30%; 95% CI, 2–50%).

Comparison between prokinetic therapies. One randomized controlled trial compared cisapride with metoclopramide in 60 non-ulcer dyspepsia patients.48 Another compared cisapride with domperidone in 84 patients,49 whilst a third compared metoclopramide with domperidone in 138 patients.50 There was no statistically significant difference between these interventions, although there was a trend for cisapride to be more effective (cisapride vs. domperidone: RRR, 48%; 95% CI, − 72% to 84%; cisapride vs. metoclopramide: RRR, 43%; 95% CI, − 75% to 81%). There was no difference between metoclopramide and domperidone (RRR, 1%; 95% CI, − 30% to 45%).

Mucosal protecting agents and antacids

Comparison with placebo. Bismuth salts were compared with placebo in five randomized controlled trials assessing a total of 311 patients.38–42 The majority of trials evaluating bismuth assessed the role of Helicobacter pylori eradication in non-ulcer dyspepsia. Trials were therefore predominantly in H. pylori-positive patients but, as bismuth salts alone are rarely successful in treating the infection, these were classified as trials evaluating the efficacy of bismuth in non-ulcer dyspepsia. The results showed a trend for bismuth salts to be more effective than placebo, which was of marginal statistical significance (RRR, 42%; 95% CI, 68% to − 4%; P = 0.07; random effects model) (Figure 5). Bismuth trials also demonstrated significant heterogeneity between the studies (Q = 18.02, d.f. = 4, P = 0.001); this was largely due to the inclusion of one trial with a low placebo response rate and a large treatment effect. There was no asymmetry in the Egger plot (P = 0.8), all trials being around the same size. Two trials evaluated both H. pylori-positive and H. pylori-negative non-ulcer dyspepsia patients and reported no obvious relationship between infection status and effectiveness.40, 43 Two papers gave dyspepsia scores as a continuous outcome and showed a statistically significant effect in favour of bismuth (overall standard mean difference = 0.57; 95% CI, 0.17–0.97).43, 44 Two of the bismuth trials reported the method of randomization and concealment.

Figure 5.

Meta-analysis of bismuth salts compared with placebo in non-ulcer dyspepsia patients.

Sucralfate was studied in two trials evaluating a total of 246 patients.45, 46 There was a reduction in dyspepsia (RRR, 29%; 95% CI, − 40% to 62%), which was not statistically significant. Antacids were not significantly better than placebo (RRR, − 2%; 95% CI, 24% to − 36%) in one randomized controlled trial evaluating 109 patients.29

Comparison between therapies. Bismuth therapy was compared with sucralfate in one trial evaluating 29 patients.44 There was no statistically significant difference between these interventions.

Discussion

This is the most comprehensive systematic review of therapies for non-ulcer dyspepsia. The results suggest that both prokinetic and H2RA therapies are more effective than placebo, as found by other systematic reviews.7, 8, 58, 59 This is the first systematic review, however, to critically evaluate the results using funnel plots, and these indicate that, although there may be benefit of prokinetics and H2RA therapy over placebo, the benefit may be less marked than the data suggest. Funnel plots describe the data in terms of the effect size (logarithm of the relative risk) of each trial on the x axis and the size of that trial (1/standard error) on the y axis.57 It is expected that small trials will show greater variation in the treatment effect, but this variation should be symmetrically distributed across the summary relative risk. Larger trials should show less variation and be close to the summary relative risk. A ‘funnel’ shape should therefore emerge with a spread of small trials at the base and a cluster of larger trials at the apex. Funnel plots of prokinetic and H2RA therapies, however, indicate that there are fewer small negative trials than expected and the larger trials show a smaller treatment effect than predicted by the meta-analysis. This is particularly the case for prokinetic therapy where the largest trials in the funnel plot show no benefit over placebo. Furthermore, all of the eligible trials evaluated cisapride, which is now unavailable in many countries60 due to the possibility of serious adverse events in heart disease patients.61 Data supporting the use of prokinetic therapy in non-ulcer dyspepsia are therefore poor, although domperidone and metoclopramide are relatively inexpensive and even a small improvement in dyspepsia symptoms with these drugs may be cost-effective.

The effect of H2RA therapy on non-ulcer dyspepsia may also be overestimated. This is indirectly supported by randomized controlled trials of proton pump inhibitor therapy in non-ulcer dyspepsia.4 These trials report smaller effects of proton pump inhibitors on non-ulcer dyspepsia symptoms4, 36 than this meta-analysis of H2RA therapy. The funnel plot suggests that, although the benefit of H2RA on non-ulcer dyspepsia may be overestimated, there may still be some effect on symptoms. There is little evidence that gastric acid secretion is abnormal in non-ulcer dyspepsia patients, and symptoms cannot be reproduced by acid infusion at endoscopy.62 Patients with a normal endoscopy may still have gastro-oesophageal reflux disease and this will respond to acid suppression.63 It is possible that the effectiveness of antisecretory therapy in non-ulcer dyspepsia is due to the treatment of a sub-group of patients with gastro-oesophageal reflux disease. H2RAs were not more effective at treating heartburn than other dyspepsia symptoms, but reflux disease is more appropriately diagnosed by validated gastro-oesophageal reflux disease questionnaires rather than by evaluating heartburn alone.64

There was a trend for bismuth salts to be effective in non-ulcer dyspepsia patients and this requires further evaluation. Non-ulcer dyspepsia is a chronic relapsing and remitting disorder, and yet all the trials identified followed up patients for less than 12 weeks. Trials evaluating therapies with a long-term follow-up are required.

Acknowledgements

The National Health Service (NHS) Executive Research and Development, Health Technology Assessment Programme (HTA project no. 96/37/01) funded this review. The views expressed in this paper are the opinions of the authors and do not necessarily reflect those of the NHS Executive. Dr Brendan Delaney is currently funded by an NHS R&D National Primary Care Career Scientist Award. The Cochrane Upper Gastrointestinal and Pancreatic Diseases Group is funded by the NHS R&D Directorate.

Ancillary