Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Probiotics for antibiotic-associated diarrhoea

Finding more information
New results

Most of us would probably agree that meta-analysis is a good thing, but not when we cannot use the results or we fail to reach a satisfying level of comfort in using its results. Bandolier was delighted when it saw a meta-analysis on the use of probiotics for antibiotic-associated diarrhoea [1]. There were seven randomised trials, but like so many meta-analyses all it told us was a lot of statistical guff, with the fact that the combined effect relative risk was 0.3966 with a 95% confidence interval of 0.275 to 0.571.

Not only does the over-precision of these figures worry us (who really worries about the fourth significant figure of the relative risk in Bridgend on a wet Wednesday afternoon?), but what use is a relative risk without some anchor? What is the incidence of diarrhoea with antibiotics? What is the rate of diarrhoea without probiotics?

So Bandolier decided to use this as a test case to see how much work would be needed to turn a good meta-analysis without utility into one that could be used.


MEDLINE and Cochrane registers, and abstracts of all major gastroenterology congresses over 15 years, were searched for studies performed since 1991. Studies had to be randomised, in any language, for the treatment of antibiotic-associated diarrhoea with Lactobacillus or Saccharomyces boulardii. Studies had to have a placebo group, have diarrhoea as a primary end point, and have a minimum of two weeks of follow up. The definition of diarrhoea was as presented in studies, except that difference in daily stool discharge was not accepted.


There were seven studies with 881 patients. No information was given on reporting quality, allocation concealment, or blinding. Two studies were in children, and the rest in adults, with one limited to elderly patients. Three studies showed a significant reduction in diarrhoea, and the overall relative risk was 0.4 (0.3 to 0.6).

We were not told what proportion of patients had diarrhoea with placebo, or with probiotic, neither overall nor for individual studies.

Finding more information

Bandolier used PubMed to obtain the abstract of each paper in the meta-analysis. One was available in full text. A brief search was conducted, again using PubMed, for any other papers that may have been published subsequently.

Two were found, one (full text) from the Italian group publishing the meta-analysis on Helicobacter eradication, and one (abstract) from the USA on patients in hospital given antibiotics. All abstracts and papers had information directly or implicitly giving the numbers of patients with diarrhoea, and in total.

New results

The rates of diarrhoea with probiotic and with placebo for nine individual trials are shown in Figure 1, where open circles represent Lactobacillus and filled circles Saccharomyces.

Figure 1: Antibiotic-associated diarrhoea with and without probiotics. Open circles used Lactobacillus and filled circles used Saccharomyces

The overall rates and NNTs for the original and extended meta-analyses, and for Lactobacillus and Saccharomyces only are shown in Table 1. The relative risk for diarrhoea with probiotic was 0.4 (0.3 to 0.6) for the original meta-analysis and 0.6 (0.4 to 0.7) for the extended analysis with nine trials.

Table 1: Results of meta-analysis for all studies, those in the original meta-analysis, and with Lactobacillus or Saccharomyces only



Diarrhoea/total (%) with





(95% CI)

All studies
10 (7.1 to 17)
Studies in meta-analysis
8.4 (6.1 to 14)
Lactobacillus only
8.9 (6.0 to 17)
Saccharomyces only
14 (7 to 108)

Using all the information, antibiotic-associated diarrhoea occurred in 23% of patients without probiotic and in 13% of patients with probiotic. For every 10 patients using daily probiotic with antibiotics, one fewer will have diarrhoea.


Most people find this an interesting topic. Hardly anyone found a relative risk useful or illuminating. Tell them that the incidence of diarrhoea falls from 23% to 13%, or the NNT is 10, or preferably both, and everyone knows what the results are. Rocket science this is not, but reporting results without utility is not clever and is without value. The good news is that an IT strategy embracing a cheap computer and web browser software can remedy all that.


  1. F Cremonini et al. Meta-analysis: the effect of probiotic administration on antibiotic-associated diarrhoea. Alimentary Pharmacology and Therapeutics 2002 16: 1461-1467.
previous or next story in this issue