Efficacy of cisapride and domperidone in functional (nonulcer) dyspepsia: a meta-analysis

THE AMERICAN JOURNAL OF GASTROENTEROLOGY © 2001 by Am. Coll. of Gastroenterology Published by Elsevier Science Inc.

Vol. 96, No. 3, 2001 ISSN 0002-9270/01/$20.00 PII S0002-9270(00)02314-5

Efficacy of Cisapride and Domperidone in Functional (Nonulcer) Dyspepsia: A Meta-Analysis Sander J. O. Veldhuyzen van Zanten, M.D., Michael J. Jones, Ph.D., Marleen Verlinden, Ph.D., and Nicholas J. Talley, M.D. Division of Gastroenterology, Dalhousie University, Halifax, Canada; Jones and Just Pty Ltd and Department of Psychology, Macquarie University, New South Wales, Australia; Urology Division, Abbott Laboratories, Abbott Park, Illinois; and Department of Medicine, University of Sydney, Nepean Hospital, New South Wales, Australia

BACKGROUND: The efficacy of prokinetic agents in functional (nonulcer) dyspepsia has been questioned based on recent trial results. We performed a meta-analysis to determine the efficacy of cisapride and domperidone in functional dyspepsia. METHODS: Computer and manual searching was used to identify placebo-controlled studies that included ⬎20 patients. The statistical analysis focused on: global assessment by the investigator, epigastric pain, early satiety, abdominal distension and nausea (all rated on four-point scales). Results are reported as odds ratios (OR) in favor of treatment. Regression analysis was performed to evaluate possible effect modifiers. The relationship between improvement in gastric emptying and symptoms was also evaluated. RESULTS: For cisapride, 17 studies met the inclusion criteria, but varying numbers of studies had to be used for the different outcome measures. For all outcome measures, there was a statiscally significant benefit in favor of cisapride: global assessment of improvement by the investigator or patient (OR 2.9, 95% CI 1.5–5.8), epigastric pain (OR 0.19, 95% CI 0.05– 0.7), early satiety (OR 0.18, 95% CI 0.9 – 0.4), abdominal distension (OR 0.32, 95% CI 0.1– 0.7), and nausea (OR 0.26, 95% CI 0.1– 0.5). Age of patient, year of publication, and country where study was performed had only small modifying effects. There were insufficient data to determine whether there is a relationship between improvement in gastric emptying and response to treatment. For domperidone, four of eight studies could be used for the analysis of global assessment of improvement by the investigator. This showed an OR of 7.0 (95% CI 3.6 –16) in favor of domperidone. CONCLUSIONS: Both cisapride and domperidone seem to be efficacious in functional dyspepsia, although this conclusion is largely based on global assessment by the investigator, which may not be an optimal outcome measure. (Am J Gastroenterol 2001;96:689 – 696. © 2001 by Am. Coll. of Gastroenterology)

INTRODUCTION Cisapride is a prokinetic agent which stimulates the motility of the digestive tract by selectively enhancing acetylcholine release in the myenteric plexus of the gut, probably via action as a partial serotonin (5-HT4) agonist (1, 2). Cisapride has been shown to stimulate antroduodenal motility and accelerate gastric emptying in patients with delayed gastric emptying (3, 4). Cisapride can also enhance the peristaltic activity in the esophagus and has an efficacy similar to histamine-2 (H2) receptor antagonists in patients with gastroesophageal reflux disease (5, 6). Domperidone is a peripherally acting dopamine antagonist (7–9). It increases lower esophageal sphincter pressure, accelerates gastric emptying, and enhances antropyloric motility (7). It has also been used in patients with gastroesophageal reflux disease and diabetic gastroparesis (7, 10). Functional (nonulcer) dyspepsia is a common functional GI disorder for which no definite cause has been established. Although the pathophysiological mechanisms that explain functional dyspepsia remain obscure, delayed gastric emptying has been considered to be a potential contributing factor (11). Currently, there is uncertainty as to whether truly efficacious treatments are available for functional dyspepsia. Several randomized clinical trials of cisapride in functional dyspepsia patients have been published. Although initial studies showed a benefit of treatment, more recent large trials were negative. For domperidone, both positive and negative study results have also been published. As a result, it is controversial as to whether cisapride or domperidone are efficacious in functional dyspepsia. We therefore performed a meta-analysis of all cisapride and domperidone studies to determine whether these prokinetic agents are efficacious in patients with functional dyspepsia. We also evaluated whether an improvement of dyspepsia symptoms was associated with changes in gastric emptying.

690

Veldhuyzen van Zanten et al.

MATERIALS AND METHODS Cisapride and domperidone-functional dyspepsia trials were identified using both computerized and manual methods of searching. Only publications that described original treatment studies were searched for but review articles were used to help identify relevant references. Nine cisapride studies and seven domperidone studies had been identified in an earlier search for a systematic overview of functional dyspepsia treatment trials (12). For computerized searching, the MEDLINE database was searched from 1986 to 1999 using the following medical subject headings: cisapride, domperidone, dyspepsia, clinical trial, and human. Only articles published in English and as full papers were evaluated. Data that was available only in abstract form were not reviewed. A fully recursive manual search of cited references in published studies was performed to help locate other studies. To evaluate whether delayed gastric emptying affects the response of cisapride or domperidone in functional dyspepsia patients, we searched the literature using the following medical subject headings: cisapride, domperidone, gastric emptying, controlled clinical trial, and human. Inclusion Criteria The criteria for inclusion for trials was as follows: 1) randomization was used for treatment allocation; 2) the study was double-blind; 3) the study was placebo-controlled; 4) the sample size was ⱖ20 patients; and 5) the target group was patients suffering from functional dyspepsia. The sample size of 20 patients was arbitrarily chosen to avoid including studies with very small sample sizes, which might introduce unreliable statistical inference. For inclusion in the meta-analysis, a study had to fulfill all five inclusion criteria. The inclusion criteria were defined a priori by two authors (S.J.O.VvZ. and N.T.). Study Evaluation and Data Extraction All studies included were carefully evaluated by the authors. A data collection form was designed, which contained information on study design, duration of treatment, dose used, age, gender, country where study was performed, year of publication, and outcome measures used. Choice of Outcome Measures and Data Extraction There was marked variation in the symptoms that were recorded and the way in which their severity was assessed. The statistical analysis was guided by the need to analyze those outcome measures that were used by the largest possible number of studies. After a detailed review and extensive discussion among the four authors, it was decided to focus the statistical analysis on the following outcomes: 1) global assessment of treatment by the investigator or the patient with the result rated as excellent versus other (this global assessment was usually reported on a four-point scale: excellent, good, fair, or poor); 2) global assessment of treatment by the investigators or the patient, with the result

AJG – Vol. 96, No. 3, 2001

rated as good or excellent versus other; 3) epigastric pain or discomfort; 4) early satiety; 5) abdominal distension; and 6) nausea. The individual symptoms were also assessed on a four-point scale. The decision to use these six outcome measures was driven not only by the outcome measures that were available in the individual studies but also by the availability of sufficiently comparable rating scales and by whether the study provided sufficient details about the distribution of the data (SD or variance). All articles reported symptom results as bar charts, either as percentages or as numbers of subjects. In some instances there were also text references to percentages displayed in charts. Measurement of bar heights was performed manually. Computer software designed for this purpose (UNGRAPH; Biosoft, Cambridge, UK) was initially tested (13) but was found to be less reliable and agreed less well with text references than manual measurement. All studies retrieved except for that by Al Quorain et al. (14) provided the global rating data as numbers and/or percentages of subjects in each category of efficacy; the study by Al Quorain et al. provided a bar chart of percentages, which were measured by hand. The accompanying legend provided validation of the measurements, which proved to concur closely with those reported in the text. Odds ratios for global assessment were calculated in two ways. One used a rating of excellent versus the remainder as the positive result, whereas the other used a rating of good or excellent as the positive case. The scales used to measure the severity of individual symptoms varied in the studies. For calculation of the ORs, the data were dichotomized into moderate or more versus mild at the end of treatment, as all scales could be reported in this way. Statistical Analysis Two forms of analysis were conducted. The first combined the information from the trials into a single estimate of effect. The second analysis evaluated the presence of potential effect modifiers that might influence the efficacy of cisapride or domperidone. Subject age and gender were considered. Age was recorded as the mean or median age (some studies only reported one statistic and they were treated as equivalent). Gender was recorded as the percentage of study subjects who were female. All trials were either randomized controlled trials or two-period crossover trials. Given the potential for carry-over effects in crossover trials, we tested whether experimental design was an effect modifier by comparing crossover to parallel group design. Year of publication was analyzed, as positive studies may be more often reported when a drug is first tested in clinical trials. Finally, as many of the cisapride studies were conducted in Belgium, we also tested whether the combined estimate of the effect of cisapride was influenced by the country in which the study was conducted (Belgium vs others). The approach of DerSimonian and Laird to random effects meta-analysis was adopted (15). This method allows

AJG – March, 2001

for true between-study variation in effect as well as for random sampling error in the calculation of standard errors. The linear model is a regression-type method in which the dependent variable is a measure of effect and the independent variables are potential effect modifiers (16). The OR was adopted as the measure of effect. The analysis was conducted on the log OR. This has an asymptotic normal distribution and known SE on which to base study weights. The homogeneity of study results was assessed through Cochran’s test of homogeneity (15). In addition, the magnitude of between-study variance was quantified by ␶2. Large values of ␶2 indicate greater levels of variance. Plots of individual study effects with 95% confidence intervals and the combined estimate of all studies are reported.

RESULTS For cisapride, a total of 74 citations were retrieved by both computer search and manual search of cited references. Of these, 56 were ineligible, leaving 18 studies for the final analysis (14, 17–33). For domperidone, a total of 41 citations were identified, of which eight met our inclusion criteria (34 – 40). Using the medical subject headings cisapride or domperidone and either dyspepsia or clinical trials, the yield of computer searches was unreliable, as several of the potentially eligible studies would have been missed. The final number of studies was reached by using different combinations of computer and manual searching. Cisapride We became aware, and subsequently confirmed, that two of the cisapride studies were in fact publications of the same result (17, 24); we therefore eliminated one of these from the analysis (17). The sample size of the 17 eligible studies varied from 22 to 135 (14, 18 –33). Five studies used a crossover design and 12 a parallel group design. One study, which compared cisapride to nizatidine and placebo, had a sample size of 330 patients of whom 109 were randomized to cisapride (33). The dose of cisapride varied from 5 to 20 mg t.i.d., and the study duration ranged from 2 to 6 wk. A global assessment of treatment effect by the investigators was available in 11 of the 17 studies. In addition, there were two studies (18, 20) that reported global assessment by the patient. When combined, this left 13 studies for global assessment as the main outcome measure. Four studies were not eligible because no global assessment by the investigator was performed (30, 32, 33) or a VAS scale was used which could not be transformed for the analysis (31). Of the 13 remaining studies, two (24, 28) only provided the number or percentage of subjects in whom the response was good or better, and by subtraction, the number of subjects in whom the response was less than good. In the study by Champion et al., results were given for excellent or good as outcome but no separate data was given for an excellent outcome (29). Therefore, there are 13 studies that applied a global assessment rating as good or better, and 10 studies that used

Cisapride and Domperidone in Functional Dyspepsia

691

Table 1. Global Response—Combined Estimates for Cisapride Studies Homogeneity Outcome Global, Global, Global, Global, Global, Global, Global, Global,

excellent* ⱖgood* excellent† ⱖgood† excellent‡ ⱖgood‡ excellent§ ⱖgood§

OR

95% CI

p Value

␶2

2.91 3.38 1.99 2.72 3.02 3.20 1.78 2.46

1,45, 5.83 2.04, 5.58 1.11, 3.57 1.70, 4.36 1.24, 7.34 1.82, 5.65 0.83, 3.83 1.44, 4.19

0.001 ⬍0.001 0.04 0.001 ⬍0.001 ⬍0.001 0.02 0.001

0.82 0.57 0.35 0.39 1.18 0.66 0.54 0.45

* Uses all available studies. For the excellent vs ⬍excellent outcome, 10 articles are used. For the ⱖgood vs ⱕmoderate outcome 13 articles are used, as three did not report “good” and “excellent” separately. † Excludes studies by Al Quorain et al. (14) and Chung et al. (25) with outlying result(s). ‡ Excludes studies by Francois et al. (18) and Hannon et al. (20), global assessment by patient. § Excludes studies mentioned in 2 and 3. Note: The odds ratios in Table 1 compare the odds of scoring in the higher category of response in cisapride-treated subjects with the odds of scoring in the higher category of response in untreated (placebo) subjects. Hence, values ⬎1.0 indicate a benefit in favor of cisapride, whereas values ⬍1.0 indicate a detrimental effect of taking cisapride. The ␹2 statistic is Cochran’s test of homogeneity of study estimates of the treatment effect, whereas ␶2 is a quantitative index of study heterogeneity. Values of ␶2 close to zero indicate little heterogeneity, whereas large values indicate substantial heterogeneity. CI ⫽ confidence interval; OR ⫽ odds ratio.

a global assessment rating that included excellent. For all studies the end of treatment result were used in the analysis although some studies reported both interim (e.g., at 2 wk) and end of study results (4 wk). The study by Champion et al. evaluated two doses of cisapride, both of which were used (29). The ORs compare the odds of scoring in the higher category of response in cisapride-treated subjects with the odds of scoring in the higher category of response in untreated (placebo) subjects. Hence, values ⬎1.0 indicate a benefit in favor of cisapride, whereas values ⬍1.0 indicate a detrimental effect of taking cisapride. The combined sample size for global response defined as excellent was 646 patients and was 970 patients if response is defined as excellent or good. Table 1 shows that treatment with cisapride offered a clear benefit, with an OR of 2.91 (95% confidence interval [CI] 1.45–5.83) if global response is defined as excellent, and as 3.38 (95% CI 2.04 –5.58) if response is defined as excellent or good. Figures 1 and 2 show the distribution of the log ratios for the two global assessment analyses. There are two studies (14, 25) that are outliers (this is clearer visually on a normal scale, but the ORs are drawn on a log scale to improve resolution for the other studies). If these two studies are excluded, the OR decreases to 1.99 (95% CI 1.11–3.57) if global response is rated as excellent and to 2.72 (95% CI 1.70 – 4.36) if response is defined as excellent or good. All studies reported global assessment by the investigator except the trials by Francois and de Nutte and by Hannon (18, 20). As can be seen in Table 1, the exclusion of these two studies had only a small influence on the ORs. These results

692

Veldhuyzen van Zanten et al.

AJG – Vol. 96, No. 3, 2001

Table 2. Univariate Effect Modifiers Outcome

Basis

Excellent

All Robust All Robust All Robust All Robust All Robust All Robust All Robust All Robust All Robust All Robust

ⱖGood

Figure 1. Plot of individual study and combined findings (excellent).

demonstrate that cisapride is superior to placebo for this outcome measure. Univariate Effect Modifiers The ORs in Table 2 compare the effects of cisapride across a number of design and study subject characteristics. Where the 95% confidence interval does not include 1.0, the corresponding characteristic can be taken to have an effect on the estimated OR for cisapride. The table is based on different numbers of studies: all studies with a global assess-

Effect Modifier

OR

Lower

Upper

Age Age % Female % Female Design Design Country Country Year Year Age Age % Female % Female Design Design Country Country Year Year

0.87 0.85 0.96 0.97 1.03 0.59 1.20 2.32 0.97 0.85 0.92 0.98 0.97 0.98 0.65 0.45 1.66 2.46 0.93 0.87

0.80 0.70 0.90 0.93 0.19 0.17 0.25 0.84 0.78 0.75 0.84 0.83 0.93 0.94 0.20 0.18 0.56 1.06 0.81 0.80

0.94 1.03 1.01 1.01 5.58 1.96 5.76 6.41 1.21 0.96 1.00 1.16 1.01 1.02 2.04 1.13 4.95 5.69 1.06 0.96

The odds ratios in Table 2 compare the effect of cisapride (as defined in Table 1) across a number of design and study characteristics. All ⫽ includes all studies for global assessment rated as excellent (N ⫽ 10) or excellent or good (N ⫽ 13). Robust ⫽ excludes the two outlying studies (14, 25). Design ⫽ randomized controlled trial vs cross-over study. Country ⫽ country of study origin: Belgium vs other. Year ⫽ year of publication. OR ⫽ odds ratio.

ment rating as excellent (10 studies), excellent or good (13 studies), and excellent with the two outlying studies excluded (eight studies), and excellent or good with the two outlying studies excluded (11 studies). Some study characteristics did reach statistical significance univariately. Age (decreasing efficacy with increasing age, OR 0.87), year of publication (diminished efficacy in more recent studies, OR 0.85), and country where study originated (OR 2.46) modified study outcomes. Gender and trial design (studies with a parallel group design compared to crossover studies) did not modify the effect of treatment. Multivariate Effect Modifiers There were no combinations of design or subject characteristics that exhibited statistically independent effects on the OR for either excellent or good or excellent outcomes.

Figure 2. Plot of individual study and combined findings (good or better).

Individual Symptoms The number of studies available for inclusion in the metaanalysis of individual symptoms was much lower than for the global assessment outcome. There were five studies available for epigastric pain or discomfort (14, 21, 22, 25, 33), four for early satiety (14, 18, 21, 24), three for abdominal distension (18, 21, 25), and five for nausea (14, 18, 21, 25, 33). It is important to note that different studies were used for different symptoms. The studies are listed in the footnote to Table 3. The ORs in Table 3 compare the odds of a particular symptom improving in cisapride-treated subjects with the

AJG – March, 2001

Cisapride and Domperidone in Functional Dyspepsia

Table 3. Combined Information Across Individual Symptoms Outcome

Basis

OR

Lower

Upper

␶2

p Value

Epigastric pain Early satiety Abdominal distension Nausea

All*

0.19

0.05

0.66

1.46

0.002

All†

0.18

0.09

0.36

0.00

0.4

All‡

0.32

0.14

0.74

0.00

0.9

All§

0.26

0.13

0.54

0.00

0.6

* Al-Quorain et al. (14), Rosch et al. (21), De Nutte et al. (22), Chung et al. (25), and Hansen et al. (33). † Al-Quorain et al. (14), Francois et al. (18), Rosch et al. (21), and Chung et al. (25). ‡ Francois et al. (18), Rosch et al. (21), and Chung et al. (25). § Al-Quorain et al. (14), Francois et al. (18), Rosch et al. (21), De Nutte et al. (22), Chung et al. (25), and Hansen et al. (33). Note: The odds ratios (OR) in Table 3 compare the odds of a particular symptom occurring in cisapride-treated subjects with the odds of it occurring in untreated (placebo) subjects. Hence, values ⬍1.0 indicate a benefit in favor of cisapride, whereas values ⬎1.0 indicate a detrimental effect of taking cisapride.

odds of it occurring in untreated (i.e., placebo-treated) subjects. Hence, values ⬍1.0 indicate a benefit in favor of cisapride, whereas values ⬎1.0 indicate a detrimental effect of taking cisapride. All four symptoms showed a statistically significant benefit of cisapride over placebo. Although for early satiety, abdominal distension, and nausea the available studies were clearly homogeneous, statistically significant between-study variability (␶ ⫽ 1.46) was observed for epigastric pain. This is reflected in a nonzero estimate of between-study variance. Domperidone Results Only eight of the domperidone studies met our inclusion criteria (34 – 41). Five studies used a parallel group design and three a crossover design. The dose of domperidone varied from 10 mg t.i.d. to 10 mg q.i.d. The duration of the studies ranged from 2 to 4 wk. As most studies reported only the percentages of patients with symptoms, individual symptoms could not be pooled for meta-analysis. Only four studies using a global assessment rated by the investigator or the patient (34 –36, 39) were analyzed, defining a response to treatment as excellent or good on this scale. The combined sample size of these four studies was 211 patients. As shown in Table 4, an effect in favor of domperidone over placebo was found, which is unlikely to have occurred by chance. Table 4. Domperidone: Global Response of Treatment Effect Rated by the Investigator or Patient as Excellent or Good Homogeneity Outcome

OR

95% CI

␹2

␶2

Global assessment rated as excellent Global assessment rated as ⱖ good

7.44

3.57, 15.51

2.02

0.00

17.38

9.16, 32.95

1.84

0.00

Data are based on Ref. 34 –36, 39. CI ⫽ confidence interval; OR ⫽ odds ratio.

693

Correlation Between Gastric Emptying and Symptoms CISAPRIDE. We searched for evidence that improvement in gastric emptying correlated with improvement in symptom severity. Of the nine potentially relevant citations identified by our search, only two investigated whether the effect of cisapride on dyspepsia symptoms is related to changes in gastric emptying (26, 30). In the study by Kellow et al. (26), gastric emptying for solids and liquids was only assessed at baseline. Contrary to expectations, a beneficial effect of cisapride was only seen in patients with normal baseline gastric emptying. The study by Jian et al. (30) showed that cisapride improved both gastric emptying of liquids and solids, and this correlated with an improvement in symptoms. However, the total number of patients with abnormal emptying was small (N ⫽ 17) and the symptom reduction did not reach statistical significance. DOMPERIDONE. Of the eight potentially eligible studies, only two met our inclusion criteria (40, 41). In the study by Sarin et al. (40), 44 patients were treated in a crossover study with either domperidone or placebo. Gastric emptying times decreased during domperidone treatment and symptoms improved, but is unclear whether there was a correlation between the two. In the study by Duan et al. (41), 60 patients were randomized to a 7-day treatment of either domperidone 20 mg t.i.d. or placebo. Gastric emptying studies using a standard meal were performed before and after treatment. The dyspepsia symptoms improved both with domperidone and with placebo. In the 29 patients with delayed gastric emptying, domperidone decreased gastric emptying time significantly compared to placebo. In these patients both bloating and early satiety improved more during treatment with domperidone.

DISCUSSION This meta-analysis of randomized, placebo-controlled clinical trials has shown that there is a significant benefit in favor of cisapride in patients suffering from functional dyspepsia. This applied to measures of global assessment of efficacy by the investigator or the patient, as well as to individual measures of epigastric pain or discomfort, early satiety, abdominal distension, and nausea. The dose of cisapride in the studies varied from 5 to 20 mg t.i.d. and treatment duration varied from 3 to 6 wk. One strength of our study is that an exhaustive search was performed to identify all the potentially relevant studies. However, for statistical analysis of individual symptoms, the number of available cisapride studies (three to five) was much smaller than for global assessment by the investigator (13 studies). Although the analysis for domperidone also suggested a benefit of active drug over placebo, the analysis was based on only four studies and used global assessment of efficacy by the investigator or the patient. A recent meta-analysis by Finney et al. also concluded that cisapride and domperidone were better than placebo in functional dyspepsia (42). But

694

Veldhuyzen van Zanten et al.

this study included only eight cisapride and four domperidone studies and therefore their results need to be interpreted cautiously. Although the data suggest a potential benefit of cisapride in the treatment of functional dyspepsia management decisions should be based on a risk-benefit assessment. Given the overall benign natural history of functional dyspepsia the recent reports of rare but serious and sometimes life threatening cardiac arrhythmias related to the use of cisapride need to be taken into consideration (43, 44). Recently, the FDA has recommended that a baseline electrocardiogram is obtained before treatment with cisapride is initiated. Clearly, physicians must weigh the potential benefits of treatment against the small risk of adverse events. An important problem in a meta-analysis such as ours is that there is considerable variation in the design, execution, and analysis of studies. A recent systematic review highlighted the methodological shortcomings in many treatment trials of functional dyspepsia patients (12). The study evaluated 52 functional dyspepsia treatment trials and found that they were limited by a suboptimal study design, unclear inclusion criteria, and a short duration of follow-up. Only five of the 52 studies used previously validated outcome measures. None of the eligible studies in this meta-analysis used adequately validated outcome measures, although they do have face validity. In future trials of prokinetic agents, it is critical that sufficient attention is paid to the recording of subjective assessments by patients or by physicians. Recently, a report on optimal study design of functional GIdisorders, which includes functional dyspepsia, was published as part of the Rome II Working Group (45). We planned to provide an overall quality score for each individual study; however, this was difficult, as studies were often clear and adequate in some areas (e.g., inclusion criteria or description of symptoms) but weak in others (i.e., measurement scales to assess severity or statistical analysis). For these reasons, we abandoned assigning quality scores to individual studies. Overall, it is fair to say that several studies seemed to be limited by weak design and poor reporting of results. A recent review of meta-analysis demonstrated that inclusion of trials of poor quality can erroneously increase the reported benefit of treatment (46). There are several other limitations that warrant caution in interpreting our results. Depending on the main outcome measure, the combined sample size for the cisapride studies varied from 646 to 970, which is small. For domperidone the combined sample size was 211, which is the likely explanation as to why the OR was higher. The variations in study design among the eligible trials made it difficult to pool data for a formal meta-analysis. We spent considerable time identifying the outcome measures that could be used for our analysis. The choice of the six outcome measures was driven by several factors. These included that the measures were reported 1) by a sufficient number of studies, 2) using scales that could be combined in the analysis (most studies used interval, or Likert scales), and 3) in sufficient detail to

AJG – Vol. 96, No. 3, 2001

assess the distribution of the data. With regard to the latter, a few studies reported only mean scores without standard deviations or data on the variance, which made it impossible to use those studies in the pooled analysis (30, 31). The scales used in the studies were all either four- or five-point scales. Fortunately, the wording used made it clear that the top two categories of the five-point scales could be equated to the top category of the four-point scales. The largest number of studies was available global assessment by the investigator, which was chosen as the main outcome measure. When two studies using global assessment by the patient were added to the total, there were 13 studies. Inclusion or exclusion of these two studies (18, 20) had little effect on the OR. Deletion of the two outlying studies with positive results did not affect our conclusions although, as expected, the OR decreased from 2.91 to 1.99 for the outcome “good or better” and from 3.38 to 2.72 for the outcome “excellent.” Notably, there was considerable heterogeneity among the studies; this is reflected by the statistical test for homogeneity (␶ statistic), which varied form 0.39 to 1.18. Another reason for caution in interpreting our results is that global assessment of efficacy by the investigator is not the ideal outcome measure (12, 45). A global assessment done by the patient may be better, as it seems unlikely that a physician rating of a patient’s symptoms is more objective than an assessment by the patient. For this outcome measure there was substantial heterogeneity, indicating that there was considerable variation among the different studies. For the four outcome measures that used severity of an individual symptom (epigastric pain, early satiety, abdominal distension, or nausea), only three to five studies could be used for the meta-analysis. Although the analysis of these four symptoms also showed a beneficial effect of cisapride, the results should be viewed with caution. We searched for possible effect modifiers. Age, year of publication, and country where study was performed showed a small but statistically significant effect depending on the number of studies included. However, the ORs were small, and none of the factors were independent predictors in a multivariate model. We have no good explanation as to why age might diminish the effect of cisapride, although it is possible that the number of 5-HT4 receptors is diminished or the receptors become less responsive with age. The small effect seen with year of publication may be a proxy for either changes in trial design (e.g., different inclusion criteria) or changes in study subject characteristics. The initial studies of cisapride were predominantly carried out in Belgium, where the drug was developed. Another important aspect to consider in study design is the duration of therapy, which was only 3– 6 wk in the studies included. No further follow-up data were reported after treatment. Short duration of follow-up has been identified as a problem in studies (12); this is surprising, as functional dyspepsia is a chronic condition with symptoms that may recur frequently over time. It is important to show

AJG – March, 2001

that therapy maintains its efficacy over longer periods of time. There is uncontrolled data suggesting that after therapy with cisapride, there can be sustained symptom relief (47), but this has not been adequately tested. Although there seems to be agreement that patients with gastroesophageal reflux disease (GERD) should be excluded from functional dyspepsia trials, in practice there undoubtedly is overlap between these two disease entities (12). A substantial proportion of GERD patients do not have endoscopic evidence of esophagitis, making reliance on heartburn as the dominant symptom the most common way by which these patients will be identified. Some contamination of functional dyspepsia studies with GERD patients is unavoidable and mimics the real-life clinical situation. In one study, the presence of esophagitis was not even an exclusion criterion (24). Cisapride has proven efficacy in patients with mild-to-moderate reflux esophagitis, so it is possible that the apparent beneficial effect in functional dyspepsia is largely explained by improvement in true reflux (5, 6). In the study by Kellow et al. (26), cisapride was beneficial mainly in patients with reflux-like dyspepsia. Future studies that include 24-h pH monitoring will help to resolve this issue. For domperidone, only global assessment by the investigator from four studies could be included in a formal metaanalysis but this did show a beneficial effect. These trials had limitations similar to the those of the studies described above. Dyspepsia and Gastric Emptying We reviewed the literature for evidence that the beneficial effect of cisapride or domperidone might be mediated through changes in liquid and/or solid gastric emptying. Only two relevant cisapride and two domperidone studies were retrieved by our search (26, 30, 40, 41). For cisapride, one study measured gastric emptying only at baseline, and the other included only 17 patients with abnormal gastric emptying. Thus, it is unknown whether the treatment response with cisapride in functional dyspepsia patients is linked to abnormal gastric emptying. For domperidone, the study by Duan et al. (41) suggests that in patients with delayed gastric emptying, there is a correlation between improvement in symptoms and gastric emptying time; but, clearly, meaningful conclusions cannot be derived from a single small study. In reviewing studies using gastric emptying it is important to keep in mind that although tests within a single center are usually standardized, marked variation may exist among different centers in the methods and ways that results of gastric emptying studies are reported (48). Conclusions This meta-analysis suggests that the prokinetic agents cisapride and domperidone are superior to placebo in improving functional dyspepsia symptoms. However, most of the studies included were limited by weaknesses in design and/or reporting of results, so the results must therefore be viewed

Cisapride and Domperidone in Functional Dyspepsia

695

with caution. Furthermore, the results are mainly based on a global assessment by the investigator as the outcome measure, which, in all likelihood, is not the optimal method for evaluating changes in everity of symptoms. A meta-analysis cannot replace methodologically sound large clinical trials, which are necessary to confirm the present results. Given the recent concern about potentially serious cardiac side effects, the physician must carefully weigh the benefits and risks before prescribing cisapride.

ACKNOWLEDGMENTS This work was supported in part by an unrestricted grant from Abbott Laboratories. S.J.O. Veldhuyzen van Zanten is the recipient of a Nova Scotia Clinical Scientist Award. Reprint requests and correspondence: S.J.O. Veldhuyzen van Zanten, Queen Elizabeth II Health Sciences Center, Victoria General Hospital Site, Room 928, Centennial Building, 1278 Tower Road, Halifax, NS B3H 2Y9, Canada. Received Apr. 28, 2000; accepted Oct. 5, 2000.

REFERENCES 1. McCallum R, Prakash C, Campoli-Richards DM, et al. Cisapride, a preliminary review of its pahrmadynamic and pharmacokinetic properties, and therapeutic use as a prokinetic agent in gastrointestinal motility disorders. Drugs 1988;36: 652– 81. 2. Van Nueten JM, Schuurkes JAJ. Pharmacodynamics of cisapride, a prokinetic agent with cholinergic properties. Digestion 1986;34:137– 41. 3. Jian R, Ducrot F, Piedelop C, et al. Measurement of gastric emptying in dyspeptic patients: Effect of a new gastrokinetic agent cisapride. Gut 1985;26:352– 8. 4. Horowittz M, Maddox A, Harding PE, et al. Effect of cisapride on gastric and esophageal emptying in insulin-dependent diabetes mellitus. Gastroenterology 1987;92:1899 –907. 5. Verlinden M. A role for gastrointestinal prokinetic agents in the treatment of reflux esophagitis. Aliment Pharmacol Ther 1989;3:113–31. 6. Baldi F, Bianchi Porro G, Dobrilla G, et al. Cisapride versus placebo in reflux esophagitis. A multicenter double-blind trial. J Clin Gastroenetrol 1988;10:614 – 8. 7. Brogden RN, Carmine AA, Heel RC, et al. Domperidone. A review of its pharmacological activity, pharmacokinetics, and therapeutic efficacy in the symptomatic treatment of chronic dyspepsia and as an emetic. Drugs 1982;24:360 – 400. 8. Weihrauch TR, Forster CF, Krieglstein J. Evaluation of the effect of domperidone on human oesophageal and gastroduodenal motility of intraluminal manometry. Postgrad Med J 1979;55(suppl 1):7–14. 9. Platteborse R, Hermans C, Loots W, et al. The effect of domperidone on pyloric activity in dog and man. Postgrad Med J 1979;55(suppl 1):15–9. 10. Horowittz M, Harding PE, Chatterton BE, et al. Acute and chronic effects of domperidone on gastric emptying in diabetic autonomic neuropathy. Dig Dis Sci 1985;30:10 –5. 11. Talley NJ, Phillips SF. Non-ulcer dyspepsia: Potential causes and pathophysiology. Ann Intern Med 1988;108:865–79. 12. Veldhuyzen van Zanten SJO, Cleary C, Talley NJ, et al. Drug treatment of functional dyspepsia: A systematic analysis of

696

13. 14.

15. 16. 17.

18. 19.

20. 21. 22. 23.

24. 25. 26. 27. 28. 29. 30.

31.

Veldhuyzen van Zanten et al.

trial methodology with recommendations for design of future trials. Am J Gastroenterol 1996;91:660 –73. UNGRAPH. Cambridge, UK: Biosoft. Al-Quorain A, Larbi EB, Al-Shedoki F. A double-blind, randomized, placebo-controlled trial of cisapride in Saudi Arabs with functional dyspepsia. Scand J Gastroenterol 1995;30: 531– 4. DerSimmonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177– 88. Egger M, Smith G, Phillips A. Meta-analysis: Beyond the grand mean ? Br Med J 1997;315:1610 – 4. Deruyttere M, Lepoutre L, Heylen H, et al. Cisapride in the management of chronic functional dyspepsia: A multicentre, double-blind, placebo-controlled study. Clin Ther 1987;10: 44 –51. Francois I, de Nutte N. Non-ulcer dyspepsia. Effect of the gastrointestinal prokinetic drug cisapride. Curr Ther Res 1987; 41:891–7. Goethals C, Van de Mierop L. Cisapride in the treatment of chronic functional dyspepsia. Results of a double-blind, placebo-controlled, crossover study. Curr Ther Res 1987;42: 261–7. Hannon R. Efficacy of cisapride in patients with non-ulcer dyspepsia. Curr Ther Res 1987;42:814 –22. Rosch W. Cisapride in non-ulcer dyspepsia. Results of a placebo-controlled trial. Scand J Gastroenterol 1987;22: 161– 4. De Nutte N, van Ganse W, Witterhulghe M, et al. Relief of epigastric pain in non-ulcer dyspepsia: Controlled trial of the promotility drug cisapride. Clin Ther 1989;11:6268. Hausken T, Berstad A. Cisapride treatment of patients with non-ulcer dyspepsia and erosive prepyloric changes. A double-blind, placebo-controlled trial. Scand J Gastroenterol 1992;27:213–7. Deruyttere M, Milo R, Creytens G, et al. Therapy of chronic functional dyspepsia: Multicentre crossover study of cisapride and placebo. Prog Med 1987;43:61– 8. Chung JM. Cisapride in chronic dyspepsia. Results of a double-blind, placebo-controlled trial. Scand J Gastroenterol 1993;28:11– 4. Kellow JE, Cowan H, Shuter B, et al. Efficacy of cisapride therapy in functional dyspepsia. Aliment Pharmacol Ther 1995;9:153– 60. Yeoh KG, Kang JY, Tay HH, et al. Disorders of the gastric mucosa. J Gastroenterol Hepatol 1997;12:13– 8. De Groot GH, de Both PSM. Cisapride in functional dyspepsia in general practice. A placebo-controlled, randomized, doubleblind study. Aliment Pharmacol Ther 1997;11:193–9. Champion MC, MacCanell KL, Thomson AB, et al. A doubleblind randomized study of cisapride in the treatment of nonulcer dyspepsia. Can J Gastroenterol 1997;11:127–34. Jian R, Ducrot F, Ruskone A, et al. Symptomatic, radionuclide, and therapeutic assessment of chronic idiopathic dyspepsia. A double-blind placebo-controlled evaluation of cisapride. Dig Dis Sci 1989;34:657– 63. Agorastos I, Akriviadis E, Goulis G. Effect of cisapride in non-ulcer dyspepsia: A placebo-controlled trial. Curr Ther Res 1991;49:870 –7.

AJG – Vol. 96, No. 3, 2001

32. Frazzoni M, Lonardo A, Grisendi A, et al. Are routine duodenal and antral biopsies useful in the management of functional dyspepsia? A diagnostic and therapeutic study. J Clin Gastroenterol 1993;17:101– 8. 33. Hansen JM, Bytzer P, Schaffalitzky de Muckadell OB. Placebo-controlled trial of cisapride, and nizatidine in unselected patients with functional dyspepsia. Am J Gastroenterol 1998; 93:368 –74. 34. Van Ganse W, Van Damme L, Van de Mierop L, et al. Chronic dyspepsia: Double-blind treatment with domperidone (R 33 812) or a placebo. A multicentre therapeutic evaluation. Curr Ther Res 1978;23:695–701. 35. Bekhti A, Rutgeerts L. Domperidone in the treatment of functional dyspepsia in patients with delayed gastric emptying. Postgrad Med J 1979;55(suppl 1):30 –2. 36. DeLoose F. Domperidone in chronic dyspepsia: A pilot open study and a multicentre general practice cross-over comparison with metoclopramide and placebo. Pharmatherapeutica 1979;2:140 – 6. 37. Englert W, Schlich D. A double-blind crossover trial of domperidone in chronic postprandial dyspepsia. Postgrad Med J 1979;55(suppl 1):28 –9. 38. Haarmann K, Lebkuchner F, Widmann A, et al. A doubleblind study of domperidone in the symptomatic treatment of chronic post-prandial upper gastrointestinal distress. Postgrad Med J 1979;55(suppl 1):24 –7. 39. Van de Mierop L, Rutgeerts L, Van de Langenbergh B, et al. Oral domperidone in chronic postprandial dyspepsia. A double-blind placebo-controlled evaluation. Digestion 1979;19: 244 –50. 40. Sarin SK, Sharma P, Chawla YK, et al. Clinical trial on the effect of domperidone on non-ulcer dyspepsia. Indian J Med Res 1986;83:623– 8. 41. Duan LP, Zheng ZT, Li YN. A study of gastric emptying in non-ulcer dyspepsia using a new ultrasonographic method. Scand J Gastroenterol 1993;28:335– 60. 42. Finney JS, Kinnersley N, Hughes M, et al. Meta-analysis of antisecretory and gastrokinetic compounds in functional dyspepsia. J Clin Gastroenterol 1998;26:312–2. 43. Vitola J, Vukanovic J, Roden DM. Cisapride-induced torsade de pointes. J Cardiovasc Electrophysiol 1998;9:1109 –13. 44. Canadian Adverse Drug Reaction. Cisapride: Interaction with grapefruit juice and drugs. 2000;10:1–2. 45. Veldhuyzen van Zanten SJO, Talley NJ, Bytzer P, et al. Design of treatment trials for functional gastrointestinal disorders. Gut 1999;45(suppl II):1169 –77. 46. Moher D, Pham B, Jones A, et al. Does quality of reports of randomized trials affect estimates of intervention efficacy reported in meta-analyses? Lancet 1998;352:609 –13. 47. Heyse PM, Rambaldo R, Hazelhoff B. Factors affecting short and long term outcome of a short therapeutic trial with cisapride in dyspeptic patients. Scand J Gastroenterol 1993; 28(suppl 195):15–24. 48. House A, Champion MC, Chamberlain M. National survey of radionuclide gastric emptying studies. Can J Gastroenterol 1997;11:317–21.