Europe to ban Bifidobacterium? Probably not

On July 3^rd, 2026, the Journal EFSA Supporting Publications released an external scientific report¹ flagging six food ingredients as possible emerging risks. To the surprise (and panic) of many, one of these ingredients was the (entire) bacterial genus Bifidobacterium.

This external scientific report must be framed in the context of the community knowledge initiative created to address possible emerging risks due to the consumption of food supplements (other than vitamins and minerals). The scope of this project is to gather information from the market, including alerts from (nutri)vigilance and monitoring systems to identify substances of concern. The ultimate goal of the project is to facilitate the prioritisation of substances that are likely to be regulated under Article 8 of Regulation (EC) No 1925/2006 which, in short and in this context, gives the European Commission the power to restrict, prohibit, or place under scrutiny substances added to foods if they pose a potential health risk to consumers. This pathway has been the legal basis, for instance, for the initial restriction, and later prohibition, of monacolins (red yeast rice) in food supplements in Europe.

The scientific report estates that even if the work did not highlight clear emerging risks, [it] did result in the identification of a shortlist of substances which could be further discussed for a possible assessment in the frame of article 8 of the regulation (EC) 1925/2006. These substances are Coleus (Plectranthus barbatus), Guggul (Commiphora mukul), Toothedclubmoss (Huperzia serrata), Black radish (Raphanus sativus L.), Ginseng (Panax ginseng), and, as mentioned above, the genus Bifidobacterium.

To reach to this conclusion, and this is very important to understand the situation, the authors follow a defined methodology.

First, case reports of adverse events are collected, and those regarded as relevant (i.e. related to food supplements, sufficiently documented, and with at least a likely causal link between consumption and outcome) are classified according to severity by applying the French Nutrivigilance system, which grades adverse events in four levels from 1 (mild) to 4 (lethal outcome), with Level 2 being intermediate effects and Level 3 corresponding to severe effects. Only those in Levels 3 and 4 are considered in this freshly-published report.

Then a literature search is performed to identify further ingredient-adverse effect connections. Articles are then sorted according to defined inclusion / exclusion criteria, and are subsequently allocated in one of these three scoring tiers: B0 (not documented); B1 (poorly documented, comprehending isolated clinical cases not supported by pathophysiological data, or animal studies); and B2 (well documented, including clinical or epidemiological studies or cases supported by pathophysiological data). The analysis published recently specifically targets the B1 score, as ingredients in that group are interpreted as potential emerging risks.

Finally, those ingredients within the B1 category are ranked on the basis of three criteria:

• The number of related cases:

  • 1 case report = 0 point

  • >1 case report = 2 points

• The number of related publications:

  • 1 publication = 1 point

  • 2 publications = 2 points

  • 3 publications = 3 points

  • >3 publications = 4 points

• The existence of Human data

  • Yes = 2 points

  • No = 0 point

Considering the vigilance signals collected in Belgium, Denmark, France, Italy, Netherlands and Portugal, ingredients ranked B1 and with a high score of concern in this report, together with the reasons for that concern, were:

Some of these ingredients with highest Score (5 or 6) are already in different stages of scrutiny, restriction or prohibition. Thus the main outcome of this exercise was the flagging of the six ingredients mentioned in the beginning, including Bifidobacterium, which scores 6 for infectious endocarditis based on:

• A single case report, which adds 0 points in the ranking methodology

• More than three articles, which adds 4 points

• The existence of human data, which adds 2 points

Plant/ Component	Adverse effects	Number of case report (1=0, >1 =2)	Number of articles (1=1, 2=2, 3= 3, >3=4)	Human data (yes = 2, no = 0)	Final score
Bifidobacterium	Infectious endocarditis	0	4	2	6

With this one may wonder, Bifidobacterium and infectious endocarditis? To understand this, one needs to first know that this system uses the term infectious endocarditis as a synonym of bacterial infection (which may be a bit confusing). These effects are, understandably, regarded as Level 3 grade (severe effects). The scientific report cites 6 documents that led to the flagging of Bifidobacterium.

The first of them (ANSES, 2019) is the case report that in fact put Bifidobacterium on the list, and describes the occurrence of an infectious endocarditis attributed to probiotic consumption in a 66-year old woman in France in 2019; the second one (Esaiassen et al., 2017) is a series of 15 case reports of bacteremia caused by Bifidobacterium species; the third one (Skiest et al., 1994) includes an individual case of anaerobic pericarditis in a 63-year-old woman in the US; and the remaining three (Brook, 2009; Brunser, 2017; Costa et al., 2018) are review papers that cite 13 different articles comprehending a total of 33 individual cases of infections and complications caused by different species of Bifidobacterium in multiple parts of the world. Taken together, it may seem that a total of 6 documents, comprehending 50 individual cases, led to the hint that Bifidobacterium may cause infectious endocarditis (or bacterial infections).

The problem with this conclusion arises when those publications are analyzed in more detail.

First, the apparent 49 cases discussed in 5 of the 6 documents (namely Skiest et al., 1994; Brook, 2009; Brunser, 2017; Esaiassen et al., 2017; and Costa et al., 2018) included in the scientific report overlap significantly. In fact, Brook, 2009 only mentions Skiest et al., 1994 as a report of anaerobic pericarditis caused by Bifidobacterium (actually, the only mention to Bifidobacterium in Skiest et al., 1994 comes from another publication, namely Epstein et al., 1992). Similarly, 10 of the 16 cases discussed in Costa, 2018 are also discussed in either Brunser, 2017 or Esaiassen et al., 2017. Also importantly, in all 4 cases of endocarditis included in Costa et al., 2018, the microorganisms isolated from the infected tissue were not members of Bifidobacterium. Similarly, and more strikingly, the case brought up by ANSES, 2019 describes an infectious endocarditis caused by (possibly) Lactobacillius larvae, L. paracasei and/or L. rhamnosus, even if it is mentioned that the person affected was taking probiotics (brand names included, which allows to trace them back to the specific strains) based on both Lactobacillus and Bifidobacterium.

With this, what seemed to be 50 cases of complications with Bifidobacterium goes down to 34.

Is 34 cases a lot? This depends on different things, like the temporal and the geographical context. Regarding time, it does indeed not seem like a lot when one knows that 9 of the 16 cases described in Brunser, 2017 come from Bourne et al., 1978, which studied samples collected between 1972 and 1977. Indeed, Costa et al., 2018’s Pubmed’s review looks all the way back to 1946, and up to 2018. The latest document in the external scientific report’s literature dates from 2019 – so, we are talking about between 47 and 72 years in total depending on how one wants to look at it. Concerning geography, studies involve cases from the US to Japan, both included, and many things in between (traveling East) – a lot of person-years involved.

Regarding clinical severity, in these 34 individual cases, deaths were reported in less than 1/4 of occasions. Even if foods should cause no adverse events (let alone deaths), it is also worthwhile to look at the background health status of the individuals involved. The most significant result in this regard is that 11 of the 35 cases involved newborns between 10 and 46 days of age (average 18 days), with all of them being pre-term and/or reported as suffering from conditions typical of pre-term birth. In adults, all 24 cases had risk factors, from recent delivery, to bowel obstructions, to metastatic cancer. In the case described in ANSES, 2019, the person’s BMI was of 12.2 kg / m2 (severely underweight).

The precise clinical conditions emerging from these cases are well described, with bacteremia leading by far (85% of cases), followed by sepsis (9%), and septic arthritis and pericarditis accounting for one case each. As mentioned earlier, the work’s methodology may cluster bacteremia, sepsis and pericarditis together as infectious endocarditis / bacterial infection.

Another key factor to consider is whether the bifidobacteria involved in these cases originated from foods or supplements containing them or not. Even if probiotic consumption was not explicitly reported in 30% of cases, and that confirmation through molecular means was not performed in almost 70% of the reports, it can be inferred that probiotic products being consumed by affected people should be involved in 40-50% of those 34 cases. Importantly, the conclusion in 20% of those 34 cases is that the patients were indeed consuming probiotics containing bifidobacteria, but in turn bifidobacteria were not recovered as the causal agents from their tissues. For instance, as discussed above, in ANSES, 2019 the person involved was taking two probiotic food supplements containing six different species of bifidobacteria and lactobacilli (eight different strains in total). However, three species only of the genus Lactobacillus were found in their tissues, and only one of them (at the species level) matched any of the species included in the food supplements they were taking. And surprisingly, the same case that Epstein et al., 1992; Skiest, 1994; and Brook, 2009 deal with determines the cause of the polymicrobial (including Bifidobacterium spp.) infection as a iatrogenic mediastinitis secondary to transbronchial needle aspiration of a subcarinal mass – in other words, microbes got in the wrong place after an accident during a medical procedure, not because the patient was taking any food supplement.

It is also noteworthy that, in 9 of those 34 reports (25%) where Bifidobacterium was identified in the affected tissue, other bacteria or fungi (such as Bacteroides fragilis, Bacteroides ureolyticus, Streptoccus oralis, Fusobacterium nucleatum, Escherichia coli, Veillonela parvula or Candida glabrata) were also isolated simultaneously.

In summary, if one attends only to unique / not duplicated cases where 1) Bifidobacterium was unequivocally, and in monoculture, responsible for any type of adverse event, and 2) it is reasonable to think that the Bifidobacterium causing the problem may have come from food consumption (just oral route meets the methodology’s inclusion criteria), only 7 individual cases remain. These cases are in fact amongst the mildest recorded, as they resulted in no deaths and involved one case of sepsis, five cases of bacteremia (which is generally milder than sepsis), and a case of septic arthritis which happened in a patient who was taking a Bifidobacterium-containing supplement and that had undergone a hip arthroplasty (which makes one think whether the route of infection was oral or otherwise). No cases of endo or pericarditis in these 7 reports. These 7 cases are described in 3 of the publications included in the report (Costa, 2018; Brunser, 2017; Esaiassen, 2017). Attending to the report’s methodology, this should take Bifidobacterium’s Score down to 5 from the current 6.

Crucially, it should not be forgotten that the case report that put Bifidobacterium on the shortlist in the first place is ANSES, 2019 which, as extensively discussed above, describes a case of endocarditis caused only by lactobacilli in a person taking two different food supplements containing both lactobacilli and bifidobacteria. If the case report that makes Bifidobacterium qualify to be flagged for risk of infectious endocarditis does not describe a situation in which Bifidobacterium is causing endocarditis (or in any infection, for that matter), should Bifidobacterium really be on this shortlist?

Finally, one may wonder if outside these quantitative metrics, the authors of those 6 documents cited in the external scientific report gave any qualitative impression of concern about Bifidobacterium. Put differently, does all this discussion contradict what the authors of the papers say? Here the answer is again negative. Those documents cited have, more directly or indirectly, the spirit to induce caution about the use of probiotics, which are otherwise regarded as exceptionally safe. However, ANSES, 2019 estates that to date, no clinical cases of endocarditis involving the consumption of probiotics containing Bifidobacterium spp. have been found in the literature. On its end, Brunser, 2017 reminds that Bourne et al., 1978 showed that only 10 out of 91,493 blood cultures (0.01%) over a 7-year period were positive for bifidobacteria. It does also estate that the episodes of sepsis due to Bifidobacterium described in the medical literature are few and that it is striking that the frequency of Bifidobacterium sepsis remains so low that these episodes become almost anecdotal phenomena.

All in all, the publication of this external scientific report should not translate in a change of the current regulatory status of Bifidobacterium in food supplements in Europe. First, because the inclusion of an ingredient in a shortlist does not necessarily mean that it will be restricted or prohibited, despite previous cases in which this has been the case. Second, because the careful review of the documentation shows that bifidobacteria as a group, as well as their presence in foods, are safe. In any case, authorities should, by all means, continue to monitor Bifidobacterium and all other food ingredients in the interest of public health.

The inclusion of Bifidobacterium on the shortlist should be contextualized with the methodology employed to generate that shortlist in the first place, which is designed to identify safety risks with a high degree of sensitivity. Since the ranking methodology pays attention to the number of documents rather than to the number of cases, individual case reports may apparently count as wider epidemiological studies (although they will be properly scored accordingly, see above).

Lastly, the nomenclature employed has generated a great deal of confusion, as infectious endocarditis is used as a synonym of bacterial infection. However, the reason for inclusion of the genus on the shortlist (infectious endocarditis) has never been described as a risk as such, with an individual case of polymicrobial pericarditis (again, caused by an invasive medical procedure) as the closest reference.

References

1.      https://efsa.onlinelibrary.wiley.com/doi/10.2903/sp.efsa.2026.EN-10226

All other references are available from the one listed here.