Drug treatment of patients with functional dyspepsia is controversial but H2 receptor antagonists have been the mainstay of treatment. For patients with symptoms suggestive of dysmotility, prokinetics such as cisapride have been used. A large number of clinical trials have been unable to produce definite answers as to whether any of these treatment modalities are truly efficacious. This is partly due to the fact that the methodology and reporting of the majority of trials evaluating the symptomatic effects of H2 receptor antagonists and cisapride are severely flawed. Based on the current literature, H2 receptor antagonists may possibly have a therapeutic gain of approximately 20% over placebo. Evaluating the therapeutic gain of cisapride is more difficult but meta-analyses indicate a somewhat larger effect.
- functional dyspepsia
- H2 receptor antagonists
Statistics from Altmetric.com
Drug treatment of patients with functional dyspepsia has been a matter of controversy for decades. The choice of therapy has been much influenced by the resemblance in clinical presentation to peptic ulcer disease, and H2 receptor antagonists have therefore been the mainstay of treatment. For patients with symptoms suggestive of dysmotility, prokinetics such as cisapride have been suggested. A large number of clinical trials in this area have been unable to produce definite answers as to whether any of these treatment modalities are truly efficacious. This is partly due to problems in designing and reporting and partly due to inherent methodological difficulties in clinical trials of a vaguely defined condition such as dyspepsia, in which the only relevant outcome measure is the patient's “gut feeling”.
This review focuses on H2 receptor antagonists and prokinetics. Only cisapride studies were included even though a number of trials have been reported evaluating domperidone. The majority of these studies were published in the 1970s and the last trial was published more than 10 years ago.1 Sample sizes in the domperidone studies were small and the methodology and reporting of these trials were severely inadequate.2,3 Furthermore, most clinicians today would prefer cisapride over domperidone if a prokinetic is prescribed.
Only trials published as a full article in English or with an abstract and tabulated results in English were included. Trials reported only in abstract form and unpublished trials were not considered. Even though unpublished data could be very useful these were not sought. However, for a complete review and meta-analysis using the Cochrane criteria, it would be essential to include data from unpublished studies because publication bias makes it likely that many of these trials showed no benefit of the active drug over placebo.
Trials that focused mainly on surrogate parameters, such as histological signs of gastritis or scintigraphic signs of gastric emptying, without a symptom evaluation, were not considered.
Functional dyspepsia, or non-ulcer dyspepsia, is a diagnosis of exclusion, based on dyspeptic symptoms in the absence of structural abnormalities on endoscopy. To what extent other abnormalities have been excluded by additional testing, for instance ultrasonography or oesophageal pH monitoring, is very variable. Consequently, most trials have included patients with a heterogeneous pathophysiology, and some studies, particularly those evaluating cisapride, have included patients with mild oesophagitis,4–7 previous peptic ulceration,7–10 and even patients with a previous vagotomy.8,11
The broad and non-specific definitions of dyspepsia that were used until the Rome definition was agreed upon in 199112 have complicated the selection of patients into trials. The Rome criteria explicitly recognise that epigastric pain or discomfort must be the predominant complaint in patients labelled as suffering from dyspepsia. Patients with predominant symptoms such as heartburn or acid regurgitation, suggestive of gastro-oesophageal reflux disease, should be excluded from the diagnosis of dyspepsia, even in the absence of structural abnormalities on endoscopy. This distinction was not made clear in many of the early trials and may thus complicate comparisons between studies over time because the inclusion criteria have obviously changed.
Evaluation of outcome
The placebo response is usually high in dyspepsia trials which should be taken into consideration when the outcome is evaluated, particularly for trials showing no benefit of the active drug. Placebo response rates in the studies evaluated in this review varied from 6% to 69%, and varied most in the cisapride studies (tables 1–3). These discrepancies reflect the different ways of defining a response, differences among study populations, and various methodological technicalities, such as the exclusion of placebo responders during a placebo run-in phase.
One of the most troublesome methodological problems is the lack of validated outcome measures. The average dyspeptic patient complains of at least three different dyspeptic symptoms apart from epigastric pain or discomfort.44 As a consequence, multiple testing of effects on a number of different symptoms may lead to the false conclusion that the active drug is superior to placebo.45 Global assessments offer more valid outcome measures.46 Furthermore, dyspeptic symptoms are not stable over time,47 and this creates specific problems for the crossover designs used in many trials evaluating cisapride.
Study populations have usually been recruited from the small proportion of dyspeptic patients referred for endoscopy and from highly specialised referral centres with a specific interest in dyspepsia. Patients with dyspeptic symptoms who obtain relief from over the counter medicine and patients who respond favourably to empirical drug treatment in primary care are less likely to be referred for endoscopy or to referral centres and thus recruited to clinical trials.48 Accordingly, there is a risk that the study populations in these trials constitute non-responders to drug treatment. As a consequence, the implications of such drug trials are uncertain or unknown for the vast majority of dyspeptic patients who are managed in primary care settings. Only six of the 45 studies evaluated in this review recruited patients directly from the primary care setting.
Given the heterogeneity of dyspepsia, it is unlikely that a single drug will work for all patients. The placebo response is high and some patients may even deteriorate while receiving active drug treatment. As a consequence, parallel group studies may mask individual responders to treatment. This problem has been addressed by special study designs, such as the single subject trial designs, multiple crossover designs,49–54 and by post hoc analysis of patient characteristics in groups of responders.
Dyspeptic symptoms are usually chronic or recurrent. It is thus a surprise that in the majority of trials patients were treated for six weeks or less, and no study included long term follow up after cessation of treatment.
H2 RECEPTOR ANTAGONISTS IN DYSPEPSIA
Twenty two studies comparing a H2 receptor antagonist with placebo were evaluated.
Parallel group studies
A summary of the trial design, number of randomised patients, inclusion criteria, and outcome measures is reported in table 1.
Epigastric pain or ulcer-like symptoms were the main inclusion criteria in seven of the studies (table 1).9,13,15–19 In the remaining studies, the dyspepsia type was not specified or a mixture of different dyspeptic symptoms was allowed.
In seven studies, the authors claimed a statistically significant benefit of the active drug over placebo.10,14,17,21–24 However, in three studies,14,22,24 a reported significant effect of the active drug could not be confirmed after a simple re-analysis of the raw data presented in the tables in the articles. Thus only four studies10,17,21,23 showed a significant effect of the H2 receptor antagonist over placebo.
Table 1 summarises the estimated therapeutic gain (difference in success rates, as defined by the individual trial, between placebo and active drug and the related 95% confidence interval) for 12 of the studies. In the remaining two studies, the reported data did not allow an estimate of therapeutic gain. Placebo response rates in the three large scale studies reporting a significant effect of the H2 receptor antagonist were 36%,10 57%,17 and 59%,21 and the therapeutic gains 14%, 20%, and 21%, respectively. In the study by Müller et al, at least one fifth of the included patients had a positive history of peptic ulcer disease or gastro-oesophageal reflux disease, which may have contributed to the significant effect of ranitidine over placebo in patients with acid related symptoms.10 The most recent study, which recruited unselected dyspeptic patients directly from primary care, was unable to detect any benefit of nizatidine over placebo.25
Seven studies have used crossover or multiple crossover designs. The study by Talley et al was unable to show any benefit in the global assessment of symptoms.55 The other six studies were single subject trials using a multiple crossover design to identify individual responders to treatment.49–54 All of these studies claimed that a small proportion of patients, typically 10–20%, obtained significantly better symptom relief during the periods on the H2 receptor antagonists compared with periods on placebo. A therapeutic gain cannot be estimated from these study designs. Patients who responded to H2 receptor antagonists were characterised by heartburn or other features suggestive of gastro-oesophageal reflux disease as well as dyspepsia.
Two meta-analyses have tried to summarise the overall symptomatic effects of H2 receptor antagonists in dyspepsia. Based on six trials, Dobrilla et al estimated a significant therapeutic gain over placebo in the order of 20%.56 This conclusion was confirmed in a more recent analysis by Finney and colleagues.57 None of the meta-analyses included unpublished data however.
CISAPRIDE IN DYSPEPSIA
In seven of the 25 studies, the main entry criteria were epigastric pain or discomfort, or so-called ulcer-like dyspepsia (tables 2 and 3).8,11,28,31,35,36,39 In the remaining 18 studies, patients were troubled by symptoms suggestive of dysmotility or a mixture of dyspeptic symptoms, including symptoms associated with gastro-oesophageal reflux disease.
Parallel group studies
A total of 16 studies compared cisapride with placebo in a parallel group design.5,6,8,11,25–36 These are summarised in table 2. Some of the early studies randomised rather few patients, they often claimed a positive response, and two of these studies allowed patients with mild oesophagitis.5,6 Seven of the 16 studies reported a significant improvement with cisapride compared with placebo,5,6,8,11,26,29,33 but five of these trials randomised patients with dysmotility-like or mixed dyspeptic symptoms.5,6,26,29,33
Five studies randomised more than 100 patients.25,26,28,34,35 Of these five studies, only the early study by Rösch showed a significant improvement with cisapride.26 The other four, all published within the last seven years, were negative.
The majority of studies recruited patients in secondary or even tertiary centres, evaluating only highly selected patients. Only three studies recruited patients directly from primary care and all were negative.25,31,34
A number of crossover studies have been reported. It is particularly difficult to review this part of the literature and the majority of these trials were hampered by imperfect methodology and poor reporting.
The main findings from nine trials are summarised in table 3.4,7,37–43 Seven of the trials had randomised fewer than 40 patients.4,37–43 The majority of the trials claimed a significant effect of the active drug.4,7,37–42 However, none of the studies complied with common standards concerning design, analysis, and reporting of crossover trials. For instance, a washout period to exclude period effects before a shift to the alternate treatment was seldom included. Furthermore, the results reported by Deruyttere et al may represent double publication, summarising the results from three previous publications even though this was not specifically stated in the paper.40
Three meta-analyses have evaluated the effects of cisapride. The analysis by Dobrilla et al based on seven studies concluded that cisapride had a therapeutic gain of 39%.56 That conclusion was based on a total of only 275 patients from six crossover studies and just one parallel group study.
In the meta analysis by Finney et al, eight studies were included.57 Based on 415 patients from three crossover trials and five parallel group studies, an overall therapeutic gain of 36% was found. However, this meta-analysis did not include the recent negative large scale studies by de Groot and de Both,34 Champion and colleagues,35 Yeoh and colleagues,36 and Hansen and colleagues.25 These four studies alone had more patients randomised compared with the eight studies included in the meta-analysis.
Using a somewhat different design, Veldhuyzen van Zanten et al have published a meta-analysis including 18 studies.58 They found cisapride to be efficacious based on global assessment rated by the investigator. However, individual symptoms such as epigastric pain, abdominal distension, and nausea were not improved. Furthermore, none of these meta-analyses included unpublished data.
The methodology and reporting of the majority of trials evaluating the symptomatic effects of H2 receptor antagonists and cisapride are severely flawed and the published conclusions should be evaluated very carefully. Based on the current literature, H2 receptor antagonists may possibly have a therapeutic gain of approximately 20% over placebo. Patients with heartburn and other symptoms suggestive of gastro-oesophageal reflux disease are most likely to respond. Evaluating the therapeutic gain of cisapride is more difficult but meta-analyses indicate a somewhat larger effect. This conclusion however is based mainly on highly selected patients who often have symptoms suggestive of dysmotility rather than epigastric pain, and whether this effect translates back to the vast majority of dyspeptic patients, who are managed in primary care, is very doubtful.
We need long term studies of better quality in unselected patients in primary care before we can draw any firm conclusions about the effects of these drugs in the target population. Head to head comparisons between different treatment modalities are also needed.
Conflict of interest: This symposium was sponsored by AstraZeneca, makers of omeprazole. The author of this paper has received sponsorship for travel and an honorarium from AstraZeneca.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.