Probiotics and Mild Cognitive Impairment: A New Trial, and a Correction to Our Own Note

Abstract

A trial recruiting at the Complutense University of Madrid is giving 100 people with mild cognitive impairment a probiotic capsule or a placebo for sixteen weeks, and measuring the result with an electrophysiological study rather than a bedside questionnaire.

When we filed this topic we described it as filling a gap: a rich mechanistic literature on the gut–brain axis with almost no human intervention data behind it. That was wrong, and the correction is the most useful part of this note. Thirty-four randomised trials have already been pooled.

What those trials produced is stranger than an absence of evidence. They produced a large effect with low certainty, and two instruments measuring the same thing disagreeing by more than threefold.

1.The Trial

Registry entry NCT06670807 is a randomised, parallel-group, triple-masked trial sponsored by the Universidad Complutense de Madrid, in collaboration with AB Biotics SA. It began recruiting on 21 March 2026 and is scheduled to complete in May 2027.

One hundred participants with mild cognitive impairment take one capsule daily before breakfast for sixteen weeks — either a probiotic formulation designated AB-MIND with the strain CECT7481, or placebo. The collaborating company makes the product under test, which is normal for this kind of trial and worth knowing before the results appear.

The primary outcome is a single line in the registry and it is the reason this trial is worth watching: change in cognitive function assessed by electrophysiological study, from enrolment to the end of treatment at sixteen weeks.

Not the Mini-Mental State Examination. Not the Montreal Cognitive Assessment. An electrophysiological measurement, which is to say something recorded from the scalp rather than scored by an examiner.

2.What We Got Wrong

Our own note on this topic said the gut–brain axis has a rich mechanistic literature and a very poor interventional one in humans, and that this trial aims precisely at that gap.

The first half is right. The second is not. A systematic review and meta-analysis published in BMC Complementary Medicine and Therapies in November 2025 by Nancy Calzada-Gonzales, Joshuan Barboza and colleagues identified 34 randomised clinical trials with 2,390 participants testing probiotics against cognitive outcomes in adults.

That is not a gap. It is a substantial body of randomised human evidence, and the reason it feels like a gap is that almost nobody quotes what it actually found.

3.The Number That Does Not Fit

Pooling the trials that used the Mini-Mental State Examination at twelve weeks, the analysis found a mean difference of 4.23 points in favour of probiotics, with a confidence interval from 2.77 to 5.68 and heterogeneity of zero.

The MMSE runs from 0 to 30. Four points is not a subtle shift on it — it is roughly the distance between categories of impairment. A supplement producing that in twelve weeks would be the most effective cognitive intervention in existence, comfortably ahead of anything licensed for dementia.

Now the same analysis, same trials, same question, using the Montreal Cognitive Assessment: a mean difference of 1.21 points, confidence interval 0.06 to 2.36. And on the categorical fluency test, 3.94 points, interval 3.20 to 4.69.

The MMSE and the MoCA are both brief screening instruments for global cognition, designed to measure substantially the same construct, and here they differ by a factor of more than three. That is not a finding about probiotics. It is a signal that something in the underlying trials is not behaving.

The authors do not oversell it. Their GRADE assessment rates the certainty of evidence as low for every cognitive outcome and very low for adverse events, and their conclusion states that although probiotics show potential benefits, the current evidence remains highly uncertain. Their own abstract contains the phrase "the evidence is very uncertain".

That is a meta-analysis reporting a very large effect and telling readers not to believe it yet. The number will travel; the caveat will not.

4.Where a Result Like That Comes From

Heterogeneity of zero across dozens of independent trials is worth pausing on, because intuition gets it backwards. Perfect consistency between studies of different strains, doses, durations and populations is not reassurance. Genuinely independent trials of a modest intervention disagree with each other.

The likelier explanations are structural. Small trials, run in similar settings, using unblinded or partially blinded examiner-scored outcomes, and published when positive. The MMSE is administered by a person to a person; if either knows the allocation, the instrument is porous in a way an electrode is not.

A second meta-analysis, published in Brain and Behavior in November 2025, pooled seven randomised trials of probiotics in patients with depression and reported improvement in cognitive symptoms with a standardised mean difference of −0.90. Its heterogeneity was 82 per cent — the opposite extreme, and in a way the more honest number, because it says plainly that these trials do not agree.

5.The Mechanism Is Not the Weak Point

None of this is an argument that the gut–brain axis is fanciful. The mechanistic literature is genuinely detailed, and the 2024 review by Loh and colleagues in Signal Transduction and Targeted Therapy sets out routes that are specific rather than hand-waving.

Short-chain fatty acids produced by gut bacteria act as ligands for G-protein-coupled receptors and inhibit histone deacetylases, which is a direct route from microbial metabolism to gene expression. Trimethylamine N-oxide, another microbially derived metabolite, promotes microglial activation and neuroinflammation. Bile acids such as UDCA and TUDCA act on the farnesoid X receptor and on TGR5 in microglia and neurons.

These are real pathways with named receptors. The problem has never been whether the gut can signal to the brain. It is whether swallowing a particular capsule of particular strains for sixteen weeks changes a person's cognition — and that is a question mechanism cannot answer on behalf of a trial.

6.Why the Madrid Endpoint Matters

Seen against that background, the choice of an electrophysiological primary outcome stops looking like a technical detail and starts looking like the point of the trial.

The existing evidence is 34 trials of examiner-scored screening tests producing an effect too large to be plausible and too consistent to be natural. The most useful thing a new trial can do is not add a thirty-fifth MMSE score. It is to measure something that does not depend on who is holding the clipboard.

That also raises the bar. Electrophysiological measures are less susceptible to expectation, and they are correspondingly harder to move. A trial that finds nothing on an evoked-potential measure will not have refuted the MMSE literature, but it will have made it considerably harder to explain.

One limitation is fixed in advance: 100 participants and sixteen weeks. This will not settle whether probiotics affect the trajectory of mild cognitive impairment, a condition measured in years. It can only show whether something measurable happens at all.

7.How We Would Read It

Which electrophysiological measure, specified how? The registry says "electrophysiological study" without naming the parameter. Event-related potentials offer many components, and a paper that reports a significant change in one of them without a pre-specification is doing something different from a paper that named it in advance.

Do the secondary cognitive scores move with it? If the electrophysiology shifts and the questionnaires do not, that is interesting and hard to interpret. If the questionnaires shift and the electrophysiology does not, the reader has learned something about the questionnaires.

And is the company's role stated in the paper as clearly as it is in the registry? AB Biotics appears as a collaborator on the registry entry. That is a disclosure made in advance, which is the good version. It should reappear in the publication.

Editorial Comment

MindHeaven® uses no probiotics in any formulation and makes no claim about them. We have no commercial interest in this trial's outcome.

We are publishing a correction to our own research note as the substance of this piece because the error we made is the one the whole category runs on. "Promising mechanism, evidence still to come" is a comfortable thing to say about an ingredient. It is often false in a specific way: the evidence has already come, it is mediocre, and the mechanism is being recited because the trials are not worth quoting.

Thirty-four randomised trials exist. They report a four-point MMSE gain that nobody in the field appears to believe, including the authors who pooled it. That is a more interesting state of affairs than an absence of data, and it is the state of affairs we should have described the first time.

Mild cognitive impairment is a clinical diagnosis and a predictor of dementia in a proportion of those who receive it. Nothing here is guidance about managing it. If memory changes have become noticeable enough to worry about, the assessment worth having is a medical one — because some causes of cognitive decline are treatable, and identifying them matters considerably more than any supplement currently marketed for the condition.