NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Hempel S, Newberry S, Ruelaz A, et al. Safety of Probiotics to Reduce Risk and Prevent or Treat Disease. Rockville (MD): Agency for Healthcare Research and Quality (US); 2011 Apr. (Evidence Reports/Technology Assessments, No. 200.)

Cover of Safety of Probiotics to Reduce Risk and Prevent or Treat Disease

Safety of Probiotics to Reduce Risk and Prevent or Treat Disease.

Show details

Discussion

Results Summary

The review demonstrates that there is a large volume of literature on probiotics. However, the literature provided only limited evidence to address the questions the review set out to answer. The search of 10 databases combined with reference screening of included studies and pertinent reviews identified 11,201 publications, and 622 studies were included in the review. Of these 622 studies, 235 studies made only nonspecific safety statements (“well tolerated”), and the remaining 387 studies reported the presence or absence of one or more specific adverse events.

The review includes a large number of randomized controlled trial (RCTs); however, the majority of these were not designed to monitored adverse events but primarily tested the efficacy of probiotics in managing, treating, or preventing clinical symptoms. The quality of included studies varied within study design categories; only a minority of trials reported adequate randomization methods, concealment of treatment group allocation, and blinding of outcome assessors to the treatment group; and studies were not powered to assess adverse events. Adverse events were poorly documented and publications seldom stated what parameters were monitored. Further, in the majority of included studies, interventions were poorly documented, lacking detail, for example, on the specific probiotic strain that was administered as well as the dose and viability.

Identified case studies indicated that fungemia, bacteremia, and sepsis may be associated with administered probiotic organisms. None of the identified case series, controlled clinical trials (CCTs), parallel and crossover RCTs reported an infections caused by the administered probiotic strains. However, these studies did not monitor routinely for such infections; reported adverse events were primarily gastrointestinal in nature. In parallel RCTs, no statistically increased risk for adverse events in the quantity of adverse events was observed, analyzing the number of participants with adverse events and reported adverse event incidences per treatment group. Exploring the nature of reported events in the literature, we found that adverse events were gastrointestinal in nature, addressed infections and infestations, or addressed other adverse events. In none of the different types of adverse events did parallel RCT show a statistically significantly increased risk for adverse events in intervention participants compared to control. Across studies, there was also no statistically significantly increased risk of serious adverse events associated with probiotic product use. Long-term effects are largely unknown as very few existing studies report on followup periods of one year or more.

Stratifying studies by probiotic genus, it was apparent that the existing literature covers primarily the genus Lactobacillus, alone or in combination with other genera, most frequently Bifidobacterium. There was some evidence from a metaregression that indicated Streptococcus interventions may be associated with a larger number of adverse events compared to other genera, but evidence from direct, head-to-head comparisons is lacking. Stratifying RCTs that used each genus exclusively, no statistically significant difference between intervention and control group participants was observed for any of the six genera. However, published reports on the genera Enterococcus, Bacillus, Streptococcus are largely absent from the literature. Saccharomyces interventions and Bifidobacterium interventions were also rare, and a substantial proportion of studies used blends of probiotic organisms.

The review aimed to address a large number of participant and intervention variables and their effect on safety. Direct evidence comparing intervention factors is largely absent from the existing literature. Few studies directly compared the safety of different product or participant characteristics. Indirect comparisons indicated that effects of delivery vehicles should be investigated further. Analyzing participant factors such as health status showed that case studies described adverse events in patients with existing health concerns, often already hospitalized when potentially probiotics associated infections occurred. However, RCTs did not indicate a statistically significantly increased risk of adverse events in healthy, medium-risk, or critically ill participant groups compared to control.

Scope and Limitations

This evidence report considers a large number of studies and addresses a large number of research questions. Unlike the majority of existing reviews, this evidence report considers only adverse events reported in studies of probiotics, and does not cover efficacy or effectiveness questions for the management, prevention, or treatment of clinical symptoms or other indications for using probiotic products. For a risk–benefit analysis, both aspects would need to be considered.

A substantial number of reviews summarizing individual studies of effects of probiotics have been published. However, existing reviews focus on selected interventions, selected probiotic genera, selected patient groups, or selected outcomes (Abad, 2009; Alfaleh, 2008; Allen, 2003; Barclay, 2007; Boyle, 2009; Boyle, 2008; Brenner, 2009; Butterworth, 2008; Chande, 2009; Chande, 2008; Chmielewska, 2010; Chou, 2008; Dendukuri, 2005; Deshpande, 2007; Deshpande, 2010; Doherty, 2009; Doron, 2008; Dugoua, 2009; Fuccio, 2009; Gawronska, 2005; Gurusamy, 2008; Holubar, 2010; Hoveyda, 2009; Johnston, 2007; Kahn Ch, 2009; Kale-Pradhan, 2010; Lirussi, 2007; Mallon, 2007; McFarland, 2005; McFarland, 2010; Miller, 2009; Moayyedi, 2008; Osborn, 2007; Petrov, 2009; Pillai, 2008; Rolfe, 2006; Sachdeva, 2009; Szajewska, 2010; Szajewska, 2005; Szajewska, 2001; Szajewska, 2004; Tung, 2009; Vouloumanou, 2009; Wang, 2009; Watkinson, 2007; Whelan, 2010; Wu, 2008; Zigra, 2007). This evidence report has a broader scope, and due to the large number of included studies, allows unique statistical analyses. Adverse events reported in intervention studies of probiotic organisms are largely rare events encountered by only a small number of participants. Thus, large sample sizes are necessary to be able to detect any statistically significant incidence rates of such adverse events.

Search

This review aimed to capture the safety of probiotics, in particular the safety of Lactobacillus, Bifidobacterium, Saccharomyces, Streptococcus, Enterococcus, and Bacillus used as probiotic agents. The search strategy was primarily designed to capture all explicitly identified probiotic studies, and steps were taken to ensure the completeness of the body of evidence of probiotic literature. We identified a large number of publications on probiotics and carefully screened full paper copies of all publications that might contain information on the safety of probiotics. Other studies that investigated the same genera in ways that resembled their use as probiotic agents but did not label their interventions as probiotic studies were not excluded but were also not sought systematically as outlined in the search strategy justification, and no claim of completeness is made. These studies were primarily identified through reference mining, that is, scanning the bibliographies of included studies and pertinent review articles. This review was not restricted to particular species, strains, patient group, clinical fields, settings, or study design, and the sought interventions included genera such as Bacillus with known pathogenic properties, hence the decision to restrict the search to probiotic studies rather than expanding it to the wider literature on the individual bacteria and yeast strains. Judging from our experience, future reviews targeted towards more specific research questions should use a combination of search terms covering both the term “probiotic” and the genus to identify those studies that used a particular strain as a probiotic agent.

This review adopted a thorough process of identifying information on the safety of probiotics by screening full paper copies of empirical studies on probiotics, regardless of whether the safety of probiotics was mentioned in the summary of the article, that is, the title or abstract of the publication. Initial experiments with search filters have shown that screening studies at the title or abstract level would have resulted in missing a large proportion of the pertinent literature. The majority of included studies were not tagged by databases as including safety information, the title and the abstract gave no indication that adverse events would be addressed in the publication, and in the overwhelming majority of studies other than case reports, safety was not the main aim of the publication.

The review focuses on published literature, and a substantial number of studies of probiotics have been published in scientific journals. However, there may also be a substantial number of unpublished studies, most likely from manufacturers of probiotics. This factor, combined with the fact that we could not be certain studies that failed to mention adverse events indeed had no adverse events, limits the utility of the review as a basis for true risk–benefit analysis of probiotics.

Probiotics

This exploratory review on the safety of probiotics lists the reported presence and absence of adverse events for interventions that used Lactobacillus, Bifidobacterium, Saccharomyces, Streptococcus, Enterococcus, and Bacillus organisms as probiotic agents. The imbalance of genera in the included studies (favoring Lactobacillus alone and in combination with Bifidobacterium) presumably reflects the research conducted to date.

We adopted a very inclusive definition of probiotics. However, there is an ongoing debate about whether yogurt should be considered a probiotic product, since yogurt contains live bacteria (e.g., Guarner, Perdigon, Corthier, et al., 2005) of genera that are associated with probiotic properties, and the debate also extends to whether there is any reason to think adverse events need to be monitored for yogurt and lactic acid bacteria products (e.g., MacGregor, Smith, Thakker and Kinsella, 2002). For this review, yogurt studies that did not explicitly report the addition of a probiotic agent, that is, a strain in addition to the yogurt starter culture, were excluded.

A distinct limitation of this review is that most of the identified studies provided insufficient information on the intervention, that is, a clear description of the microbes that were included in the investigated probiotic product. The lack of identification or proper classification of the administered probiotic organisms is a safety concern in itself. A large number of published studies did not report the strain of the probiotic agent included in the preparation. Given that the efficacy of probiotics is often considered strain specific, the informational value of these studies has to be questioned. Lack of documentation is hindering efficacy as well as safety evaluations (EFSA, 2009; Shane, 2010) and limits overviews necessary for consumers and policymakers.

A further limitation is the uncertain reliability of the reported product details. For this literature review, we rely on the information reported by the study authors. Very few studies reported using accepted methods (or any methods) to test the content of preparations given to participants. The exact organisms as well as any contaminants present in the preparations are pertinent information. For example, included studies indicated that the species used was Lactobacillus sporogenes however; the species designation Lactobacillus sporogenes is now considered an invalid name for Bacillus coagulans (Becker, 1950; De Clerck, 2004; Jung, 2009). Similarly, some studies reported on Streptococcus faecium and Streptococcus faecalis, which have been transferred to the genus Enterococcus (Schleifer, 1984). A study published in 2006 conducted a survey of commercial probiotic strains and found that 28 percent of the strains intended for use in humans as probiotics were misidentified at the genus or species level (Huys, 2006). Other reports show that products can contain more species than noted on the product labels (Marcobal, 2008; Underwood, 2009).

Also, over the time span covered by our literature search, many of the employed organisms may have undergone mutations (spontaneous or otherwise), identification techniques have improved (e.g., revealing them to be less similar to a more familiar strain or to belong to a different genus than previously thought), and taxonomic name changes were introduced (see, e.g., Masco, 2004; Mattarelli, 2008; No Author, 2008; Li, 2006; Posteraro, 2005; Morita, 2009).

Finally, we identified a large number of studies that gave a blend of different probiotic organisms to participants. These studies individually do not permit to attribute reported harms to a particular genus, species, or strain. Metaregressions can to some extent trace effects across studies, but this process cannot replace adequate study designs to investigate the safety of probiotic strains.

Intervention Studies

This report was explicitly limited to assessing the outcomes of interventions (as opposed to merely passive or accidental exposure). We identified a large number of intervention studies in the international literature assessing the effects of Lactobacillus, Bifidobacterium, Saccharomyces, Streptococcus, Enterococcus, and Bacillus used as probiotic agents. A number of publications exists that systematically collated example cases of fungemia associated with Saccharomyces cerevisiae (e.g., Munoz, 2005), or infections associated with Lactobacillus (e.g., Aguirre, 1993; Husni, 1997), or Bifidobacterium (e.g., Bourne, 1978). However, we considered only those case descriptions that reported a preceding intervention, that is, the purposeful use of probiotics. This limitation also pertains to reports from hospitals describing outbreaks of fungemia such as reports on an intensive care unit (ICU) where patients did not purposefully consume probiotics, but the yeast was reported to linger in the ICU (Cassone, 2003). One of the included case studies (Perapoch, 2000) also reported on an infant who appeared to have contracted an infection from an infant treated with Saccharomyces cerevisiae who later developed fungemia; hence, spread of infections should also be monitored in research studies.

The review considered studies without study design restrictions and it includes a large number of different study designs such as parallel and crossover RCTs, CCTs, case series, and case studies. However, the literature search did not identify any observational cohort studies comparing two cohorts or retrospective case-control studies on the safety (or even the efficacy) of organisms used for their probiotic properties; all observational data came from case series following only one intervention group and case studies. The reason for this lack of large-scale observational studies of probiotic safety is unclear but may be the result of a general presumption of probiotic safety on the part of epidemiologists (and the failure to implicate them as the cause of any particular conditions). A 2002 epidemiological study addressing a similar question assessed changes in the incidence of Lactobacillus-associated bacteremia in Finland after a rapid increase in the use of Lactobacillus rhamnosus GG as a probiotic agent. The study found no increase in the incidence of Lactobacillus-associated bacteremia in the population, although a small proportion of isolates matched the strain of the probiotic agent, using the typing technology available at that time (Salminen, 2002).

Safety

The review identified a large number of relevant publications addressing the safety of probiotic products. For RCTs, we identified a similar volume of publications that addressed the potential efficacy of probiotic preparations but not their safety. It is not possible to extrapolate from the lack of mention of adverse events that no adverse events occurred in interventions (e.g., the adverse events associated with a particular trial might be reported in an accompanying or subsequent, not-yet-published, article). Even fewer RCTs reported on the presence and the absence of specific adverse events.

The review identified a large number of publications that made vague safety statements such as “the intervention was well tolerated” and “there were no adverse events.” We compiled these vague references to safety to allow a complete overview of the existing literature, but these studies were analyzed separately from studies with more specific statements. This group of studies reported no information on what was monitored or how “well tolerated” was defined. For an evidence report such as this whose purpose is to synthesize the evidence, these studies are of little informational value.

When publications reported that there were no adverse events, we did not make inferences from this statement to specific outcomes. Although it may appear plausible to assume that this means no death or hospitalizations occurred, this assumption is very problematic and should not replace actual empirical evidence on the safety of probiotics. The safety of probiotics has only recently been considered as an issue warranting further investigation (Liong, 2008). Older publications may not have thought to associate such harms with an intervention considered completely harmless. In order to advance the empirical evidence on the safety of probiotics, studies should monitor and report the presence and also the absence of specific harms.

For this review we extracted all reported adverse events, regardless of whether the authors of the publication considered these in their summary statement regarding the safety of probiotics. We also included outcomes regardless of the author’s assurance that the event was unrelated to the intervention. Such judgments are difficult to make and may change with increasing knowledge of the safety of probiotics. Very few publications appear to have addressed the assessment of the strength of association between adverse event and intervention systematically, as reported for example in Gibson (2009).

Safety reviews on probiotics have focused on various aspects of safety such as toxicity, the potential for translocation, and antibiotic resistance or other virulence factors (Ishibashi, 2001; Sanders, 2010; Yazdankhah, 2009). This report operationalized safety as the presence or absence of unintended adverse health events in probiotics interventions for human participants. We document the quantity, quality, and nature of adverse events reported in research studies using probiotics to reduce risk of and prevent or treat disease in vivo.

Efficacy studies for which the efficacy outcome was the mitigation of an adverse event (e.g., efficacy of probiotics in preventing or treating antibiotic-induced diarrhea or other negative health outcomes) were excluded unless (1) the outcome was actually exacerbated in the probiotic treatment group compared to baseline or to a control group and this outcome was one of the main safety findings of the paper (stated in the abstract of the publication, so-called treatment failures); or (2) the safety of the probiotics, themselves, was also explicitly addressed in the publication. This operationalization is not without problems but it is a pragmatic solution adopted in other recent overviews of the safety literature (e.g., Pitrou, Boutron, Ahmad & Ravand, 2009).

Particular outcomes addressed in this review warrant further investigation as a risk-benefit analysis in a review that includes all studies reporting on a particular outcome such as all-cause mortality. Such a review would need to include all studies addressing the outcome, regardless of whether the outcome was considered a measure of efficacy or an unintended effect.

Key Questions

Key Question 1. What is the evidence that the active and lyophilized forms of probiotics (Lactobacillus, Bifidobacterium, Saccharomyces, Streptococcus, Enterococcus, and Bacillus) as single ingredients or in combination with other probiotics or prebiotics in all delivery vehicles (and formulations) when used to cure, treat, mitigate or prevent a disease or reduce disease risk are safe in the short term? Long term?

The question of whether probiotic interventions are safe cannot be answered with sufficient confidence based on the existing literature. The existing literature includes primarily the genera Lactobacillus, alone and in combination with other genera, often Bifidobacterium; adverse events associated with other genera are not well documented.

Case studies indicated that primarily fungemia, but also bacteremia, and incidences of sepsis have been linked to administered probiotic organisms. Although the confidence of matching strains has only recently been improved through DNA-based matching methods, the existing reports indicate that an association between administered probiotic strains and observed infections must be considered (Liong, 2008).

RCTs, CCTs, and case series did not report that they routinely monitor for the kinds of infections identified in case reports. This is particularly distressing as the identified case studies span a long period; the infectious potential of probiotic organisms is not a recent observation (Jensen, 1976; Richard, 1988). Most controlled trials did not state what harms were monitored, and the safety of the probiotic products was not addressed systematically. Poor reporting of adverse events is not specific to studies on probiotic products but a general concern of intervention studies (Ioannidis, 2004).

None of the identified case series, CCTs, or parallel and crossover RCTs reported an infections caused by the administered probiotic strains. However, these studies did not monitor routinely for such infections. The absence of reliable evidence on adverse events should not be mistaken for evidence of the absence of adverse events. The adverse events reported in RCTs in the current literature do not suggest a widespread risk, but future studies that explicitly monitor for the safety issues of concern are needed to quantify the actual risk of specific adverse events in intervention studies.

Frequently reported individual adverse events were deaths that occurred during the study followup period; many gastrointestinal incidences such as diarrhea, constipation, or nausea; and respiratory infections. These types of outcomes were reported for both study arms, participants using probiotics as well as participants in control groups. Across studies most incidences were distributed evenly across treatment groups; nonetheless, there were individual studies such as the PROPATRIA trial reported by Besselink et al. (2008), a study of failed effectiveness reported a higher mortality rate in the probiotic treatment group than in the control group in patients with acute pancreatitis, which indicates that individual outcomes such as mortality should be monitored. In particular, as the mechanism of action must be investigated further, the study reported no incidences of infections caused by the administered probiotics organisms (Lactobacillus and Bifidobacterium strains). In a further publication, this mortality rate was determined to be increased in those taking probiotics who had organ failure, as compared to those who did not (Besselink, 2009). The analysis of individual outcomes also suggests that treatment failures should be highlighted in current research. Although treatment failures were not considered per se for this review, failed efficacy was sometimes considered a safety concern (Besselink, 2008; Boyle, 2008) and a central outcome of the study. Individual outcomes such as mortality should be assessed in a risk–benefit analysis that includes the outcome regardless of whether it was investigated as a safety concern or efficacy measure (i.e., where probiotics were given to reduce mortality).

To approach the question of safety of probiotics, we also systematically investigated the quantity of adverse events reported in probiotics studies. This information is meaningful only in comparison to a control group, a comparable group with similar patient characteristics, co-interventions, and other similar circumstances that permit investigation of whether adverse events are increased with probiotics use. We investigated two alternative measures, the number of patients with adverse events in each treatment group and the number of adverse-event incidences per treatment group. Each measure has inherent advantages and disadvantages, and the measures are not identical, as a single participant can experience multiple adverse events. Across all individual studies and identified adverse events, parallel RCTs did not indicate a statistically significantly increased risk of adverse events in either of the complementary measures. However, it has to be considered, though, that the existing literature is dominated by Lactobacillus-based interventions, both in combination with several other genera or alone.

Finally, the current literature also does not permit statements on the long-term safety of probiotics. With few exceptions, the existing literature reports on short- and medium-term use of probiotics assessed for a short or medium-term followup period. Research on probiotics has increased dramatically in recent years and studies in the near future may report more information on long-term effects of probiotics.

Key Question 2. What are characteristics and associations of the reported harms in Question 1?

The reported adverse events were primarily gastrointestinal in nature, others concerned infections and infestations, and a large group of studies did not fit any particular category in the published system used to classify adverse events (DHHS, 2009). While the case studies primarily reported infections suspected or confirmed to be caused by an administered probiotic organism, the majority of other studies reported gastrointestinal incidences. In the included RCTs, there was no indication that participants using probiotic organisms have a higher risk of experiencing gastrointestinal adverse events than those not using them and this was also the case for infections and infestations and all other reported adverse events across studies. Studies rarely reported efforts to monitor harms specific to probiotic product interventions, including infections due to the administered strains. Hence, evaluations of the safety might change with future, more targeted, assessment of adverse events (Liong, 2008).

There is a lack of studies investigating potential interactions between probiotics and other, concomitantly administered, medications. The descriptions of cases experiencing serious adverse events suggest that either multiple medications or the underlying condition may have contributed to the severe adverse events reported but studies systematically addressing interaction effects are lacking.

We identified only a very small number of studies addressing acquired antibiotic resistance as a patient outcome with clinical relevance. Evidence for potential harms came from case studies in patients with multiple morbidities. Reported resistance pertained only to selected antibiotics. However, it has to be noted that we restricted the current review to patient outcomes, only where antibiotic resistance and translocation were described as clinical adverse events were these eligible for inclusion in the review. This excluded, for example, in vitro and animal research on the potential, or lack of potential, for antibiotic resistance and translocation that has been published for the investigated genera (Abe, 2010; Corthesy, 2007; Ishibashi, 2001).

Key Question 3. What is the evidence that harms of Lactobacillus, Bifidobacterium, Saccharomyces, Streptococcus, Enterococcus, and Bacillus differ by product and delivery characteristics?

We set out to answer a large number of research questions related to the interventions and delivery characteristics. However, identified studies lacked detail in their description of administered probiotic organisms. Many studies did not specify which probiotic strains were investigated, nor was there indication that intervention preparations were tested for identity of the included organisms, viability, or contaminants.

The question of genus-specific safety profiles is not easy to answer with the existing literature. The review included probiotic organisms that were very different in nature (bacterial as well as yeast strains) with different histories and research experiences of using the genera as probiotic products (e.g., Lactobacillus versus Enterococcus). The number of identified fungemia case reports associated with of Saccharomyces boulardii [cerevisiae] outnumbered case reports of infections reported for the bacterial strains. However, RCTs, CCTs, and case series investigated primarily Lactobacillus, alone or in combination with Bifidobacterium strain interventions; the available evidence, including reports of the absence and the presence of adverse events as well as effectiveness studies, is very unbalanced across genera.

The kind of postmarket reports of adverse events that participants might encounter when using probiotic products had to be elicited from studies that often investigated products that included different genera or gave different probiotic genera for very different purposes, to different participant groups, in different doses and potencies. Very few studies provided head-to-head comparisons of different genera. For the included RCTs, we undertook stratified analyses for each genus in studies that used organisms from one genus only, for example, all studies using exclusively Lactobacillus organisms. Stratified analyses by probiotic genus showed no increased risk of adverse events for any of the genera in studies using the genus in question exclusively. In addition, we undertook a metaregression and investigated each genus as a moderator in studies that used a particular genus alone or in combination with other genera (e.g., all studies including a Lactobacillus strain). There was some indication that interventions including Streptococcus strains showed a higher risk of adverse events compared to the other genera. However, this result was based on a small number of studies given the paucity of studies using genera other than Lactobacillus and direct evidence is missing.

Included studies used unique interventions that comprised a large number of different species and strains to investigate the efficacy, and in some cases the safety, for use as probiotic agents. Typically, there were too few comparable studies to enable individual safety statements for species or strains: many studies used interventions that included more than one probiotic organism so that it was not possible to link encountered adverse events to specific species or strains, and as outlined before, the documentation and validation of the interventions as well as the monitored adverse events were lacking. Other factors, such as a history of safe use of species in the food production, data on the prevalence of opportunistic infections, or reports of resistance to antibiotic or antifungal medications, may be considered to determine the potential for safe use. (see e.g., EFSA opinion, 2007; [Cote, 2006.]). However, these factors do not preclude the occurrence of rare adverse events, and such known properties of genera or species are only useful if there is evidence to suggest that all strains within the genus or within a species can be expected to behave similarly. Assuming that because a genus or individual species has low toxicity, no strain of the genus or species and no intervention including organisms of that genus or species can cause adverse events in intervention studies appears to be an overgeneralization.

There is also a lack of studies directly comparing product characteristics such as the mode of delivery. Indirect comparisons across the RCTs identified in this review indicated that the potential effect of different delivery vehicles should be investigated further. Subgroups indicated more adverse event incidences in the treatment group when probiotics were taken in a yogurt or other dairy product than when taken in any other vehicle. It must be kept in mind that no study actually compared adverse events between a yogurt/other dairy vehicle and any other vehicle within the same study; nevertheless, there are alternative explanations for such an observation. Probiotic organisms might maintain greater viability in dairy than nondairy vehicles, or the adverse events are actually attributable to lactose intolerance. Given that many consumers consume probiotics as part of dairy or yogurt products, this effect should be further investigated in direct comparisons. The possibility that the use of a particular food as a vehicle for probiotic organisms might alter their viability (and therefore the potential efficacy and toxicity) has been explored in a number of studies (Champagne, 2005), and some have reported that Lactobacillus rhamnosus GG isolated from 15 different manufactured food products (carriers) showed strain differences that could affect both efficacy and safety (Grzeskowiak, 2010).

The only included studies that compared the form of probiotic organisms directly compared viable and heat-killed organisms. Heat-killed organisms are not included in prominent definitions of probiotics; hence, this comparison is of minor interest. There was no indication that active forms were associated with a higher number of adverse events. The characterization of organisms was too poor in included studies to allow a systematic investigation of the influence of the form. Also seldom tested or reported was the viability of the administered organisms: Considering that probiotics are live organisms and that they presumably need to remain live to be fully functional, it is concerning that few studies demonstrated that they were indeed able to maintain the evaluated organisms in a live and robust state. Related to this concern, Bacillus species are capable of forming spores, which would affect the count of viable organisms in a preparation. Furthermore, because several of the genera of interest are primarily anaerobic, exposure to oxygen during storage could easily affect viability. Another factor that might lower the potency of probiotic products is the failure to consider the potential for cryogenic damage during lyophilization and/or storage and to compensate by adding a cryoprotectant (see e.g., Savini, 2010).

We did not identify conclusive evidence in the existing literature showing that interventions with a mixture of different organisms reported more adverse events than studies using one probiotic strain only or that synbiotics (mixtures of prebiotics and probiotics) differ from probiotics; however, there is a lack of direct comparisons. Although the risk of adverse events (as well as the efficacy) is not necessarily comparable across species and strains, direct head-to-head comparisons are largely absent in the literature and in practice, probiotic interventions often included several different probiotics genera, species, and strains.

Key Question 4. How do the harms of Lactobacillus, Bifidobacterium, Saccharomyces, Streptococcus, Enterococcus, and Bacillus vary based on (a) dose (cfu); (b) timing; (c) mode of administration (e.g., catheter); (d) age (all ages, including infants), gender, ethnicity, disease or immunologic status of the patient; (e) relationship to efficacy?

Only a few primary studies explored the effect of intervention and participant characteristics on safety. Both the variation in definitions of high and low dose across published studies and other factors such as the inherent differences in the compared organisms as outlined previously precluded a systematic evaluation of a dose-response relationship.

Very few published studies were identified that investigated the effects of long-term use of probiotics, that is, intervention durations of 1 year or longer; information on the safety of long-term use is lacking. Given the current research interest (Shane, 2010) studies will hopefully provide needed evidence on long-term interventions.

There were few descriptions of the time of onset of harms relative to treatment and the further clinical course of adverse events. In the few studies that reported on the time of onset of gastrointestinal effects, most effects were observed within in the first 3 days of treatment. The onset of infections tended to occur 1 to several weeks after initiation of probiotics use; however, this information is primarily based on case studies and was not systematically reported. A further pertinent question may be the optimal time for administering probiotics, that is, early to prevent, rather than aiming to treat or improve particular conditions, which may be associated with the risk–benefit ratio of interventions (Arciero, 2010; Sanders, 2010).

In indirect comparisons across all identified RCTs in this review, we found no evidence that a particular mechanism or route of administration of probiotic organisms (e.g., through enteral feeding) was associated with an increased risk of an adverse event relative to a control group. In the literature, serious adverse events associated with probiotic use have been linked to catheter use (e.g., Sanders, 2010). However, the route of administration is closely linked to the health status of participants.

With regard to the health status of participants, there was some indication that health status is associated with the risk for an adverse event when using probiotics. The majority of case studies reporting serious adverse events described a critically ill patient or someone suffering from multiple morbidities when they contracted a serious infection potentially caused by probiotic organisms. There was some indication in the metaregressions that health status may predict an increased risk of adverse events associated with probiotic organisms. However, a subgroup analysis of all controlled trials enrolling critically ill participants did not show a statistically significantly increased risk of experiencing adverse events for participants using probiotic organisms compared to control group participants with similar patient characteristics. Critically ill patients may be more prone to experience adverse events; however, these were not associated with the use of probiotics; adverse events were equally distributed across treatment groups. Further large controlled studies are needed to identify any increased risk for rare but pertinent adverse events, and the risk–benefit ratio should be considered (also Whelan, 2010).

For studies enrolling patients with compromised health, it would appear appropriate to use a data monitoring committee. A study by the Society for Clinical Trials’ DAMOCLES Study Group found that only about 25 percent of articles presenting the main results of clinical trials mentioned having used a data monitoring committee to ensure the appropriate collection of data throughout the trial (Sydes, 2004). Such committees would also be helpful in standardizing the collection of adverse event data in large, well-powered trials as well as in some smaller trials in populations of interest; a data monitoring working group has provided a set of guidelines (STC, 2006).

To assess the role of the age in the safety of probiotics, we stratified studies according to the age of participants and undertook separate analyses for studies in children, adults, or elderly participants. The stratified analyses did not indicate an increased risk of adverse events in any of the subgroups associated with the use of probiotics compared to corresponding control group participants. However it has to be noted that very few studies were identified that reported on elderly participants.

The identified case studies described more male than female patients. In the RCTs, we investigated the results of subgroups in female only and male only studies as well as analyzing the percent of female participants as a factor in a meta-analysis. In these indirect comparisons across RCTs, we found no indication that encountered adverse events relative to control group incidences depend on the gender of the participants.

The included studies did not provide enough information to investigate whether probiotic safety is associated with racial/ethnic characteristics. It should be kept in mind that the majority of included studies were conducted in European countries where ethnic characteristics are rarely assessed in research studies. The research field needs to advance much further in order to be able to answer such specific questions regarding the safety of probiotics; such evidence is not available for other more established interventions (such as antibiotics use) either.

In total, 59 percent of included studies were explicitly described as effective by the study authors for the various applications of probiotic use under investigation. We found no indication that the efficacy of an intervention was related to the number of encountered adverse events across all included RCTs.

Key Question 5. How often does harm associated with Lactobacillus, Bifidobacterium, Saccharomyces, Streptococcus, Enterococcus, and Bacillus lead to hospital admission or lengthened hospitalization?

While several case studies reported a new hospitalization potentially associated with the consumption of a product including Saccharomyces, Lactobacillus, or Bacillus strains, none of the case series, CCTs, and parallel and crossover RCTs reported that a probiotics intervention led to a hospitalization in the intervention participants. A comparison of all reported hospitalizations regardless of the perceived association with the intervention treatment indicated no statistically significant risk in probiotics interventions compared to the number of hospitalizations in control group participants. However, the number of hospitalizations due to adverse events was only explicitly reported on in a few of the included studies. Older publications may not have associated a hospitalization with probiotics intake, and several studies were in participants already hospitalized. As outlined previously, the safety of probiotic products has only recently been considered as an issue warranting further investigation (Liong, 2008).

A proportion of included studies reported on the presence or absence of serious adverse events following the Food and Drug Administration definition. Results for serious adverse event varied across RCTs, sometimes favoring the probiotics group and sometimes the control group, and differences across probiotic and control group were not statistically significant. The same result was obtained for Lactobacillus and Saccharomyces interventions, but there were too few studies (Bifidobacterium) or no studies (Streptococcus, Enterococcus, Bacillus) in order to analyze serious adverse events for other genera, as studies did not report on the presence or absence of serious adverse events. The reporting of adverse events appears to have improved in recent years, presumably due to stricter guidelines and higher standards imposed by journals, for example, making it mandatory to report on adverse events when reporting the results of RCTs (e.g., Item 19 of the CONSORT statement, “All important harms or unintended effects in each group”). Relevant to this review is that the reporting of the presence and absence of infections has increased in particular, possibly a reaction in part to the PROPATRIA trial reported by Besselink et al. (2008).

We also investigated pertinent subgroups that were of particular interest to this evidence report. Most notably, we did not find evidence that health-compromised patients were at increased risk of experiencing more serious adverse events than health-compromised control group participants. However, it has to be taken into account that the monitoring and reporting of adverse events is lacking, existing interventions were again primarily Lactobacillus interventions, and future assessments may come to different conclusions as the evidence base improves.

Key Question 6. How does harm associated with Lactobacillus, Bifidobacterium, Saccharomyces, Streptococcus, Enterococcus, and Bacillus relate to use of concomitant antibiotics, confounding diet therapies, corticosteroid use, immune suppressants, or other potential confounders?

Multivariate analyses in primary research studies are suitable to systematically trace interactions between cointerventions and probiotic use. In studies where some of the participants use these cointerventions while others do not, this factor and its effect on the study outcome can be investigated. We did not identify studies meeting the review inclusion criteria that reported statistical interactions between concomitant antibiotics, diet therapies, corticosteroid use, or immune suppressants and probiotics.

Although the risk of adverse events in general might be higher in participants on multiple medications, the crucial issue for this Key Question is whether participants in probiotics interventions are more likely to experience adverse events compared to corresponding control group participants. Interactions between comorbidities and cotreatments are complex research questions (Fitzgerald, 2010). For example, we might assume an interaction between corticosteroids and probiotics when studies in participant samples using corticosteroids report a higher risk ratio of adverse events than other studies. In subgroup analyses of identified studies in which the intervention participants as well as the control group participants received corticosteroids, we found no statistically significantly increased risk of adverse events for intervention participants compared to control.

Probiotic interventions have been the focus of much research interest for the prevention of side effects associated with antibiotics (Abernethy, 2008; Cots, 2008; D’Souza, 2002; Doron, 2008; Elmer, 1998 Jack, 2010; Johnston, 2005; Johnston, 2006; Kale-Pradhan, 2010; Katz, 2006; Marshall, 2008; McFarland, 2005; McFarland, 2009; McFarland, 2006; Oldfield, 2008; Rohde, 2009; Ruszczynski, 2008; Szajewska, 2005; Szajewska, 2006; Wilcox, 2009; Young, 1998; Zou, 2009). While efficacy studies for the prevention of side effects were not eligible for inclusion in the review, we included those studies that addressed side effects of probiotics in addition to side effects of antibiotics where feasible, through the design and the adverse event monitoring of the study. Across RCTs, there was no evidence for a statistically significantly increased risk of adverse events for intervention participants compared to controls or an interaction between antibiotics and probiotics.

We identified only a few studies with concomitant diet therapies. Studies in participants using immune suppressants were also largely absent in the existing literature and patients on immune suppressants were systematically excluded from a number of RCTs. The existing evidence base is not sufficient to draw any meaningful conclusions from adverse events observed in the few studies that addressed these patients.

Views

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...