• abdominal pain
• irritable bowel syndrome

We read the randomised controlled trial (RCT) by Zhou et al. 1 The authors recruited a highly selected group of patients with increased intestinal permeability and postinfection IBS with diarrhoea (IBS-D), defined by an unconfirmed enteric infection in the 12 months prior to symptom onset, and randomised them to 8 weeks of treatment with either glutamine or placebo. The population screened also had other curious features. For example, only four had rapid orocaecal transit on lactulose hydrogen breath testing, erroneously described as small intestinal bacterial overgrowth. This appears low, compared with other studies examining this issue in patients with IBS.

There are other minor issues with the trial conduct and analysis. First, compliance was checked by telephone calls, not a particularly robust method. Second, the efficacy analysis was per-protocol, rather than intention-to-treat, and minimal information on dropouts was provided, other than that they ‘voluntarily withdrew’. Finally, the extremely low adverse event rate in a population with IBS is also unusual. Did none of the patients have headaches, respiratory tract infection, or require antibiotic therapy, leading to withdrawal?

Glutamine reversed intestinal permeability and improved symptoms of IBS, according to the IBS Symptom Severity Score (IBS-SSS).2 The benefit of glutamine in this patient group was unprecedented, compared with all known therapies for IBS-D.3 The absolute difference in response rates between glutamine and placebo, defined as a decrease in IBS-SSS of ≥50 points, was 73.8% (p<0.0001), equating to a number needed to treat of <2 for glutamine. The mean change in IBS-SSS at 8 weeks was also vastly different between glutamine and placebo (−120.04 vs −5.58, p<0.0001). This is one of the most remarkable demonstrations of efficacy, relative to placebo, ever seen in an IBS treatment trial.

Placebo response rates in IBS are high, estimated at 37.5% in a previous meta-analysis of RCTs,4 although many of these trials used different endpoints to judge efficacy from that of Zhou et al.1 However, in another RCT of the probiotic Symprove, which used an identical endpoint, the placebo response rate was 41.2%, and the change in IBS-SSS at completion of therapy was −63.3 with Symprove, compared with −28.3 with placebo.5 The reasons for the very low placebo response rate in this trial, and the huge treatment effect of glutamine, are unclear. One possibility is that the placebo (whey protein) induced symptoms, exaggerating the effect of glutamine. Indeed, in a cross-over rechallenge RCT in patients with IBS reporting gluten sensitivity, whey protein was suspected as another possible culprit, although further study did not confirm this.6

The major issue, however, is the lack of a power calculation or sample size estimation, in both the manuscript and on ClinicalTrials.gov, a requirement for all RCTs.7 Trials in IBS-D usually recruit hundreds of patients to show a minimum clinically important difference in treatment effect of 10%–15%.8 9 This suggests that the study is too small, and has thrown up an extreme, inflated result erroneously. This is a less well-recognised problem of underpowered studies, which are more notorious for their tendency to lead to falsely negative results. In this alternative scenario, inflation occurs when, in order to claim success (discovery), an association has to pass a certain threshold of statistical significance, and the study has suboptimal power to make the discovery at the requested threshold of statistical significance.10

In conclusion, the remarkable results seen with glutamine led the trial authors to call for the conduct of larger RCTs. However, for the reasons summarised above, we suspect other investigators may find it difficult to replicate these results.