Does Creatine Cause Hair Loss? Tracing the Claim to Its Source

Some claims survive because nobody checks them and some survive because checking them is hard. This one survived for sixteen years on a single measurement in twenty rugby players, and the measurement was not of hair. This is research journalism and contains no usage guidance, per our editorial standards.

The original study, in full

Everything downstream comes from one paper: van der Merwe, Brooks and Myburgh, Clinical Journal of Sport Medicine, September 2009 (PMID 19741313).

The results were clear and, on their own terms, statistically strong. Total testosterone did not change. DHT rose 56% after seven days of loading and remained 40% above baseline after fourteen further days of maintenance (P < 0.001). The DHT:T ratio rose 36%, remaining 22% elevated (P < 0.01).

The authors' conclusion is worth quoting exactly, because it is more modest than what was built on it: creatine supplementation "may, in part, act through an increased rate of conversion of T to DHT. Further investigation is warranted."

The step that was never taken

DHT is the androgen implicated in androgenetic alopecia — that is why finasteride, which inhibits its production, treats male-pattern hair loss. So a rise in DHT is a coherent place to start a hypothesis about hair.

It is not a finding about hair. Between "a hormone associated with a condition rose" and "the condition occurs" sit a magnitude question, a duration question, a scalp-tissue question and a susceptibility question, none of which a three-week serum measurement in twenty athletes can address. The 2009 paper did not claim otherwise. The internet did.

The paper also drew a formal response: a comment and an author reply published in the same journal in 2010 (Clin J Sport Med 2010;20(3):220). A single small crossover study attracting a published critique is ordinary scientific process, and it is another sign that the result was treated as provisional by the people reading it closely.

Sixteen years later, somebody measured the hair

In 2025 a group across Iran, Canada and the United States ran the study the claim always needed (Lak et al., Journal of the International Society of Sports Nutrition 2025;22(sup1):2495229, PMID 40265319).

Forty-five resistance-trained males aged 18–40 were randomised to 5 g/day creatine monohydrate or 5 g/day maltodextrin placebo for twelve weeks, maintaining their habitual diets and training. Thirty-eight completed. Blood was taken at baseline and week twelve for total testosterone, free testosterone and DHT. Hair follicle health was assessed with the Trichogram test and the FotoFinder system, giving hair density, follicular unit count and cumulative hair thickness.

There were no group-by-time interactions for any hormone or hair outcome (p > 0.05). Total testosterone rose and free testosterone fell across the study in both arms — creatine ∆124 ± 149 ng/dL against placebo ∆216 ± 203 ng/dL for total testosterone — but those movements were independent of supplementation. DHT, the DHT:T ratio and every hair parameter showed no significant difference between groups.

The authors describe their result as "strong evidence against the claim that creatine contributes to hair loss."

Now the part a reader is entitled to

That sentence comes from a paper published in the journal of the International Society of Sports Nutrition, and the conflict-of-interest statement printed with it is substantial. One author is the chief executive and co-founder of the ISSN, which the statement describes as an academic non-profit that may be sponsored by companies that manufacture, market and sell dietary supplements, and is a scientific advisor to several named creatine brands. Another is a scientific advisor to a creatine brand and has received creatine donations from a manufacturer for research. A third owns a consultancy providing paid services to dietary supplement manufacturers.

None of that makes the measurements wrong. Trichogram and FotoFinder counts are not opinions, and a null result is not the outcome a sponsor's marketing department would commission. But this site applies the same rule to a supplement it likes as to a compound it does not: who funded the study is part of the study, and we have said the same about the creatine efficacy meta-analyses, whose authors disclose the same kinds of relationship.

The other limits are ordinary and matter more. Thirty-eight completers is a small sample for a follicle endpoint. Twelve weeks is short for a process measured in years. The participants were healthy men aged 18–40, not men with diagnosed androgenetic alopecia, who are the population the worry is actually about. A null result in that group does not establish a null result in a man already losing hair.

How much literature is there, in total?

Very little, which is the most surprising thing on this page. A PubMed search on 2026-09-11 for creatine together with dihydrotestosterone or DHT returns 33 records of any kind — every species, every design, including papers where the two terms merely co-occur. Restricting a creatine-and-hair search to clinical trials returns twelve records, of which eleven are cancer, dermatology or neurology trials that mention creatine incidentally; the twelfth is the 2025 study above.

A claim that has shaped the purchasing decisions of a very large number of people rests on a body of work that one person could read in a day.

What can honestly be said

• The hair-loss claim originated in a hormone measurement, not a hair measurement, in 20 men, over three weeks.
• That measurement was real: DHT did rise, substantially, and the paper is sound on its own narrow terms.
• The inference from it was never tested for sixteen years.
• When it was tested, in 38 men over 12 weeks with objective follicle imaging, no effect on hair or on DHT was found.
• That test is one study, industry-adjacent, short, and in the wrong population to close the question for men who are already balding.

Against a large body of trials finding no excess of side effects overall, the hair question now looks like what it probably always was: a plausible mechanism that turned into a certainty on the way to the forums, and shrank back to a hypothesis as soon as anyone measured the outcome it was about.

Sources and dates

All PubMed searches run 2026-09-11 via E-utilities.

Figures are reported as the sources state them. Nothing here is a recommendation or an instruction.