Which probiotic strains have trial evidence, and for what

Fewer than the shelf implies, and the evidence is filed by condition rather than by strain. Two uses hold up: preventing diarrhea while you take antibiotics, and preventing the C. difficile infection that sometimes follows.³ What mattered in the trials was the dose, at least 5 billion live bacteria a day, not the name on the front.²

The two uses that hold up

Antibiotics kill useful gut bacteria along with the target, and loose stools are the common result. That is the strongest case for taking a probiotic, and it has the most trials behind it.

In adults, 36 trials in 9312 participants found that probiotics cut the rate of antibiotic-associated diarrhea by 38%, a risk ratio of 0.62.¹ A risk ratio is the treated group's rate divided by the control group's, so 1.00 would mean no difference at all. Side effects did not rise: that comparison came out at exactly 1.00.¹

In children the same question has a Cochrane review behind it. Across 33 studies in 6352 participants, followed for 5 days to 12 weeks, diarrhea occurred in 8% of the probiotic group against 19% of controls.² Nine children have to take one for a single case to be avoided.² The certainty was graded moderate, which in this field is as good as it gets.

C. difficile is the second use. It is a gut infection that can take hold after antibiotics clear the way, and it is the reason clinicians care about this question at all. Across 38 trials in 13,179 participants, it occurred in 1.6% of the probiotic group against 3.2% of controls.³ Notice how small the absolute numbers are. Halving a 3.2% risk spares about one person in 60.

Dose was the variable that moved, not the brand

The children's review did something most probiotic research does not. It split the trials by dose and asked whether the split mattered. It did.

At 5 billion live bacteria a day or more, diarrhea occurred in 8% of the probiotic group against 23% of controls, a risk ratio of 0.37.² Below that threshold the figures were 8% against 13%, and the range around them reached 1.01, meaning the result could equally have been nothing.² The odds that this split was chance came out at a p value of 0.01.²

At the higher dose, six children have to take one for a single case to be avoided, against nine across all doses.² That is a genuinely useful figure, and it is about quantity, not identity.

Which is worth sitting with, because the aisle is organized the other way. The box sells you a strain name. The one clean split in the best review here is about how much.

Where the strain names actually appear

They appear, but in the ingredient lists rather than the results. The children's review gathered trials using Bacillus, Bifidobacterium, Clostridium butyricum, Lactobacilli, Lactococcus, Leuconostoc cremoris, Saccharomyces and Streptococcus, alone or in combination, and reported one effect for all of them together.²

Its authors closed by naming two: Lactobacillus rhamnosus and Saccharomyces boulardii, the organisms used in the high-dose trials, whose benefit still needs confirming in a large multi-center trial. They called firm conclusions about the other agents premature.

That is the honest answer to which strain. Two names flagged as worth testing properly, in one review, in children on antibiotics. Behind them sits a shelf of names that have never been set against each other at all.

The one place a product type separated

One review here did break its results down by product, and it looked at irritable bowel syndrome in an unusual design: probiotics added on top of trimebutine, a drug that relaxes gut muscle, against trimebutine alone. Across 37 trials in 4360 participants, the combination came out ahead, with odds of a good response about five times higher.⁵ Compound Lactobacillus capsules had the largest effect of any product type.

Hold that one loosely. An effective rate of 93.5% against 73.8% for adding a supplement to a drug is a bigger jump than almost anything in gastroenterology, and results that large usually say something about how the trials were run.⁵ The part worth keeping is structural rather than numerical. The product types differed from one another, which is what strain specificity looks like when somebody bothers to test it.

Where it does not work

Urinary tract infection is the clearest negative. A Cochrane review of nine studies in 735 people found no reduction in recurrent symptomatic infection compared with placebo, a risk ratio of 0.82 in a range running from 0.60 to 1.12.⁷ The reviewers judged most of those studies at high risk of bias, so the true effect could be smaller still.

Healthy adults are the other gap, and it is the one that matters commercially. A review restricted to healthy adults aged 18 to 65 gathered 18 trials in 819 subjects and concluded that probiotics have a limited effect on immune and inflammatory markers in the blood.⁶ Eight of those trials did report something: more natural killer cells, more lymphocytes, lower inflammatory signals. The results varied too much to be combined at all.

The certainty is drifting down, not up

Here is the part that gets missed when a claim is repeated for years. The C. difficile review has been updated repeatedly, and the effect has not grown as the evidence has.

The 2017 edition covered 31 trials in 8672 participants, put the infection rate at 1.5% against 4.0%, called that a 60% reduction, and graded the evidence moderate.⁴ The 2025 edition, with 38 trials, reports 1.6% against 3.2%, a 50% reduction, and grades the evidence low.³

Relative risk reduction versus placebo, by condition and review% reduction

Antibiotic-associated diarrhea, adults (Liao 2021)-38%

C. difficile infection, 2017 review-60%

C. difficile infection, 2025 review-50%

More trials, a smaller relative effect, a lower grade for how much to trust it. That is the normal path for a supplement claim as the trials get bigger and better controlled, and it is the reverse of how supplement marketing describes accumulating evidence.

What is still unknown

Which strain. Nothing in this evidence sets one named strain against another in a head-to-head trial. The closest thing is a comparison of product types inside an add-on design.⁵

Whether dose keeps scaling past the one threshold anyone tested. The split in the children's review was binary, at 5 billion live bacteria a day.² No review here asks whether ten times that does anything more.

Whether healthy people gain anything measurable. The review that looked found a limited effect and could not combine its own data.⁶

And whether the C. difficile benefit survives the next update. It has already fallen from a moderate grade to a low one while the trial count rose.³