Probiotics

Esmaeilinezhad et al. (Cochrane Database of Systematic Reviews, 2025; PMID 40931979) pooled 38 randomized trials in 13,179 participants and found fewer cases of C. difficile-associated diarrhea when probiotics were given with antibiotics (RR 0.50), about one case prevented for every 65 people treated, and rated that evidence low certainty, downgraded for few events and publication bias. The earlier version, Goldenberg et al. (Cochrane Database of Systematic Reviews, 2017; PMID 29257353), had rated it moderate. The largest single trial, PLACIDE (Allen et al., Lancet, 2013; PMID 23932219), randomized 2,981 older inpatients to a lactobacilli and bifidobacteria preparation or placebo and found no significant reduction (12 versus 17 cases; RR 0.71, 95% CI 0.34 to 1.47). Smaller positive trials, Hickson et al. (BMJ, 2007; PMID 17604300) and Gao et al. (American Journal of Gastroenterology, 2010; PMID 20145608), each tested a different product. WHY C: probiotics are not one intervention, the trials that found a benefit tested different strains, and the largest trial found none, which is a replication that failed. The pooled signal is real enough to record and too uncertain, and too spread across products, to grade higher. Benefit does not generalize to probiotic products as a class.

Tabrizi et al. (Eur J Pharmacol, 2019; PMID 30959049): meta-analysis of randomized trials of probiotic and synbiotic supplementation, reporting changes in inflammatory markers in patients with diabetes. Grade C for this use: the evidence is randomized and pooled, but every endpoint is a surrogate marker, the trials pool probiotics with synbiotics rather than isolating either, and no trial reports a clinical outcome. This is a different and much weaker claim than the C. difficile prevention above, which is why the two are graded separately.

Taverniti et al. (Microorganisms, 2026; PMID 42354894): 140 healthy adults randomized to CRL 1505 or placebo, double-blind, 122 analyzed; episodes defined by study clinicians. The primary endpoint was the share of people with at least one, two or three episodes. Only the three-or-more threshold reached significance, and only at 16 weeks (12.7 percentage points fewer, p = 0.03); one or more and two or more did not. Total episodes were fewer on the probiotic (53 versus 78, p = 0.05). Three of the authors work for the company licensed to make the strain and five for the institute that owns it. WHY C: a single trial, from the strain's owner and licensee, positive on one of its three primary thresholds, that nobody has independently repeated. The result attaches to this one strain at this dose, not to probiotics as a class.