Technical Note

Probiotics in healthy adults: a small signal for mood and sleep, none for anxiety

“Mood, stress and sleep” is one of the fastest-growing positionings for probiotic products, and it is usually sold as a single package. A systematic review and meta-analysis published this week in Pharmacological Research pooled 20 placebo-controlled trials in healthy adults and pulled that package apart: depression scores and sleep quality improved, with moderate-certainty evidence, while anxiety showed no significant effect, with very-low-certainty evidence. For formulators, the size of the effect and the fact that it was pooled across strains matter as much as the direction.

Published 1 November 2026 · 7 min read

What the study did

Design

Systematic review and meta-analysis of randomised controlled trials comparing probiotics with placebo for at least 4 weeks in healthy adults.

Search: PubMed, Cochrane Library, Web of Science and Embase, from inception to 18 September 2026.
Included: 20 studies, 1,508 participants.
Outcomes: depression scores, anxiety scores and sleep quality, pooled as standardised mean differences (SMD).
Analysis: random- or fixed-effects models chosen according to heterogeneity; certainty of evidence rated with GRADE.

Result

Depression scores: SMD −0.23 (95% CI −0.36 to −0.10; I² = 4.67%). GRADE: moderate.
Sleep quality: SMD −0.37 (95% CI −0.52 to −0.22; I² = 41.67%). GRADE: moderate.
Anxiety scores: SMD −0.11 (95% CI −0.31 to 0.09; p = 0.27; I² = 61.41%). GRADE: very low.
Subgroup analyses were exploratory and unadjusted; no conclusion could be drawn about intervention duration or strain type.

The authors' conclusion: probiotic supplementation may alleviate depressive symptoms and improve sleep quality in healthy adults, while its effect on anxiety remains inconclusive. They call for larger, standardised trials to confirm the findings and clarify strain-specific effects.

What it does and does not establish

What it establishes: across placebo-controlled trials in people without a diagnosed condition, probiotics as a category were associated with lower depression scores and better sleep-quality scores, consistently enough for the authors to rate both as moderate certainty. The depression estimate in particular is tight and homogeneous (I² under 5%).

Three points need careful reading:

The effects are small. An SMD is a difference expressed in standard deviations, which is how scores from different questionnaires get pooled. By the conventional benchmarks (0.2 small, 0.5 medium), −0.23 for depression is at the small end and −0.37 for sleep sits between small and medium. The abstract does not name the instruments used, so these figures cannot be turned back into points on any particular scale.

“Depression scores in healthy adults” is not depression. These participants were healthy; what moved was a symptom score within the non-clinical range. The result says nothing about probiotics for depression as a clinical condition, and should not be cited as if it did.

For anxiety, “not significant” is not “no effect”. The anxiety estimate is both heterogeneous (I² 61%) and very low certainty. The honest reading is “not established” — neither shown to work nor shown not to.

One methodological note, which is our observation rather than the authors': choosing between fixed- and random-effects models according to observed heterogeneity is a data-driven choice. Where I² is near zero, as for depression, the two models give similar answers, so it is unlikely to change the picture here. It is still worth knowing when comparing this estimate with meta-analyses that pre-specified a random-effects model.

The decisive detail: the pooled estimate belongs to no strain

The 20 trials used different organisms, and the authors state plainly that no conclusion can be drawn regarding strain type. The SMDs above are therefore an average over a category, “probiotics”, not a property of any one product.

A strain with no trial of its own does not inherit this result. Being the same genus or species as an organism in the pool is not evidence.

Even a strain that was in the pool cannot claim the pooled number. Its own trial may have contributed a larger, smaller or null effect; the meta-analysis averages them.

A useful sense of scale: 1,508 participants across 20 studies is an average of about 75 per trial. A conventional power calculation (two-sided α = 0.05, 80% power) for an effect of SMD 0.25 needs roughly 250 participants per arm. In other words, most individual trials in this field are too small to detect, on their own, an effect of the size the pool suggests, which is why strain-level evidence is so often inconclusive and why it so often gets borrowed from the category. That calculation is ours, not the paper's.

We made the same point from the other direction in our note on probiotics with iron, where a null result from a blend without strain designations could not be used against any defined strain. Evidence follows the strain, whether the pooled result is positive or negative.

A note from an ingredient manufacturer

We supply lactic acid bacteria and bifidobacteria, the groups most probiotic mood and sleep trials draw from. A favourable category-level meta-analysis is exactly the kind of result an ingredient supplier is tempted to borrow. We will not: this pooled estimate does not support a mood, sleep or stress claim for any of our strains, or for anyone else's. We are writing about it because it shows clearly where the category evidence ends and the strain evidence has to begin.

What this means for formulators and brand owners

1. Cite the strain's own trials, not the category

This meta-analysis is good background for explaining why the field is worth investigating. It is not substantiation for a product. Substantiation means a trial of the same strain, at a comparable dose, in a comparable population.

2. Do not bundle mood, sleep and anxiety

“Mood, stress and sleep” reads as one benefit on a label, but in this analysis the three outcomes behaved differently. The anxiety arm of that bundle is the weakest part of the evidence. Language should follow the outcome that was actually measured and supported.

3. Stop at the population that was studied

Healthy adults, symptom scores, at least four weeks. Moving from “healthy adults” to people with diagnosed depression, anxiety disorders or insomnia crosses a population and an endpoint at the same time.

4. If you are designing a trial, size it for a small effect

If the plausible effect is in the SMD 0.2–0.4 range, a 60-person pilot will most likely end up inconclusive. Name the primary outcome and instrument in advance, and power the trial for the effect size the field actually shows, not the one a marketing brief hopes for.

If you are assessing this direction

Where we can help is the strain layer: confirming what an organism is at species and strain level, assembling strain-level documentation and quality-control records, and checking whether a cited study used the same strain, dose and population. We will not use any study to support an efficacy claim for a product, ours or anyone else's.

References

Liu Q, Zhang L, Wang Y, Li J, Li Y. Effects of probiotic supplementation on depression, anxiety, and sleep quality in healthy participants: A systematic review and meta-analysis. Pharmacological Research, 2026. doi:10.1016/j.phrs.2026.109716. PubMed record
Cohen J. Statistical Power Analysis for the Behavioral Sciences, 2nd ed. Lawrence Erlbaum Associates, 1988. (Conventional effect-size benchmarks.)

This note summarises and interprets published research in order to explain how pooling across strains, effect size and study population bound the applicability of evidence. It is not a claim about the efficacy or safety of any product, is not directed at any company or brand, and is not medical or dietary advice. Persistent low mood, anxiety or sleep problems should be assessed by a physician. Findings apply only to the populations, doses, formulations and conditions described in the original source. Permitted claims are governed by the current regulations in the market concerned.