Creatine and Hair Loss: What the Trials Actually Measured

Every few months the same screenshot resurfaces on Reddit and TikTok: a percentage, a hormone acronym, and a caption warning that the tub of creatine in the cupboard is quietly thinning a hairline. The number is real — it comes from a genuine peer-reviewed trial. What the posts never mention is what that trial measured, what it did not, and what happened when a research group eventually ran the study that looks at the hair itself.

Where the claim starts

An expert review published by the International Society of Sports Nutrition traces the vast majority of speculation about creatine and hair loss to a single study Antonio 2021. That study, published in 2009 by a South African group, was a double-blind, placebo-controlled crossover trial in college-aged male rugby players, conducted at a rugby institute during the competitive season, with a six-week washout period between arms van der Merwe 2009. Twenty players volunteered van der Merwe 2009; the review that later dissected the trial reports that sixteen completed it Antonio 2021.

The creatine condition ran seven days of loading — 25 g of creatine monohydrate plus 25 g of glucose per day — followed by fourteen days of maintenance at 5 g of creatine plus 25 g of glucose per day. The placebo condition matched the sugar alone: 50 g of glucose per day for seven days, then 30 g per day for fourteen van der Merwe 2009Antonio 2021. The listed outcome measures were serum testosterone and dihydrotestosterone — DHT, a more potent androgen the body converts from testosterone — with the ratio between them, plus body composition, taken at baseline and after seven and twenty-one days van der Merwe 2009.

Serum testosterone did not change after seven days of loading or after the further fourteen days of maintenance. DHT rose by 56% after the loading week and remained 40% above baseline after maintenance (P < 0.001). The ratio of DHT to testosterone rose by 36% after loading and remained 22% above baseline after the maintenance dose (P < 0.01) van der Merwe 2009. Those are percentage changes from baseline, not absolute concentrations. Free testosterone — the fraction the body draws on to make DHT — was not among the measurements Antonio 2021.

It is worth being precise about what the authors themselves concluded, because both halves matter. They proposed that creatine supplementation "may, in part, act through an increased rate of conversion of T to DHT" — a hypothesis about mechanism, and not a hair finding, since hair was not measured. But they also wrote that further investigation was warranted because of the high frequency of creatine use and "the long-term safety of alterations in circulating androgen composition," and their statement of clinical relevance said that, particularly in relation to dihydrotestosterone, creatine's long-term clinical safety "cannot be guaranteed" van der Merwe 2009. So the trial did raise a long-term androgen-safety question. What it did not raise was baldness. That inference was added afterwards, by other people.

What the trial did not measure

The most important fact about the 2009 study is the simplest one: hair does not appear among its outcome measures. There were no hair counts, no standardised scalp photographs, no follicle assessments, no record of shedding. Supplementation ran for twenty-one days per arm van der Merwe 2009. Male-pattern baldness, by contrast, is characterised in the genetics literature as "a common and highly heritable trait characterized by androgen-dependent, progressive hair loss from the scalp" Heilmann-Heimbach 2017. A three-week hormone measurement in sixteen athletes cannot tell you what a scalp looks like in a decade.

The percentages also need reading carefully, and the ISSN review took them apart in detail. Before supplementation began, DHT was 23% lower in the creatine group than in the placebo group — 0.98 nmol/L against 1.26 nmol/L — so the creatine arm started from an unusually low floor. In absolute terms it then rose by 0.55 nmol/L after seven days of supplementation and by 0.40 nmol/L after twenty-one, while the placebo response drifted down by 0.17 and 0.20 nmol/L over the same periods. That combination of a low starting point, a small rise, and a small fall in the comparison condition is what the review argues produced the headline "statistically significant" increase. Throughout, the review notes, the DHT values and the DHT-to-testosterone ratio remained well within normal clinical limits Antonio 2021.

The 2025 hair trial, which revisited the same numbers, adds two useful details. It reports the creatine condition's absolute values as rising from 0.98 ± 0.37 nmol/L to 1.53 ± 0.5 nmol/L after the loading phase and 1.38 ± 0.45 nmol/L after maintenance, and the placebo condition's as slipping from 1.26 ± 0.52 to 1.09 ± 0.40 and then 1.06 ± 0.43 nmol/L — while stating explicitly that the placebo decrease was not statistically significant. It also notes that every value, in both conditions and at every time point, sat inside the normal physiological range it cites for healthy adult males aged 18 to 59: 0.8 to 3.5 nmol/L Lak 2025.

There is one further confounder that matters in a squad of rugby players mid-season: the ISSN review points out that intense resistance exercise on its own can cause increases in these androgenic hormones Antonio 2021. And the sample is what it is — twenty young men from a single rugby squad, sixteen of whom finished. That is a reasonable size for a mechanistic crossover study, and a thin foundation for a population-level claim about hair.

Why DHT is in the story at all

The rumour is not absurd, and it is worth explaining why. The clinical trials of finasteride open by describing androgenetic alopecia — male-pattern hair loss — as caused by androgen-dependent miniaturisation of scalp hair follicles, with scalp DHT implicated as a contributing cause; finasteride inhibits type II 5-alpha-reductase, which lowers serum and scalp DHT by blocking the conversion of testosterone Kaufman 1998. The ISSN review puts the association more cautiously than the internet does: changes in these hormones, and particularly DHT, "have been linked to some (but not all) occurrences of hair loss/baldness" Antonio 2021.

The drug evidence is nonetheless substantial. In two one-year trials, 1,553 men aged 18 to 41 with male-pattern hair loss received oral finasteride 1 mg per day or placebo, and 1,215 men continued in blinded extension studies for a second year, with efficacy judged by scalp hair counts, patient and investigator assessments, and expert-panel review of photographs. In a one-inch-diameter circle of balding vertex scalp holding 876 hairs at baseline, finasteride produced increases of 107 and 138 hairs versus placebo at one and two years respectively (P < .001) — a difference that reflects both modest regrowth on the drug and continued loss without it, since treatment with placebo resulted in progressive hair loss Kaufman 1998. Worth knowing when weighing that result: the lead author's listed affiliation is the clinical research department of Merck Research Laboratories, the drug's manufacturer Kaufman 1998.

So suppressing DHT substantially, for a year or more, demonstrably changes hair counts. That is the plank the rumour stands on. But the inference runs backwards and skips several steps: a drug that suppresses an androgen for twenty-four months is not the mirror image of a percentage change in circulating DHT over twenty-one days in one small crossover trial. And the way to close a gap that wide is not to argue about mechanisms. It is to give people creatine and look at their hair.

The trial that measured the hair

That is what a randomised controlled trial published in 2025 set out to do. Forty-five resistance-trained males aged 18 to 40 were recruited and randomly assigned to creatine monohydrate at 5 g per day or a placebo of 5 g of maltodextrin per day, maintaining their habitual diets and training routines. Blood was collected at baseline and after twelve weeks for total testosterone, free testosterone and DHT, and hair follicle health was assessed directly using the Trichogram test and the FotoFinder system — hair density, follicular unit count and cumulative hair thickness. Thirty-eight participants completed the study, nineteen per group, with no significant differences in baseline characteristics; one of them was missing hair assessments, so the hair analyses rest on thirty-seven. Outcomes were analysed by repeated-measures ANOVA with a false-discovery-rate correction, in R, by a blinded investigator who was unblinded only after the analysis was finalised Lak 2025.

The between-group result was null, and it is worth stating exactly what that means, because the within-subject picture was not static. There were no group-by-time interactions for any hormone or hair-related outcome, and no group main effects; there were no significant differences between creatine and placebo in DHT, in the DHT-to-testosterone ratio, or in any hair-growth parameter. But statistically significant time main effects did appear in both groups, at p < 0.001 each, for total testosterone, free testosterone, the DHT-to-testosterone ratio and the DHT-to-free-testosterone ratio: total testosterone and DHT-to-free-testosterone rose over the twelve weeks, while free testosterone and DHT-to-testosterone fell, regardless of which supplement people took. Total testosterone changed by 124 ± 149 ng/dL in the creatine group and 216 ± 203 ng/dL in the placebo group; free testosterone changed by −9.0 ± 8.7 pg/mL and −9 ± 6.4 pg/mL respectively. Raw DHT concentrations did not change at all, and neither did creatinine or estimated glomerular filtration rate Lak 2025. In other words: things moved over three months in both arms, and the supplement was not what moved them.

On the hair itself the authors report no differences in any outcome — hair count, density, anagen rate, telogen rate, total follicular units, terminal and vellus rates, and cumulative thickness. They describe their trial as, in their words, "the first study to directly assess hair follicle quality and quantity following creatine supplementation," and their stated conclusion is that the results "refute the common claim that creatine causes baldness" and provide strong evidence against the claim that creatine contributes to hair loss Lak 2025.

The authors are candid that their findings sit against the 2009 result. They write that their results contradict previous findings of increased conversion of testosterone to DHT, and attribute the discrepancy to methodological differences — most obviously that they used no loading phase and ran for twelve weeks rather than three Lak 2025. The 5 g daily dose is described in the paper as a commonly ingested dose of creatine, not as a figure derived from any survey of what users take Lak 2025.

Two things should temper how much weight this trial carries on its own. The first is the author group's relationships. The paper's own disclosures record that one author is the chief executive and co-founder of the International Society of Sports Nutrition and a scientific advisor to creatine brands including Creapure and Bear Balanced; another is a sport nutrition advisor to the same society and has received creatine donations from Creapure for research; a third has received research funding or in-kind gifts from nutrition and sports-nutrition companies. The acknowledgments thank a company described as the official representative of BPI Sports USA, while the paper states that no specific funding or institutional support was received for the study Lak 2025. One of those authors also co-wrote the 2021 review this article leans on for the reanalysis of the 2009 percentages Antonio 2021. Readers are entitled to weigh that overlap.

The second is the authors' own list of limitations. The trial enrolled only males, chosen because DHT has been linked to male-pattern baldness, and they say research in females may be needed. It ran twelve weeks, and they accept that a longer or higher-dose study could alter androgens and hair, while noting they found no trend in that direction. Only plasma samples were taken, so hair androgen levels were not assessed directly. And family history of hair loss was not determined — a limitation they flag by citing evidence that males whose biological fathers had hair loss were twice as likely to experience it themselves Lak 2025.

What the rest of the hormone evidence shows

If creatine were meaningfully pushing androgens around, the effect ought to surface in the wider hormone literature too. The 2021 ISSN paper that gathers that literature is an expert-consensus narrative review rather than a meta-analysis: it is built around twelve questions, each answered by one researcher chosen for expertise on that topic, with the final manuscript reviewed and approved by all authors. There is no systematic search protocol and no pooled effect estimate Antonio 2021.

What it reports is this. Besides the 2009 rugby trial, twelve further studies had examined creatine and testosterone, at doses ranging from 3 to 25 g per day for periods from six days to twelve weeks. Two of them reported small, physiologically insignificant increases in total testosterone after six and seven days of supplementation; the remaining ten reported no change in testosterone concentrations. Five of those studies also measured free testosterone, and none of the five found an increase. The review's own summary is that the body of evidence it examined "does not indicate that creatine supplementation increases total testosterone, free testosterone, DHT or causes hair loss/baldness" Antonio 2021.

The same caveat applies here as to the 2025 trial. The review's author group overlaps with the International Society of Sports Nutrition position stand that describes creatine as safe and effective — Antonio, Kreider and Candow appear on both Kreider 2017 — and two of its authors also appear on the 2025 hair trial Lak 2025. That does not make the arithmetic wrong. It does mean the picture would be stronger with an independent replication.

What is still open

The 2025 trial is the most direct evidence available on this question among the studies cited here, and it is not the last word. Twelve weeks is short relative to a trait its own authors describe as progressive; thirty-eight completers is a modest sample, thirty-seven for the hair outcomes; and the participants were resistance-trained men aged 18 to 40 whose family history of hair loss was never recorded — yet men carrying that inherited risk are the group in whom any real effect would be most likely to surface. The authors themselves call for work in females, for longer durations and higher doses, for direct assessment of hair androgen levels, and for family history to be monitored in future research Lak 2025. The 2009 authors' own call for further investigation of the long-term safety of altered circulating androgen composition also remains on the table van der Merwe 2009.

So a residual question remains, and it is a narrow one: what happens over years, at ordinary doses, in men who carry the genetic risk. What the 2025 trial does answer is the claim the rumour actually makes. Hair follicles were measured under creatine and under placebo, and no difference was found between them Lak 2025.

What the genetics evidence says

The question people are really asking — will I lose my hair? — has a better-evidenced answer, and it has nothing to do with supplements. A genome-wide meta-analysis pooled 10,846 early-onset male-pattern-baldness cases and 11,672 controls of European descent across eight independent cohorts and identified 63 genome-wide-significant risk loci. Its authors estimated that those 63 loci accounted for around 39% of the phenotypic variance in that early-onset case-control sample, derived as a correlation coefficient in a multivariate linear model of the significant index SNPs — a figure specific to that sample and that phenotype definition rather than a population parameter Heilmann-Heimbach 2017.

A later analysis in a much larger cohort gives the heritability picture more directly. Studying 205,327 European males in UK Biobank, it described male-pattern baldness as strongly heritable and polygenic, reporting pedigree-based heritability of 0.62 (standard error 0.03) estimated from close relatives and SNP-based heritability of 0.39 (standard error 0.01) estimated from conventionally unrelated males. It detected 624 near-independent genome-wide loci, and those loci together contributed SNP-heritability of 0.25 (standard error 0.01), of which 26 X-chromosome loci explained 11.6% Yap 2018.

Those numbers are not interchangeable, and the coincidence between two of them is a trap worth naming. Variance explained by a specific set of loci in an early-onset case-control sample is a different quantity from heritability estimated across a population cohort, and the 0.39 SNP-heritability figure is not the same thing as the 0.25 contributed by the 624 genome-wide-significant loci. Both analyses were also conducted in European-ancestry samples, so neither speaks directly to other populations. What the genetics evidence cited here does support is the shape of the answer: much of the explanation for male-pattern baldness sits in inherited variation. If a man's father, uncles and grandfathers receded early, that inheritance is doing far more work than anything in a shaker bottle.

What creatine's safety file contains

The reason this matters is that the trade-off is lopsided. The International Society of Sports Nutrition's position stand — a review of the literature rather than a trial — states that short- and long-term creatine supplementation, at up to 30 g per day for five years, is safe and well tolerated in healthy individuals and in a number of patient populations from infants to the elderly, and that creatine monohydrate is the most extensively studied and clinically effective form for use in nutritional supplements in terms of muscle uptake and high-intensity exercise capacity Kreider 2017. Those statements are dose- and duration-bounded, and the passages quoted here do not address hair, DHT or androgens at all.

A separate systematic review of creatine in health and disease concludes that supplementation has several health and therapeutic benefits across the lifespan and reports that creatine has been found safe in a number of populations, while explicitly stating that additional safety and tolerability studies are needed in pregnant women and in those trying to conceive Kreider 2021. Its disclosure is relevant: it appeared in a special issue sponsored by AlzChem, a creatine manufacturer, and its lead author chairs that company's scientific advisory board Kreider 2021.

The most concrete long-term human data come from an open-label, non-randomised study of 98 Division IA college football players followed over 21 months, with fasting blood and 24-hour urine samples at nine points — 0, 1, 1.5, 4, 6, 10, 12, 17 and 21 months. Participants were grouped afterwards by how long they had supplemented: 44 who did not take creatine, 12 who took it for up to six months, 25 for seven to twelve months, and 17 for twelve to twenty-one months. A 69-item panel — 54 quantitative blood and urine markers plus 15 qualitative urine markers, covering metabolic markers, muscle and liver enzymes, electrolytes, lipid profiles, haematological markers, lymphocytes and renal function — showed no significant differences among groups on multivariate testing of the quantitative panel (p = 0.51), no clinically significant interactions among groups on univariate analysis, and no apparent differences on the qualitative urine markers. The authors' conclusion is hedged and comparative: long-term supplementation, up to 21 months, "does not appear to adversely effect markers of health status in athletes undergoing intense training in comparison to athletes who do not take creatine" Kreider 2003. Hair was not among those 69 markers — by construction, the panel covers serum, whole blood and urine — which is precisely why a trial that measured follicles was worth running Lak 2025.

For what the supplement is actually for, and what the performance and cognitive evidence looks like, see our full review of creatine's evidence base. For what the first few weeks typically involve, including the water-weight question that trips most people up, see what happens in week one.

If you're genuinely worried about your hairline

Two things follow from the evidence above. The first is that dropping creatine to protect a hairline trades a documented benefit against a risk that a randomised trial looked for, measured directly in follicles, and did not find any group difference on Lak 2025. The second is that if hair loss is already a live concern — early recession, a strong family history, noticeable shedding — the useful move is a dermatologist, not a supplement audit. Among the interventions discussed here, finasteride works by blocking the enzyme that converts testosterone to DHT, and in the two-year trial programme described above it increased hair counts relative to placebo while the placebo groups continued to lose hair Kaufman 1998. Whether it suits any individual is a medical decision, not a supplement-forum one.

The honest bottom line is less dramatic than the screenshot. One small crossover trial measured a hormone over three weeks, never looked at hair, and its authors asked for more work on long-term androgen safety rather than declaring a hair risk. A larger, longer randomised trial looked at the hair and found no difference between creatine and placebo. For anyone with baldness in the family, that is a reason to see a doctor about the hair — not a reason to change a supplement.

References