- Aloysius L D'Souza, research fellow (, )
- Chakravarthi Rajkumar, senior lecturer and honorary consultant,
- Jonathan Cooke, statistician,
- Christopher J Bulpitt, professor of geriatric medicine
- Care of the Elderly Section, Faculty of Medicine, Imperial College School of Medicine, Hammersmith Hospital, London W12 0NN
- Correspondence to: A L D'Souza
- Accepted 7 November 2001
Objective: To evaluate efficacy of probiotics in prevention and treatment of diarrhoea associated with the use of antibiotics.
Design: Meta-analysis; outcome data (proportion of patients not getting diarrhoea) were analysed, pooled, and compared to determine odds ratios in treated and control groups.
Identification: Studies identified by searching Medline between 1966 and 2000 and the Cochrane Library.
Studies reviewed Nine randomised, double blind, placebo controlled trials of probiotics.
Results: Two of the nine studies investigated the effects of probiotics in children. Four trials used a yeast (Saccharomyces boulardii), four used lactobacilli, and one used a strain of enterococcus that produced lactic acid. Three trials used a combination of probiotic strains of bacteria. In all nine trials, the probiotics were given in combination with antibiotics and the control groups received placebo and antibiotics. The odds ratio in favour of active treatment over placebo in preventing diarrhoea associated with antibiotics was 0.39 (95% confidence interval 0.25 to 0.62; P<0.001) for the yeast and 0.34 (0.19 to 0.61; P<0.01 for lactobacilli. The combined odds ratio was 0.37 (0.26 to 0.53; P<0.001) in favour of active treatment over placebo.
Conclusions: The meta-analysis suggests that probiotics can be used to prevent antibiotic associated diarrhoea and that S boulardii and lactobacilli have the potential to be used in this situation. The efficacy of probiotics in treating antibiotic associated diarrhoea remains to be proved. A further large trial in which probiotics are used as preventive agents should look at the costs of and need for routine use of these agents.
What is already known on this topic
Probiotics are well known for their microbiological properties and have been used to treat gastrointestinal and vaginal mucosal infections
Conflicting results have prevented probiotics from being accepted as viable alternatives to conventional treatments for antibiotic associated diarrhoea
The commercial availability of probiotics is increasing
What this study adds
Probiotics may prevent antibiotic associated diarrhoea
The potential of specific probiotics to prevent Clostridium difficile infection secondary to the use of antibiotics should be re-examined
A large trial looking at the efficacy of probiotics in preventing antibiotic associated diarrhoea, particularly in elderly patients, with an emphasis on the optimal dose and cost benefits is needed
Biological agents (“biotherapeutic agents” or “probiotics”) have been used to treat a variety of infections, most notably infections of mucosal surfaces such as the gut and vagina (box). After the discovery and development of antibiotics, the value of these traditional treatments diminished. Now, however, we are being forced to look at alternatives to antibiotics to combat the ever increasing number of infections that occur because of excessive use of antibiotics.
Probiotics and their uses
Probiotics are live organisms that improve the microbial balance of the host
Probiotics have special properties that make them useful in fighting infections of mucosal surfaces such as the gut and vagina
Different species of lactobacilli have the potential for use in clinical practice as also the yeast Saccharomyces boulardii
Probiotics are becoming increasingly available as capsules and dairy based food supplements sold in health food stores and some supermarkets
The relative lack of side effects makes probiotics a possible way of preventing antibiotic associated diarrhoea
The term “probiotic” was first used to describe “a live microbial supplement, which beneficially affects the host by improving its microbial balance.”1 Since then, research has looked at possible clinical uses for these agents and in 1995, when a greater understanding of their properties had developed, the term “biotherapeutic agents” was proposed to describe micro-organisms with specific therapeutic properties that also inhibit the growth of pathogenic bacteria.2
A number of agents have been isolated and studied with a view to clinical use. Streptococcus thermophilus and Lactobacillus bulgaricus, commonly used in the dairy food industry, were among the first to be studied. Other strains that have been used are Bifidobacterium bifidum, B longum, Enterococcus faecium, Saccharomyces boulardii, L acidophilus, L casei, and Lactobacillus GG. However, doctors are still reluctant to use these agents in clinical practice.
In this paper, we review the results from various trials carried out to study their benefits. We also look at the properties of biotherapeutic agents and options for further research.
Materials and methods
We searched Medline between 1966 to 2000 with the terms “probiotics,” “biotherapeutic agents,” “lactobacilli,” “antibiotic associated diarrhoea,” and “Clostridium difficile.” We also searched the Cochrane Controlled Trials Register and the Cochrane Database of Systematic Reviews. We restricted the search of Medline to published literature that had an English abstract; reviews identified by the searches of Medline and the Cochrane Database gave information about trials without an English abstract.3–6 We included all randomised double blind trials that compared the effects of probiotic therapy and placebo (both given in combination with antibiotics). We independently assessed articles and abstracts, and we each put forward articles for inclusion.
Overall, we identified 38 relevant papers on the use of probiotics. We excluded 28 of the 38 papers—three single blind trials,7–9 two letters on experimental use of probiotics, 10 11 three case series,12–14 and three trials originally done for another indication.15–17 We also excluded five trials that studied only traveller's diarrhoea 3 18–21 and 12 trials that studied infectious diarrhoea. 4–6 22–30 Of the latter, one trial was in normal subjects exposed to a bacterial challenge,4 one in patients infected with HIV,6 nine in children with diarrhoea of mixed causes, and one in adults with diarrhoea of mixed causes.26 We excluded the trial in HIV patients because the results may not be generalised to other patients.6
Ten double blind placebo controlled trials were relevant to our area of interest (nine published in English and one in French).31–40 Our meta-analysis included nine that looked at prevention of diarrhoea. We excluded the other trial, which looked at treatment of diarrhoea40: this trial had two arms looking specifically at outcome of treatment—one of recurrent diarrhoea caused by Clostridium difficile and the other of non-recurrent disease caused by this organism.
Meta-analysis and data abstraction
The meta-analysis was carried out according to the recommendations of the QUOROM statement.41 The key outcome data taken for analysis included the sample size (table 1), treatment regimens, and numbers of patients in both arms of the study who had an absence of diarrhoea (table 2).
Quantitative data synthesis and validity assessment
We used the percentage of patients without diarrhoea in the probiotic and placebo groups as an outcome measure. We defined diarrhoea as “a change from the patient's normal bowel habit, with two or more loose or watery stools for at least two days.” We performed three separate analyses: one for the four prevention trials using S boulardii, one for the five trials using lactobacilli or enterococci, and one on pooled data from all nine trials.
We ensured that the odds ratios were not heterogeneous across the trials by performing tests of homogeneity across all nine trials (P=0.246), across the four yeast trials (P=0.065), and across the five non-yeast trials (P=0.573). As these tests did not achieve significance, we combined the information in the tables, using the Mantel Haenszel method. We also calculated summary odds ratios and 95% confidence intervals for these analyses, and we plotted a graph by using the log of the odds ratios to determine the benefit of treatment over placebo.
A funnel plot (fig 1)42 did not show any publication bias; it showed that the larger studies found benefit and the smaller studies gave results varying from good to no benefit. Two further tests were carried out to investigate possible publication bias—the Begg and Mazumdar adjusted rank correlation test for publication bias and the Egger et al regression asymmetry test for publication bias.
The Begg and Mazumdar adjusted rank correlation test is a direct statistical analogue of the visual funnel graph (in fig 1). Note that both the test and the funnel graph have low power for detecting publication bias. The Begg and Mazumdar procedure tests for publication bias by determining if there is a significant correlation between the effect estimates and their variances. When this test was used on the data, a P value of 0.297 was obtained.
The Egger et al regression asymmetry test and the regression asymmetry plot tend to suggest publication bias more often than the Begg approach. The Egger test detects funnel plot asymmetry by determining whether the intercept deviates significantly from zero in a regression of the standardised effect estimates against their precision. Egger's test for bias gave a P value of 0.08 and the confidence interval included zero.
The Begg and Egger tests provide graphical output (figs 2 and 3). In figure 3, the regression asymmetry graph plots the standardised effect estimates (odds ratio) against precision (1 divided by standard error) along with the variance weighted regression line and the confidence interval about the intercept. Failure of this confidence interval to include zero indicates asymmetry in the funnel plot and may give evidence of publication bias. Guide lines at x=0 and y=0 are plotted to help determine visually whether zero is in the confidence interval. As this test did include zero, we concluded that publication bias was not present in our meta-analysis.
Nine trials were included in the final analysis (fig 4). The study regimens used probiotics combined with one antibiotic or a variety of antibiotics (table 2). All trials studied the efficacy of a probiotic in the prevention of antibiotic associated diarrhoea. The numbers of patients and the duration of follow up varied greatly from study to study, but the patients' characteristics were similar for the active treatment and placebo groups within each study.
We calculated the odds ratio on the basis of the proportion of patients free of diarrhoea on treatment compared with that in control groups. After tests of homogeneity, summary odds ratios and 95% confidence interval limits were provided for the combined data of the four trials that used S boulardii (yeast trials), the five non-yeast trials, and all nine trials together. The combined odds ratios for the four yeast trials and for the five non-yeast trials were similar (0.39 (95% confidence interval 0.25 to 0.62) and 0.34 (0.19 to 0.61), respectively); both favoured active treatment over placebo in the prevention of antibiotic associated diarrhoea. The odds ratio for pooled data from all nine trials was in favour of active treatment over placebo in the prevention of antibiotic associated diarrhoea (0.37; 0.26 to 0.53). Six studies showed a significant benefit of probiotic treatment compared with placebo (P<0.05) (fig 5). 31 33 34 36 37 39 One study showed benefit for only a subgroup of patients who did not receive non-antibiotic drugs likely to induce diarrhoea, such as magnesium hydroxide (for constipation), lactulose, and bisacodyl (for hepatic encephalopathy).32
McFarland et al showed that S boulardii lessened diarrhoea associated with antibiotics that contain a β lactam ring and prevented recurrent C difficile infection when given in combination with standard antibiotics. 37 40 There was no benefit, however, when S boulardii was used to treat primary infection with C difficile.40
Lactobacillus GG was successfully used to treat a group of patients with recurrent diarrhoea caused by C difficile.10 Vanderhoof et al showed a significant reduction in antibiotic associated diarrhoea, using this agent in children.39 The other trials in our meta-analysis used different strains of lactobacilli, two trials used two different strains of lactobacilli in combination. 32 35 Wunderlich et al showed that the strain of enterococcus strain SF68 that produced lactic acid was useful.34 In the Orrhage study, a combination of L acidophilus and Bifidobacterium survived in the human gut and reduced the faecal counts of clostridia.36
Our meta-analysis of trials that used live organisms to prevent diarrhoea associated with antibiotics shows that probiotics may be effective in preventing antibiotic associated diarrhoea. We had only a small number of trials in our meta-analysis, and it should be noted that the different antibiotics used in the trials may have altered the risk of patients getting diarrhoea and their response to the probiotics. Although probiotics have been used to prevent or treat diarrhoea of other causes—namely traveller's diarrhoea and infantile infectious diarrhoea—we did not include trials that investigated probiotics in these indications; however, most of these studies showed positive results, and some reviews have been encouraging.43
The way in which probiotics affect the gut has drawn much interest. To combat the problems of gastrointestinal infection, a probiotic must be non-pathogenic and must act against pathogens by different mechanisms from antibiotics—for example, by competition. More importantly, they should have a fairly rapid onset of action and survive the challenges of gastric acid, bile, or concurrent antibiotics. It is also desirable that they modify immune processes to destroy the invading organism. Saccharomyces boulardii and lactobacilli display these common properties.
A few live organisms have been used in many trials. S boulardii, a non-pathogenic yeast, is one such organism. It has a growth temperature of 37°C, rapidly colonises the bowel, does not alter the normal gut flora, and is cleared from the colon after treatment is discontinued.44 Of the four yeast trials, two studies individually showed significant benefit, 31 33 37 but one did not38; differences in the dose and duration of treatment with S boulardii and variations in the period of follow up may explain this disparity. Interestingly, S boulardii can also destroy the receptor site for C difficile toxin A and B by producing a protease47; this could explain how S boulardii was noted to reduce the frequency of toxin B positivity.40 This finding was criticised,45 but it was also supported.46
The other probiotic agent used widely in clinical trials is the Lactobacillus species. The mechanism of action of lactobacilli may be through multiple means: Lactobacillus GG has shown beneficial effects on intestinal immunity, it increases the numbers of cells that secrete immunoglobulin G and other immunoglobulins in the intestinal mucosa, and it stimulates the local release of interferon.48 It also facilitates antigen transport to underlying lymphoid cells, and shows increased uptake in Peyer's patches.48 Lactobacillus GG has also been shown to produce an antimicrobial substance that inhibits the growth of Escherichia coli, streptococci, C difficile, Bacteroides fragilis, and Salmonella.49 L casei shirota also showed good survival in the gut in separate studies, and mucosal antibody titres (specific to lactobacilli) were increased in the presence of this agent. 50 51 Although there was no discernible change to the numbers of clostridia or enterococci, there was an increase in the numbers of excreted bifidobacteria—a normal bowel anaerobe. 50 51 It is possible that this increase in bifidobacteria interferes with the pathogenic potential of C difficile.
Advantages of S boulardii over current clinical practice include its ready availability in the form of brewer's yeast, its easy administration, and the remarkable cost effectiveness of its use compared with vancomycin when infection occurs.11 However, there is a risk of fungaemia in immunocompromised patients52 and further large trials to document safety are needed before use of this agent will be accepted widely. Some papers report the development of septicaemia in immunocompromised patients and of endocarditis in those with damaged or artificial heart valves who have been treated with lactobacilli 53 54; it would seem prudent to avoid using lactobacilli in such patients.
Probiotics are a possible solution in the prevention of antibiotic associated diarrhoea. Clostridium difficile infection is increasingly prevalent in today's hospital setting, particularly in elderly patients, in whom 10-20% of such cases occur.55 The incidence of antibiotic associated diarrhoea depends on the antibiotic used and each individual patient's risk factors. The standard regimens to treat colitis associated with Clostridium difficile are metronidazole and vancomycin; although these drugs are successful in 80% of cases, about 20% of patients suffer from recurrence.56 In light of the need to control costs in these days of managed health care, we must re-examine the benefits of using live organisms. Whether the use of probiotics can actually reduce the length of hospital stay by reducing the incidence of infection with C difficile and the need to use antibiotics such as metronidazole and vancomycin are issues that need to be addressed in a clinical trial.
Our meta-analysis of nine trials shows that biotherapeutic agents may be useful in preventing antibiotic associated diarrhoea, but it provides little support for a role of probiotics in the treatment of such diarrhoea.
The increasing availability, lower costs, and relative lack of side effects of probiotics contrast with the problems associated with current antibiotic regimens. Commercially available strains are being marketed in capsules and yoghurt based drinks, but their potential benefit needs further investigation. It would be wrong to credit the proved benefits of one strain to an untested but closely related strain.57 Data from trials have provided us with clear evidence on the efficacy of some strains in the gut, but we still need to see confirmation of their clinical benefit.
Contributors: CJB had the original plan for the meta-analysis, researched individual papers, offered statistical advice, and edited the final paper. AD'S conducted the online and other searches, researched individual papers, extracted output data from individual trials, and wrote the paper. JC was responsible for the statistical analysis, calculated final output data and odds ratios, drew the final graph, and conducted tests of publication bias. CR researched individual papers and assisted in the preparation of the article.
Editorial by Barbut and Meynard
Funding No funding was sought or obtained for the conduct of this study.
Competing interests None declared.