Gastroenterology
Volume 137, Issue 3 , Pages 766-769, September 2009

Developing Irritable Bowel Syndrome Guidelines Through Meta-analyses: Does the Emperor Really Have New Clothes?

  • Michael Camilleri, MD (Guest Associate Editor)

      Affiliations

    • CENTER Program, College of Medicine, Mayo Clinic, Rochester, Minnesota
    • ,
  • Emeran A. Mayer, MD (Associate Editor)

      Affiliations

    • Center for Neurobiology of Stress, David Geffen School of Medicine at UCLA, Los Angeles, California

published online 29 July 2009.

Article Outline

 

Clinical guidelines are often based on systematic reviews and meta-analyses. There have been many such meta-analyses regarding the treatment of irritable bowel syndrome (IBS), even though individual high quality studies used in these meta-analyses are sparse. The American College of Gastroenterology (ACG) IBS Task Force1 recently conducted a comprehensive, methodologically sound appraisal of all therapeutic modalities used in IBS based on meta-analyses of each therapeutic class, and published an evidence-based position statement on the management of IBS. The significant effort in systematically analyzing the evidence is enhanced by the interpretation by several experts forming the task force under the auspices of a national organization. The report has a carefully worded abstract that identifies pitfalls. For example, the text specifies:

1.“Trials suggest psyllium fiber, certain antispasmodics, and peppermint oil are effective in IBS patients … although the quality of the evidence is poor.”

2.“Some probiotics may be effective in reducing overall IBS symptoms, … but more data are needed.”

3.“Tricyclic antidepressants and selective serotonin reuptake inhibitors have been shown to be effective in IBS patients of all subtypes … trials generally are of good quality but the limited number of patients included in trials. …”

4.“Psychological therapies may also provide benefit to IBS patients although the quality of evidence is poor.”

This comes in an era when meta-analyses in the area of IBS seem to be published more frequently than original articles on novel therapies, and treatment recommendations are sometimes based on the meta-analyses of published meta-analyses!

However, we perceive that meta-analyses of trials are only really useful if each included trial actually meets widely accepted criteria for a high-quality study, in particular adequate sample size and inclusion of a placebo arm, or an active comparator arm if the treatment is likely to unblind the study, as with centrally acting agents. Quality of trials based upon the Jadad et al2 or other criteria typically address process rather than content. Although forest plots (graphical displays that illustrate the relative strength of treatment effects in multiple studies that address the stated medication class for the disease of interest) depict the weight of trials conducted, they also partly reassure the reader about the sample size and weighting of the odds ratio calculated. However, this weighting discounts the quality of the trial design, which (like beauty) is often in the eye of the reviewer(s) of the published data who are assigned the task to include or reject the paper from subsequent analysis. For example, meta-analysis of poorly designed studies should never be used to make any treatment recommendations, and this was the recommendation of Klein3 in the original analysis of IBS treatment trials.

The variations in the standards and criteria used for inclusion of trials in a meta-analysis are illustrated by the recent publications of meta-analyses of probiotics in IBS. Brenner et al4 identified a total of 16 randomized, controlled trials (RCTs) that met selection criteria, but 11 studies showed suboptimal study design with inadequate blinding, inadequate trial length, inadequate sample size, and/or lack of intention-to-treat analysis, and none of the studies provided quantifiable data about both tolerability and adverse events. These authors conclude there are inadequate data to comment on the efficacy of probiotics other than B. infantis (for which there are 2 evaluable trials, 1 of which exposed only 24 patients to that specific probiotic). Conversely, the ACG task force (which included some of the same authors) and the separate publication emanating from the task force5 identified 19 RCTs, whose quality was declared generally good and provided odds ratios for dichotomous data or continuous data for the whole class of probiotics. So the reader might ask, “Which meta-analysis is correct? Whose advice should I accept?” It becomes even more complicated when the reader finds 3 or more other meta-analyses of probiotics in IBS that have been published in the past 2 years!

Back to Article Outline

Should Treatment of Patients With IBS Be Altered as a Result of These Meta-analyses? 

The conclusions of meta-analyses reflect the studies included, the degree of homogeneity or heterogeneity of the trials, the size and results of each individual trial, the specific dose of drug used, and the potential of publication bias. Thus, before assimilating these results into treatment guidelines or into practice, it is important to realize that the relative strengths of the different analyses, and hence the evidence for each summary (eg, number needed to treat [NNT]), are not the same across the different classes of therapies. The authors of the systematic analyses have provided important details regarding trial heterogeneity and publication bias for each class of agent. Considerable or statistically significant heterogeneity is reported for several classes of IBS treatments. It is important that this detailed information should not be lost in any summary guidelines that may be forthcoming, or in the information that will be communicated or remembered by busy clinicians. It is often useful to examine individual studies selected in these studies, especially when their relative risk (RR) seems to be out of synchrony with the majority of studies reported for a specific drug class.

Acting Drugs: Tricyclics, Selective Serotonin Reuptake Inhibitors, and Noradrenalin Serotonin Reuptake Inhibitors 

The report suggests there is an overall clinical benefit with different forms of antidepressants and psychological therapies. However, the analysis shows considerable heterogeneity of studies of centrally acting drugs, inconsistent end points used across studies, and small sample sizes of individual studies. Interestingly, a lack of efficacy in individual, appropriately sized, high-quality studies (eg, the intent to treat analysis with the Drossman et al6 study of desipramine) contrasts with the efficacy shown with smaller studies or summarized in a meta-analysis of small studies.7

The analysis of the antidepressants7 included 13 studies comparing antidepressants with placebo in the treatment of IBS: There was a total of 789 patients, 432 who received active therapy and 357 placebo. The studies were conducted in secondary (n = 7) or tertiary (n = 6) care centers, which raises questions about the generalizability of the results. A more recent meta-analysis of tricyclic agents in 7 trials8 identified an overall benefit for IBS symptoms and for abdominal pain (latter based on 2 studies); it is intriguing to note that the largest, best designed trial with a tricyclic agent6 was excluded by the authors because the authors concluded8 that Drossman et al included patients with functional GI disorders other than IBS.

The ACG report1, 7 suggests an RR of IBS symptoms persisting or remaining unimproved after treatment with antidepressant therapy was significantly lower compared with placebo (RR, 0.66; 95% confidence interval [CI], 0.57–0.78). The analysis revealed considerable, although statistically insignificant heterogeneity among studies (I2 = 26.4%; P = .17) and Funnel plot asymmetry (P = .02) suggested publication bias. Despite these “red flags,” the authors proceeded to calculate the NNT with antidepressant therapy to prevent IBS symptoms persisting was 4 (95% CI, 3–6).

In the forest plot summarizing the psychotropic medications, the studies by Vahedi et al9, 10 with amitriptyline and fluoxetine demonstrate RR values that favor treatment over placebo. It is important, however, to note that in these 2 studies from Iran, the 44 patient study of fluoxetine is characterized by a <10% response rate to placebo, which is quite atypical. Moreover, in the study of 10 mg amitriptyline versus placebo, it is relevant to note that amitriptyline is not effective on the primary end point of pain and the significant change is observed relative to baseline in the trial of 54 patients. The comparison of number of participants reporting posttreatment pain in amitriptyline and placebo groups was not significant. It is also interesting to note the report of complete responses (ie, absolutely no IBS symptoms) of 63% in the 10 mg amitriptyline group and 28% in the placebo group; the complete response rate is so atypical relative to experience in clinical practice in Western countries that it begs the question whether such meta-analyses are compromised by societal differences, for which there is no easy correction.

Overall, the ACG IBS Task Force analysis comes to a different conclusion on the efficacy of antidepressants from a prior robust analysis11 that separated studies according to end points: a dichotomous or continuous outcome for relief of abdominal pain or dichotomous global assessment of improvement. For all groupings, antidepressants were not significantly better than placebo in the Cochrane review.11 There are other considerations in IBS patients that renders difficult the interpretation of trials and meta-analyses.

First, there is well-known, high sensitivity of IBS patients to any side effects, and it is considered difficult if not impossible to blind a study with either one of these drugs, especially the tricyclic agents, which cause anticholinergic side effects,12 Unless there is an active comparator with similar side effects, it is nearly impossible to blind the studies, and hence to claim that differences from placebo are not the result on unblinding owing to the adverse effects. Second, clinical trials do not typically exclude patients with anxiety and depression assessed by a structured psychiatric interview; hence, it cannot be ascertained whether the benefit in IBS is related to improvement of the psychiatric comorbidity rather than the IBS itself. Third, trials are usually not sufficiently powered to allow for a subgroup analysis based on responders to this class of medication. Thus, it remains unclear whether clinical practice or guidelines could be enhanced by evidence-based predictors of response.

Recently published studies on cognitive behavioral treatments have provided robust data on the efficacy of this treatment approach,13 but most of the data on which the meta-analysis is based are small, and do not meet quality criteria to justify inclusion. Indeed, a recent Cochrane analysis concluded that psychological interventions may be slightly superior to usual care or waiting list control conditions at the end of treatment, and the clinical significance of this “superiority” was considered debatable,14 the sustainability of the effect was questionable, and the meta-analysis was significantly limited by issues of validity, heterogeneity, small sample size, and outcome definition.14 In fact, the efficacy was substantially different7 in the 9 trials from 1 group (Blanchard et al; n = 270; RR, 0.59) relative to 11 studies (1,008 patients in all other groups' studies) conducted in other centers (RR, 0.71), although a significant difference was not detected (P = .35). Incidentally, both groups of trials had significant heterogeneity, and there was evidence of significant publication bias.7 In addition, the efficacy of psychotherapies for the bowel dysfunction of IBS is not routinely addressed in such studies to determine whether the efficacy applies to all symptoms of IBS.

In summary, we are not implying that pharmacologic or nonpharmacologic therapeutic approaches aimed at the central nervous system do not work; however, we believe strongly that treatment recommendations should not be based on meta-analyses of studies that are of insufficient quality, but rather on the few available studies that meet current quality standards (eg, Drossman et al6 and Lackner et al13).

Antispasmodics and Muscle Relaxants 

The ACG task force concluded that fiber, antispasmodics, and peppermint oil were all more effective than placebo in the treatment of IBS.15 We focus on antispasmodics. The conclusion of the ACG IBS Task Force meta-analyses suggesting potential for drugs such as antispasmodics and peppermint oil to relieve IBS symptoms, confirmed results of previous meta-analyses of Poynard et al16, 17 and Quartero et al.11 For example, in Poynard's analyses, the antispasmodics cimetropium bromide, pinaverium bromide, trimebutine, octilonium bromide, and mebeverine were reported to be efficacious.6, 16 The conclusions on peppermint oil are based on 4 studies (1:1 peppermint oil to placebo) in a total of only 392 patients; statistically significant heterogeneity was detected among the studies.

Probiotics 

In general, there is increasing evidence that probiotics help to relieve some IBS symptoms. There are at least 5 meta-analyses published recently based upon pooling of 8–14 trials4, 5, 18, 19, 20: They all acknowledge that the individual trials varied in relation to the length of treatment (range, 4–26 weeks), dose, organisms, and strengths of probiotics used. Given the different biological effects that are attributed to different probiotic species and strains (eg, effects on local immune function, barrier function, permeability, antibacterial effects, binding on endogenous irritants like bile acids, and other unknown effects), it seems inappropriate to group all probiotics as though they were 1 class of agents.

The ACG IBS Task Force1, 5 specifically identified the main limitation that there were a variety of species, strains, and doses of probiotics used, and initially stated, “It was difficult to reach a conclusion about the optimal probiotic strategy to use in patients with IBS… . The dichotomous data suggest that all probiotic therapies show a trend for being efficacious in IBS. In contrast, the continuous data suggest that Lactobacilli have no impact on symptoms, probiotic combinations improve symptoms in IBS patients, and there was a trend for Bifidobacteria to improve IBS symptoms.” The section concludes that because almost all probiotic combinations contained both Bifidobacteria and Lactobacilli, and the latter did not have an effect in the continuous data meta-analysis, it is possible that Bifidobacteria are the active agent in probiotic combinations. We do not concur with such extrapolations, and perceive that more robust clinical trials comparing different single and combination probiotic preparations are needed if such statements are to be evidence based.

The meta-analysis by McFarland and Dublin19 concluded that probiotic use was associated with improvement in global IBS symptoms compared with placebo (pooled RR, 0.77; 95% CI, 0.62–0.94), and with less abdominal pain compared with placebo (RR pooled, 0.78; 95% CI, 0.69–0.88). They also determined that too few studies reported data on other IBS symptoms or on specific probiotic strains to allow estimation of a pooled RR. Similarly, the meta-analysis by Brenner et al4 questions the conclusion regarding the class of probiotics and considers there is insufficient evidence about the efficacy of probiotics in IBS.

Serotonergic Agents 

The conclusions reached by the ACG Task Force regarding 5-HT3 antagonists (alosetron and cilansetron and 5-HT4 agonists [eg, tegaserod] are generally consistent with other published meta-analyses in the literature21, 22, 23). These studies are individually characterized by sound methodology in terms of design, sample size, uniformity of conduct, enrichment of study population based on symptom scores during a baseline, no-treatment observation period, and data acquisition. Therefore, the conclusions regarding these classes of medications seem to be appropriate. Not surprisingly, the NNT required for treatment of IBS with these serotonergic medications, based on large, robust trials involving several thousands of patients, is generally 2–3 times greater than the NNT for medications such as peppermint oil or antidepressants estimated from suboptimal individual studies!

Back to Article Outline

Conclusion 

Better designed individual studies of all these forms of therapies would provide more robust information from which clinicians and patients can derive more predictable information on what to expect from these therapies. The sample size of patients from all the trials in many classes of drugs obtained over >20 years (eg, for centrally acting medications or antispasmodics) is often equal to the sample size from single multicenter drug trials in IBS with the 5-HT3 antagonists and 5-HT4 agonists in the last decade. Therefore, one should question the generalizability of the information and any future guidance proposed on these classes of drugs. The vast majority of studies included in the meta-analyses have insufficient sample sizes and other methodologic flaws that should exclude them from analysis, except for the drug classes of 5-HT3 antagonists and 5-HT4 agonists. This analysis of the literature strongly suggests that only the more recent trials (including CBT) meet quality criteria, and recommendations should be based exclusively on these studies, and should not be mixed, despite the attraction of meta-analyses, with older, low-quality trials.

Physicians should become more conversant with methodology, strengths, pitfalls, and interpretations of meta-analyses, especially the claimed NNT. The documentation of significant heterogeneity and publication bias within such systematic reviews and meta-analyses should alert the physician to take the recommendations related to that analysis with a large grain of salt. This applies for several classes of therapies, including probiotics, psychotropic medications, and antispasmodics, to name a few. Conversely, recommendations based on meta-analytical data from large uniform clinical trials (eg, with serotonergic agents) provide realistic and practical information that should be viewed favorably by practicing physicians. It is a paradox that many of the agents in this category are not available for prescription, because significant adverse effects have been associated with them! There remains significant unmet need in IBS, and one should continue to question the generalizability of the information and any future guidance proposed on the classes of drugs that is based on the recent systematic review.1

Back to Article Outline

References 

  1. Brandt LJ, Chey WD, Foxx-Orenstein AE, et al. An evidence-based systematic review on the management of irritable bowel syndrome. Am J Gastroenterol. 2009;10:S1–S35
  2. Jadad AR, Moore RA, Carroll D, et al. Assessing the quality of reports of randomized clinical trials: is blinding necessary?. Control Clin Trials. 1996;17:1–12
  3. Klein KB. Controlled treatment trials in the irritable bowel syndrome: a critique. Gastroenterology. 1988;95:232–241
  4. Brenner DM, Moeller MJ, Chey WD, et al. The utility of probiotics in the treatment of irritable bowel syndrome: a systematic review. Am J Gastroenterol. 2009;104:1033–1049
  5. Moayyedi P, Ford AC, Talley NJ, et al. The efficacy of probiotics in the therapy of irritable bowel syndrome: a systematic review. Gut. 2008 Dec 17;[Epub ahead of print]
  6. Drossman DA, Toner BB, Whitehead WE, et al. Cognitive-behavioral therapy versus education and desipramine versus placebo for moderate to severe functional bowel disorders. Gastroenterology. 2003;125:19–31
  7. Ford AC, Talley NJ, Schoenfeld PS, et al. Efficacy of antidepressants and psychological therapies in irritable bowel syndrome: systematic review and meta-analysis. Gut. 2009;58:367–378
  8. Rahimi R, Nikfar S, Rezaie A, et al. Efficacy of tricyclic antidepressants in irritable bowel syndrome: a meta-analysis. World J Gastroenterol. 2009;15:1548–1553
  9. Vahedi H, Merat S, Rashidioon A, et al. The effect of fluoxetine in patients with pain and constipation-predominant irritable bowel syndrome: a double-blind randomized-controlled study. Aliment Pharmacol Ther. 2005;22:381–385
  10. Vahedi H, Merat S, Momtahen S, et al. Clinical trial: the effect of amitriptyline in patients with diarrhoea-predominant irritable bowel syndrome. Aliment Pharmacol Ther. 2008;27:678–684
  11. Quartero AO, Meineche-Schmidt V, Muris J, et al. Bulking agents, antispasmodic and antidepressant medication for the treatment of irritable bowel syndrome. Cochrane Database Syst Rev. 2005;2:CD003460
  12. Thiwan S, Drossman DA, Morris CB, et al. Not all side effects associated with tricyclic antidepressant therapy are true side effects. Clin Gastroenterol Hepatol. 2009;7:446–451
  13. Lackner JM, Jaccard J, Krasner SS, et al. Self-administered cognitive behavior therapy for moderate to severe irritable bowel syndrome: clinical efficacy, tolerability, feasibility. Clin Gastroenterol Hepatol. 2008;6:899–906
  14. Zijdenbos IL, de Wit NJ, van der Heijden GJ, et al. Psychological treatments for the management of irritable bowel syndrome. Cochrane Database Syst Rev. 2009;1:CD006442
  15. Ford AC, Talley NJ, Spiegel BM, et al. Effect of fibre, antispasmodics, and peppermint oil in the treatment of irritable bowel syndrome: systematic review and meta-analysis. BMJ. 2008;337:a2313
  16. Poynard T, Regimbeau C, Benhamou Y. Meta-analysis of smooth muscle relaxants in the treatment of irritable bowel syndrome. Aliment Pharmacol Ther. 2001;15:355–361
  17. Poynard T, Naveau S, Mory B, et al. Meta-analysis of smooth muscle relaxants in the treatment of irritable bowel syndrome. Aliment Pharmacol Ther. 1994;8:499–510
  18. Hoveyda N, Heneghan C, Mahtani KR, et al. A systematic review and meta-analysis: probiotics in the treatment of irritable bowel syndrome. BMC Gastroenterol. 2009;9:15
  19. McFarland LV, Dublin S. Meta-analysis of probiotics for the treatment of irritable bowel syndrome. World J Gastroenterol. 2008;14:2650–2661
  20. Nikfar S, Rahimi R, Rahimi F, et al. Efficacy of probiotics in irritable bowel syndrome: a meta-analysis of randomized, controlled trials. Dis Colon Rectum. 2008;51:1775–1780
  21. Andresen V, Montori VM, Keller J, et al. Effects of 5-hydroxytryptamine (serotonin) type 3 antagonists on symptom relief and constipation in nonconstipated irritable bowel syndrome: a systematic review and meta-analysis of randomized controlled trials. Clin Gastroenterol Hepatol. 2008;6:545–555
  22. Evans BW, Clark WK, Moore DJ, et al. Tegaserod for the treatment of irritable bowel syndrome and chronic constipation. Cochrane Database Syst Rev. 2007;4:CD003960
  23. Lesbros-Pantoflickova D, Michetti P, Fried M, et al. Meta-analysis: the treatment of irritable bowel syndrome. Aliment Pharmacol Ther. 2004;20:1253–1269

PII: S0016-5085(09)01186-X

doi:10.1053/j.gastro.2009.07.030

Gastroenterology
Volume 137, Issue 3 , Pages 766-769, September 2009