Original Articles
Problems and challenges in the design of irritable bowel syndrome clinical trials: experience from published trials

https://doi.org/10.1016/S0002-9343(99)00086-8Get rights and content

Abstract

The last two decades have seen many studies that are of inadequate design and power. This report focuses on what we have learned from the 25 randomized, controlled studies that included at least 30 patients during the period 1976–1998. The most important finding has been that the median placebo response was 47% (range, 0–84%), which is approximately three times the size of the difference between placebo and drug response, median 16% (range, −17–64%). This tells us the importance of reassurance and the powerful nonspecific therapeutic effects of entering patients into clinical trials in irritable bowel syndrome (IBS). Patients should be stratified according to the dominant symptoms that are relevant to the drug’s intended effect. A randomized, double-blind, controlled, parallel group study appears the most robust design. Minimizing the placebo response reduces the numbers needed to detect a significant difference. The optimum length of trial is probably >3 months, because the placebo effect takes approximately 12 weeks to start to recede. Dose titration should maximize the chance of detecting a benefit.

Section snippets

What have we learned from 1976 to 1998?

This section is based on an analysis of 25 randomized controlled trials published during this period that have included >30 patients and excludes many more with either smaller numbers or inadequate design. These 25 trials studied the efficacy of antispasmodics (13), bran (4), antidepressants (3), loperamide (3), cisapride (2), and isphagula (1).

Patient definition

Early studies used diagnostic criteria for IBS that varied from study to study. Whereas most required that patients should have normal barium studies and in some cases blood tests, few specified symptoms or their frequency. Although variable bowel habit was usually required, this was not quantified. More importantly, pain was not always included in the definition, which was effectively one of exclusion. In 1990 a working party produced the “Rome criteria,”2 which defined the syndrome more

Specific endpoints

Global assessments of severity relate poorly to specific symptoms and more to quality-of-life issues, such as how symptoms interfere with patients’ lives.7 Using global response as an endpoint has a logic, because this is what the patients want to improve, and may be more informative for evaluating centrally acting drugs than specific changes in bowel habit. However, global response measures miss some specific effects. Only a panacea or a centrally acting psychotropic drug would treat all

Placebo response

As already mentioned, the placebo effect is much larger than any drug effect, so we need to understand and control for this or else it will overshadow any specific benefit of the drug. Several studies have documented in detail the placebo-related decrease in anxiety, which occurs steadily from trial entry for the next 6 to 8 weeks.15 Others have documented the decrease in anxiety after being told an investigation is normal.17 For patients with low initial anxiety, there is a substantial and

Effects of lengthening trial

On current evidence, as presented above, the ideal of a low placebo response (<20%) may be achieved by running the trial for ≥3 months. Such a trial may require significantly fewer patients, but would it be easier to conduct? There is not much evidence on this point.

Minimal symptoms

The practice of excluding patients with minimal symptoms makes sense, because in clinical practice such patients should not receive drug therapy. Any trial using a large proportion of mild cases would be considered to have studied the wrong patient group. Furthermore, one would predict that the placebo response would be high, and thus the numbers required for a significant result would be very large. However, the threshold for entry needs to be realistic. Previous studies using a daily

Effect of placebo response on crossover design

Patients who receive active treatment in the first part of the study may well respond differently if a placebo response is still developing. Crossover designs reduce the numbers of patients needed to be recruited and are potentially more powerful, but are difficult to interpret if there is a big order effect. It is quite common to find a carry-over effect with, for example, laxatives, with which the improvement in constipation continues for some time. This means that an active drug taken first

Dose titration

The idea of adjusting the dose according to the patient’s response makes a great deal of sense, because this allows one to take into account different body weights and drug metabolism. It also allows one to use the largest dose tolerable, thus ensuring an optimal balance between therapeutic and associated but undesirable effects. Some studies have used this approach. Thus, Cann et al9 allowed patients to increase or decrease the dose of loperamide or placebo according to the response. They

Side effects

It is important that these are systematically reported, both because side effects may limit subsequent use of the treatment outside a clinical trial and also because they effectively unblind the trial, giving rise to the risk of bias in both the patient and investigator. One study, using the tricyclic antidepressant desimipramine, used atropine as a control, in addition to placebo, because the dry-mouth side effect of desipramine could not be controlled in any other way.29 The atropine had no

Documenting recruitment and drop out

Keeping a record of what proportion of patients encountered in clinical practice were considered suitable for recruitment and how many dropped out has been done in very few studies, which are written as if this never occurred. This is vital, because most authors would like to believe that the results of their trial will be applicable to the entire population of patients seen with IBS. If, however, only a select group of patients agree to enter, then this may not be valid. Recording the

Statistics and power

The use of appropriate statistics and ensuring that enough subjects are included to ensure adequate power to answer definitively the question under study is accepted by all. Many studies in the literature used far too few patients to show an effect unless this was extremely large. A recent review of published studies of smooth muscle relaxants16 reported the median number of patients studied to be 29 (range, 8–178) with a mean proportion of patients improving on the drug of 0.35 and on placebo

Conclusions

The large placebo response and relative ineffectiveness of therapy tell us a lot about IBS. Previous studies have, in retrospect, been poorly designed and analyzed. Patients should be stratified according to their dominant symptoms, because many drugs with specific effects are likely to benefit one group specifically and may actually aggravate some patients’ symptoms. Trials with newer drugs acting more on perception or coping behavior may require stratification by psychological parameters

References (30)

  • K.B. Klein

    Controlled treatment trials in the irritable bowel syndromea critique

    Gastroenterology

    (1988)
  • J. Svedlund et al.

    Controlled study of psychotherapy in irritable bowel syndrome

    Lancet

    (1983)
  • D.A. Drossman et al.

    Identification of sub-groups of functional gastrointestinal disorders

    Gastroenterol Int

    (1990)
  • K. Schutze et al.

    Double-blind study of the effect of cisapride on constipation and abdominal discomfort as components of the irritable bowel syndrome

    Aliment Pharmacol Ther

    (1997)
  • N.A. Chaudhary et al.

    The irritable colon syndrome

    Q J Med

    (1962)
  • J.E. Kellow et al.

    Effects of acute psychologic stress on small-intestinal motility in health and the irritable bowel syndrome

    Scand J Gastroenterol

    (1992)
  • M.A. Kamm et al.

    Steroid hormone abnormalities in women with severe idiopathic constipation

    Gut

    (1991)
  • B.A. Hahn et al.

    Patient-perceived severity of irritable bowel syndrome in relation to symptoms, health resource utilization and quality of life

    Aliment Pharmacol Ther

    (1997)
  • N. Hovdenak

    Loperamide treatment of the irritable bowel syndrome

    Scand J Gastroenterol

    (1987)
  • P.A. Cann et al.

    Role of loperamide and placebo in management of irritable bowel syndrome (IBS)

    Dig Dis Sci

    (1984)
  • P.A. Cann et al.

    What is the benefit of coarse wheat bran in patients with irritable bowel syndrome?

    Gut

    (1984)
  • A. Prior et al.

    Double blind study of ispaghula in irritable bowel syndrome

    Gut

    (1987)
  • J. Snook et al.

    Bran supplementation in the treatment of irritable bowel syndrome

    Aliment Pharmacol Ther

    (1994)
  • P.S. Efskind et al.

    A double-blind placebo-controlled trial with loperamide in irritable bowel syndrome

    Scand J Gastroenterol

    (1996)
  • E.M. Quigley et al.

    Ambulatory intestinal manometrya consensus report on its clinical role

    Dig Dis Sci

    (1997)
  • Cited by (0)

    View full text