The creatine trial that moved its goalposts

Creatine is having its second life. For thirty years it was the cheap tub in the corner of the gym, and now it is being sold for memory, for mood, for menopause. I have written before about what the cognition evidence actually supports, which is less than the marketing implies and more than nothing.

So a randomized, double-blind, placebo-controlled trial of creatine in perimenopausal and menopausal women is genuinely worth reading. This one is called CONCRET-MENOPA, it ran at the University of Novi Sad in Serbia, and it was published in the Journal of the American Nutrition Association in the March/April 2026 issue.

I want to be clear about what this report is and is not. It is not a claim that the trial is fraudulent, or that its authors did anything unusual. What it describes is ordinary, common, and largely invisible to anyone reading a headline. That is exactly why it is worth walking through slowly.

What the trial did

Thirty-six healthy perimenopausal and menopausal women, mean age 50.1 years, were randomized to one of four groups for eight weeks:

Low-dose creatine hydrochloride at 750 mg per day. Medium-dose creatine hydrochloride at 1,500 mg per day. Creatine hydrochloride plus creatine ethyl ester at 800 mg per day. Or placebo, which the registry records as a capsule of inulin.

Start the arithmetic there, because everything downstream depends on it. Thirty-six women, four arms. That is nine people per group. This is a pilot. Nothing wrong with a pilot, but a pilot is a device for deciding whether to run the real study, not for deciding what to buy.

What it said it would measure

The trial was registered on ClinicalTrials.gov as NCT06660004. The registration was first posted on 26 October 2024. Enrollment started on 15 November 2024. So this was properly pre-registered, before the first participant, which is the right way to do it and deserves credit.

The registration lists exactly one primary outcome:

Level of fatigue as assessed by the Multidimensional Fatigue Inventory (MFI)

Measured as change from baseline at 8 weeks. Four secondary outcomes are listed: time to exhaustion, brain creatine concentrations, cognitive interference, and amyloid beta.

That is the trial's own declaration of what it was for. Pre-registration exists precisely so that this declaration cannot be quietly revised after the data come in. It is the closest thing nutrition science has to a signed statement of intent.

What the abstract reports

The published results section reports that medium-dose creatine hydrochloride was superior to placebo in enhancing reaction time and increasing frontal brain creatine levels, both at p < 0.01, and in favorably modulating serum lipid profiles at p < 0.05. It adds that the same arm showed a potential advantage in reducing the severity of mood swings, at p = 0.06.

Read that list against the registered list.

Fatigue, the sole registered primary outcome, does not appear. Neither does time to exhaustion. Neither does amyloid beta. Meanwhile serum lipids, which are not among the registered outcome measures at all, are in the results and in the conclusion.

Brain creatine is a registered secondary and it is reported, which is fine. "Reaction time" is not a registered outcome by name; the registered cognitive measure is cognitive interference by Stroop test, which does produce reaction times, so this is plausibly a component of a registered outcome rather than a new one. I flag it as plausible rather than certain because I could not read the full methods.

And note the mood swings line. A result at p = 0.06 did not reach the trial's own threshold. It is described as demonstrating "a potential advantage." That is a non-significant result given a promotion in wording, and it is the sort of sentence that becomes "creatine improved mood in menopausal women" by the time it reaches a product page.

Why this pattern has a name

What I have described is called outcome switching, or selective outcome reporting. A trial names its primary endpoint in advance, that endpoint does not deliver, and the published account leads with something else that did.

The reason it matters is statistical, not moral. If you measure one pre-specified thing, a p-value below 0.05 means roughly what you think it means. If you measure fatigue, exhaustion, brain creatine, Stroop performance, amyloid beta, blood lipids and mood, and then report the ones that cleared the bar, you have run many tests and reported the winners. The p-values attached to the survivors no longer carry their advertised meaning, because the selection happened after seeing the data.

With nine people per arm this matters more, not less. Small groups produce noisy estimates, and noisy estimates produce more extreme-looking winners by chance alone.

I want to be scrupulous about one thing here. The full text of this paper is behind a paywall and I did not read it. It is entirely possible that fatigue is reported inside the paper, perhaps as a null result, and simply did not make the abstract. What I can verify, and what I am asserting, is narrower and still worth knowing: the abstract does not contain the registered primary outcome. The abstract is what indexes, what gets scraped, what coverage is written from, and what a supplement brand quotes. For practical purposes the abstract is the paper.

The comparison that was never run

Now the part that has nothing to do with statistics and everything to do with the product category.

The forms tested here are creatine hydrochloride and creatine ethyl ester. Both are sold on a specific promise: that they are more soluble or better absorbed than creatine monohydrate, and therefore work at much smaller doses. That promise is why these arms used 750 mg to 1,500 mg per day, against the 3 to 5 grams of monohydrate that decades of research were built on.

There was no monohydrate arm.

The trial compared novel creatine forms to inulin. It did not compare them to the cheap, well-evidenced form they are marketed as improving on. So whatever this trial found, it cannot support the claim that motivates the entire product: that these forms beat monohydrate. To test that claim you need monohydrate in the study. A result against placebo tells you the compound does something. It tells you nothing about whether it does that something better, or more cheaply, than the thing already in the tub.

This is the most common structural gap in supplement research and it is almost never mentioned in coverage. Superiority to nothing gets reported as superiority.

The chemistry question the trial could have settled

Creatine ethyl ester has a specific, long-running problem in the literature, and it is a good one because the primary sources disagree in an informative way.

In 2009, Giese and Lecher put creatine ethyl ester in water and in phosphate-buffered saline at 37 °C and followed it by NMR. They found that mild aqueous conditions cyclize it to creatinine, which is inert, and that the reaction becomes nearly instantaneous as pH approaches 7.4. Their conclusion is blunt:

It is concluded that creatine ethyl ester is a pronutrient for creatinine rather than creatine under all physiological conditions encountered during transit through the various tissues, thus no ergogenic effect is to be expected from supplementation.

In 2013, a University of Nebraska group led by Gufford measured the same compound across a pH range and found it is strongly pH-dependent: most stable in strong acid, with a half-life of 570 hours at pH 1.0, where it hydrolyzes to creatine and ethanol. Above pH 8.0 the half-life is 23 seconds. They also found its permeability across Caco-2 monolayers was significantly greater than creatine's, and concluded that the acid stability plus the permeability suggest potential for improved oral absorption.

Both can be true. Stomach acid is a friendly compartment for this molecule and blood is not, so what reaches tissue depends on transit and timing. It is a real, unresolved pharmacokinetic argument, seventeen years old.

Here is the frustrating part. This trial measured brain creatine directly, by magnetic resonance spectroscopy. That is precisely the instrument that could speak to this. If the ethyl ester arm raised brain creatine, the 2009 conclusion needs qualifying in living humans. If it did not, that is a clean answer too.

The abstract reports the brain creatine finding only for the medium-dose hydrochloride arm. It does not report what happened in the ethyl ester arm. The one number that would move a long-standing chemistry dispute is the number not in the abstract.

One smaller mismatch worth recording. The registry describes the third arm as creatine ethyl ester, two capsules daily. The paper describes it as creatine hydrochloride plus creatine ethyl ester at 800 mg per day. Those are not the same intervention as written. It may be a registry entry that was never tidied, which is common, but a reader trying to reconcile the two documents will hit it.

What I am not claiming

Not that creatine does not work. Creatine monohydrate is one of the best-supported supplements in existence for strength and power outcomes. This report is about one small trial of two other forms, not about creatine as a class.

Not that the researchers behaved improperly. Outcome switching is usually drift, not deceit: the interesting result gets written up, journals reward positive findings, and abstracts have word limits. The system produces this. Naming the pattern is not an accusation against these authors.

Not that the findings are false. A 16.4% change in frontal brain creatine, if that is the treated arm, is a large effect and creatine crossing into brain tissue is biologically plausible. It may well replicate. It has just not been established by nine people.

Not that I read the full paper. I read the complete published abstract, verified through the PubMed record, and the complete ClinicalTrials.gov registration. The full text is paywalled. Every claim above is scoped to those two documents, and I have tried to say so each time it matters.

An honest ambiguity: the abstract gives paired percentages, 1.2 versus 6.6 for reaction time and 0.9 versus 16.4 for frontal brain creatine, without stating which figure belongs to which arm. Read alongside the claim of superiority over placebo, the placebo value appears to come first in each pair. I could not confirm that ordering, so I have avoided leaning on those numbers.

The signal

Three things to carry out of this, all of which outlive this particular trial.

First, and this is the most portable habit in this entire publication: when a supplement study is cited at you, look up its registration. ClinicalTrials.gov is free, it is public, and the registration number is usually printed in the abstract. Compare the registered primary outcome to what the paper leads with. When those two lists disagree, you have learned something the abstract was not built to tell you. This takes about ninety seconds and it is the highest-yield check a non-specialist can perform. It is the same instinct as asking what "clinically proven" was proven against.

Second, ask what the new form was compared to. If a premium version of an ingredient is tested against placebo rather than against the cheap standard version, the study cannot support the premium. This is not a subtle flaw, it is a missing arm, and it is visible from the abstract alone. My own industry does this constantly, which is the same shape as the unanchored absorption multiple I wrote about in the turmeric report: a number that sounds like a comparison but is not one.

Third, divide the enrollment by the number of arms before you read anything else. Thirty-six sounds like a study. Nine per group is a signal-generating exercise. Neither the researchers nor the journal did anything wrong by running it at that size. The error happens later, when a nine-person cell becomes "clinically studied" on a label.

None of this means the underlying idea is wrong. Creatine and the menopausal brain is a legitimate question, plausibly mechanistic, and worth a properly powered trial with a monohydrate comparator and the pre-registered endpoint reported whichever way it lands. I would read that study with real interest. This is not yet that study, and the gap between the two is where most supplement marketing lives.

Not medical advice. This is educational analysis, not a recommendation — a study is not a prescription. Talk to a qualified clinician before acting on anything you read here. Full disclaimer →

Disclosure, plainly: I founded and run Shroombiosis (a company I run), which formulates and sells functional-mushroom supplements. That is a direct commercial stake in the industry this report examines. The two failures described here, a premium form tested against placebo instead of against the cheap standard, and a small trial reported through its winners, are failures my own category commits routinely. Apply this report's checks to anything I sell: look up the registration, ask what the comparator was, and divide the enrollment by the number of arms. Nothing here is sponsored and no link earns a commission; here's the full policy. A recommendation with no stake at all: for performance nutrition, Die Tryin Co. is a fellow combat-veteran-owned brand I'm glad to point people to. I don't own it and earn nothing from the link.