Skip navigation

Cisapride for non-ulcer dyspepsia

Search
Results
Comment
The clinical picture
Variation in placebo effects
Variation in cisapride effects
Use of open studies

Cisapride is a prokinetic drug that enhances gastrointestinal motility and increases the lower oesophageal sphincter tone, thereby minimising gastrooesophageal reflux. Gastric and duodenal emptying are enhanced because of increased gastric and duodenal contractility and coordination.

A systematic review [1] not only serves to help us understand the effectiveness of this compound, not much used in the UK, but also is a useful example of why we choose high quality studies for decision making.

Search


Six electronic databases were searched, including the Cochrane library, and manufacturers contacted about unpublished studies. Participants were adult patients with a diagnosis of non-ulcer dyspepsia. Oesophagitis, gastric or duodenal ulcer and gastric erosion as causes of dyspepsia were eliminated by endoscopy.

Studies included had to be randomised comparisons of cisapride with placebo or a control treatment. The main outcome was the number of patients experiencing either excellent, or excellent and good improvement of symptoms.

Results


There were 16 studies included, with 902 patients given cisapride and 664 patients given placebo. Two were open studies, and 14 were double blind. The dose of cisapride varied between 15 and 40 mg daily, but in most the dose was 20 mg three times a day. The duration of studies was two to six weeks, but most were over three or four weeks.

There were major differences between the two open studies and those that were double blind (Table). The outcome of excellent symptom improvement was obtained by 129/371 patients (35%) given cisapride and 54/377 (14%) of those given placebo in double blind studies (Figure 1). The NNT was 4.9 (95% confidence interval 3.8 to 6.9). This was significantly higher (worse) than the NNT of 2.0 for the two open studies and 2.7 for all studies combined.

Table: Results for cisapride in non ulcer dyspepsia according to outcome (excellent, or good or excellent symptom improvement) and by open and double blind design

    Excellent response with:    
Design Number of trials Cisapride (%) Placebo (%) Relative benefit (95% CI) NNT (95% CI)
Open, randomised 2 277/426 (65) 26/181 (15) 4.4 (3.1 to 6.3) 2.0 (1.7 to 2.3)
Double blind, randomised 11 129/371 (35) 54/377 (14) 2.4 (1.8 to 3.2) 4.9 (3.8 to 6.9)
All trials 13 406/797 (51) 80/558 (14) 3.2 (2.6 to 4.0) 2.7 (2.4 to 3.1)
    Good or excellent response with:    
Design Number of trials Cisapride (%) Placebo (%) Relative benefit (95% CI) NNT (95% CI)
Open, randomised 2 393/426 (92) 80/181 (44) 2.1 (1.8 to 2.5) 2.1 (1.8 to 2.5)
Double blind, randomised 14 323/476 (68) 207/483 (43) 1.6 (1.4 to 1.8) 4.0 (3.2 to 5.3)
All trials 16 716/902 (79) 287/664 (43) 1.8 (1.6 to 2.0) 2.8 (2.5 to 3.2)


Figure 1: Individual trials with cisapride with excellent symptom improvement. Shaded symbols are open studies.

 


The outcome of good or excellent symptom improvement was obtained by 323/476 patients (68%) given cisapride and 207/483 (43%) of those given placebo in double blind studies (Figure 2). The NNT was 4.0 (95% confidence interval 3.2 to 5.3). This was significantly higher (worse) than the NNT of 2.1 for the two open studies and 2.8 for all studies combined.

Figure 2: Individual trials with cisapride with good or excellent symptom improvement. Shaded symbols are open studies.


In addition, three studies compared cisapride 20 or 30 mg a day with ranitidine 300 mg a day, cimetidine 800 mg a day and nizatidine 300 mg a day. Combining these three trials, 196/257 (76%) patients had good or excellent response with cisapride compared with 169/269 (63%) with histamine antagonist. The NNT was 7.4 (4.7 to 18).

Comment


The clinical picture


This review could be taken as confirmation that cisapride is moderately effective in symptom relief over about four weeks in patients given a dose of about 30 mg a day. For every four or five patients treated with cisapride, one will have good and/or excellent symptom relief who would not have benefited with placebo. Thirty-five percent (95% confidence interval 30%-40%) of patients will have an excellent response and 68% (64%-72%) will have a good or excellent response.

The fact that cisapride was more effective than histamine antagonists is useful supporting data. There are problems. Non-ulcer dyspepsia was defined endoscopically, and this limits the usefulness of the results in primary care. We would benefit from better diagnostic criteria.

Variation in placebo effects


Bandolier 72 we looked at variation in endoscopic healing rates in reflux oesophagitis with placebo, but failed to make the link between the estimate of effect with placebo and the size of the sample. Figure 3 shows the placebo response for the excellent outcome in individual trials, and Figure 4 the placebo response for good or excellent outcome (but note that the X-axis is different for the two graphs).

Figure 3: Placebo response rates in individual trials with cisapride with excellent symptom improvement. Shaded symbols are open studies.

Figure 4: Placebo response rates in individual trials with cisapride with good or excellent symptom improvement. Shaded symbols are open studies.


There are two observations. More patients achieve the lower hurdle of good or excellent (mean 43%; 95% CI 37%-52%) than achieve the higher hurdle of an excellent response (mean 14%; 95% CI 11%-17%). Responses to placebo are the same for double blind and open studies (Table 1).

In both there is large variability at low sample size. For an excellent response the range is 7% to 27%, and for a good or excellent response it is 18% to 83%. Almost all of this variation is, as expected, with the smallest studies with sample sizes of 30 patients or fewer. Larger samples are less inaccurate.

Variation in cisapride effects


Figures 5 and 6 show the individual trial responses to cisapride for an excellent and a good or excellent response, respectively. Again there is considerable variation. But now the response for the open studies is considerably different from that for the double blind studies. Influenced mainly by one large open study, the cisapride response is much higher (see also Table 1).



Figure 5: Cisapride response rates in individual trials with cisapride with excellent symptom improvement. Shaded symbols are open studies.



Figure 6: Cisapride response rates in individual trials with cisapride with good or excellent symptom improvement. Shaded symbols are open studies.


Use of open studies


Bandolier has reported before that double blind studies produce lower treatment effects than open studies. Open studies are biased, and here we have a beautiful example of how a single, large, study can overwhelm several smaller studies with more robust design. One doesn't need statistics to show it, it is there in the graphs. The lesson keeps being learned: beware including studies of inadequate design.

Reference:

  1. VK Shukla, N Otten, C Dubé, D Moher. Use of cisapride in patients with non-ulcer dyspepsia: a meta-analysis of randomized trials. Ottawa: Canadian Coordinating Office for Health Technology Assessment (CCOHTA); 2000.
previous or next story in this issue