Study selection and inclusion criteria
In this review we chose to only include studies that had a clear definition of antibiotic-associated diarrhea, to be able to compare their results in a systematic way. However, studies lacking a precise definition of diarrhea may still provide valuable information, and it could be a subject for future discussions how to interpret them and whether to take them into account when formulating recommendations. Furthermore, the strict definition of diarrhea used in some studies means that the protective effect of probiotics against AAD may have been underestimated [
67]. Given the scope of the review, we searched for clinical trials involving the use of antibiotics, but we didn’t apply strict inclusion criteria regarding the kind of antibiotic used. We didn’t look for studies using specific treatments, nor did we exclude studies that did not indicate which antibiotics they used, since diarrhea can be a side-effect of many. Five of the studies that we included did not specify which antibiotic was administered to the patients during the clinical trial. Of the remaining 27 studies, 21 enrolled patients taking different antibiotics, including antibiotic such as broad-spectrum penicillins and cephalosporins associated with a high-risk of AAD.
Only some studies reported which antibiotics were used among the characteristic of patients following treatment [
58,
65,
71]. Kotowska et al. [
65] suggested that the probiotic they tested (
Saccharomyces boulardii) may be effective in preventing diarrhea caused by amoxicillin with clavulanate and by intravenous ceforuxime, but they also mentioned that they could not make definitive conclusions regarding differences in the probiotic’s efficacy against different classes of antibiotics. Similarly, other studies could not detect significant differences in this regard, either because of a small sample size (relative to the number of antibiotics tested) or because of the low incidence of diarrhea in the study, or both [
75,
82]. Twenty-two of the studies included in this review included a power analysis. Of these, 11 detected a significant difference between treatment and placebo, including two studies that were underpowered according to their power analysis [
56,
65]. Of the 11 studies that did not detect a significant difference between the treatment and the placebo, four were underpowered [
72‐
74,
85]. By ensuring that clinical trials have enough power it would be possible to identify which probiotics are most effective in preventing diarrhea caused by specific antibiotics, and clinicians would be able to recommend different probiotic products based on the antibiotic therapy prescribed to their patients.
An important aspect in the design of clinical trials is the inclusion of a placebo control group. This kind of control allows clinicians to account for the placebo effect, which is a well-recognized phenomenon in clinical practice. In clinical trials with dairy products the placebo would ideally consist of a specifically developed product with organoleptic properties very similar to the dairy product containing the probiotics. However, in case of the clinical trials testing probiotic dairy drinks included in this review, the placebo is often a different product. Since we are not aware to which extent these product differences affect the placebo response in individual patients, it may be opportune for producers of probiotic dairy drinks to develop products that can be administered as more appropriate placebos in clinical trials.
Criteria for recommendations
In this review, we adopted strict criteria to derive recommendations from the results of our meta-analysis. Specifically, we decided to limit strong recommendations for commercial products for which the specific probiotic combination was tested, and not the single species separately, and for which the efficacy of the composition is supported by at least three clinical trials. This approach, although necessary to ensure evidence-based decision making, is limiting, since there are likely other products on the market whose exact composition has not been tested but that may be effective in preventing AAD. In fact, many of the works that we reviewed that assessed the efficacy of multi-strain probiotics (more than three strains) in reducing the risk of AAD, concluded that these products had a significant effect on risk reduction [
63,
64,
86].
For dairy products, we recommended those that were shown to have a positive effect in a clinical trial, but it is possible that products from other brands, with a similar formulation, may be as effective as those tested. For example, we reviewed here a study showing a positive effect of the probiotic dairy drink Actimel (Danone) in preventing diarrhea caused by antibiotics [
62], and we subsequently included this product in the list of one-star recommendations. The brand Yakult produces a dairy drink containing a strain of
Lactobacillus casei that has been shown to be virtually identical to the strain used by the brand Actimel [
87], providing an argument for a one-star recommendation to the Yakult dairy drink without the need to conduct additional clinical trials.
Although our recommendations are based on different criteria and are not limited to children, they are in line with those of the European Society for Pediatric Gastroenterology, Hepatology and Nutrition (ESPGHAN) working group [
53]. The strain
L. rhamnosus GG, for which we make a three-star recommendation, was also strongly recommended for the prevention of AAD in children by the ESPGHAN working group, on the basis of a moderate quality of evidence. The working group also gave a strong recommendation to
S. boulardii. However, we could not do the same on the basis of our analysis, because we pooled results of patients of different ages. In the forest plot in Additional file
4,
S. boulardii shows a positive effect in the prevention of AAD in children [
65], a positive trend in adults [
68,
75], but no positive effect in elderly [
66,
70]. Age is one of the factors that should be taken into account when evaluating health benefits of probiotics. In general, differences in the inclusion criteria, in the methods used to conduct the meta-analysis and in the criteria used to formulate recommendations will result in different evidence-based advice.
Factors affecting the efficacy of probiotics
Multiple factors can determine the efficacy of probiotic products in specific therapeutic contexts. Firstly, the efficacy of a product can be influenced by its strain composition. One of the most studied probiotic strains is
Lactobacillus rhamnosus GG, which has been repeatedly proven effective in reducing in the incidence of diarrhea in antibiotic-treated patients and in treating other gastrointestinal disorders [
88]. Different strains of
L. rhamnosus may not be equally effective in preventing the incidence of side effects of antibiotics [
13], and the same is true for other probiotic species. Clinical trials should always specify which probiotic strain they tested, however this is not always the case, making it difficult to evaluate and compare their results. Furthermore, genetic variability has been observed among “identical” strains of LGG [
89], so even when studies indicate precisely which strain they used it is not possible to exclude the possibility of within-strain differences affecting the results of the trial. Apart from strain composition, the formulation of a probiotic product (specific combination of strains) may affect its efficacy. This effect may be particularly significant in dairy products, since the quality of the product will vary depending on the specific strains used during the fermentation, and whether they are included during the process or added as ingredients to the final product. In this review we have analyzed dairy products and food supplements separately, and we have only combined probiotic products with the exact same strain composition and formulation, in order to minimize the effect of these factors on the results of the meta-analysis.
Apart from strain composition and probiotic product formulation, specific individual differences (age, specific health condition, genetic factors and differences in the composition of the gut microbiome) might play a role in the efficacy of probiotics, as is evident in some of the trials we reviewed.
The largest study included in this review contained almost 3000 subjects, as reported by Allen [
81]. This study showed no significant effect of probiotic versus placebo. However, it included elderly participants (over 65) who may be more susceptible to adverse effects of antibiotics.. The efficacy of probiotics varies across different age groups, and is influenced by the type of antibiotic administered and the duration of the therapy. In fact, higher incidence rates of AAD were previously observed in older patients also subjected to prolonged antibiotic exposure [
84], so the same factors may partly explain the observation of the study by Allen. Furthermore, in the study by Allen antibiotic therapy could last up to 7 days before starting the probiotic treatment, and probiotics may be more effective when administered during the entire period of susceptibility. In fact, a meta-regression analysis conducted by Shen et al. [
90] showed that probiotics were significantly more effective in reducing the risk of
Clostridium difficile infection when administered closer to the first antibiotic dose, and similar considerations could be applied to the use of probiotics to prevent AAD.
The efficacy of probiotics in preventing AAD also depends on the dose. A daily intake of at least 5 × 10
9 CFU is associated with significant efficacy for AAD [
41,
91], and it has been shown that higher probiotic dose is linked to greater efficacy [
60,
84]. Although only few dose-effect studies have been performed, they observe a positive correlation between dose and AAD risk [
92].
Since so many factors can affect the efficacy of probiotics in prophylaxis, researchers should be rigorous in setting up clinical trials and in providing as much information as possible about them. Studies should report characteristics of the probiotic (strain, dose and duration of therapy), of the antibiotic (type of antibiotic, duration of the therapy) of the patients (age group, diagnosis) and accurate definitions of measured outcomes and adverse effects. In this way, results from different trials can be assessed, compared and used as a basis to formulate recommendations. Individual factors, that are not routinely monitored in clinical trials, may influence the incidence and gravity of side effects and the efficacy of probiotics. For example, each individual has their own unique microbiota, and the impact of a given antibiotic on the composition and stability of different microbial ecosystems can be different; therefore, a specific probiotic strain or combination of strains may not have the same efficacy for every person. Especially for some patients, for example those who are frequently treated with antibiotics such as elderly in care facilities, it is certainly worth being flexible and trying different probiotics until the most effective one has been found. Future research can guide the formulation of personalized therapies.