RU58841, pyrilutamide, and clascoterone: which one is actually better?

If you have spent any time reading about hair loss you will have met these three names. RU58841, the one people buy from research-chemical websites. Pyrilutamide, usually written KX-826, the one from a Chinese company that just reported a successful late-stage trial. And clascoterone, also called CB-03-01 or Breezula, the one already approved as an acne cream and now finishing a large trial for hair.

They are trying to do the same thing. None of them reduces how much dihydrotestosterone, DHT, your body makes. All three go after the androgen receptor instead, the protein DHT has to sit inside before it can tell a hair follicle to shrink. Occupy that pocket with something else and the hormone arrives to find the seat taken.

So which one works best? The honest answer is that the question cannot be settled with the evidence that exists, and the reasons why turn out to be more useful than any ranking would have been. There has never been a study of any kind comparing any two of these three molecules. A search of the medical literature for any paper mentioning clascoterone alongside either RU58841 or pyrilutamide returns nothing at all.¹

What can be done is to look at what each one has actually demonstrated, in public, in a form somebody outside the company can check. That produces a much clearer picture than the binding-affinity arguments people usually have, and it separates the three quite decisively.

Almost every online argument about these drugs is really an argument about binding affinity, and almost all of it rests on a misunderstanding of what binding affinity is.

Here is the picture people carry around: one drug molecule and one DHT molecule approach a receptor, the one that binds more tightly wins, and that is that. It is a duel. It is also wrong.

What actually happens is constant, restless traffic. Molecules drift into the receptor pocket and fall out of it again, millions of times, and at any given instant some fraction of your receptors are occupied and the rest are empty. Binding affinity does not decide a contest. It sets the odds on each encounter. The formal version of this is the law of mass action, and the number people quote, the dissociation constant or Kd, has a precise meaning under it: Kd is the concentration at which exactly half the receptors are occupied at any instant. Getting occupancy up near total requires a concentration many times higher than the Kd.²

Two consequences follow, and they matter more than any affinity number.

The first is that concentration is at least as important as affinity. A drug that binds somewhat less tightly than DHT can still occupy most of the receptors if there is far more of it around. A drug that binds more tightly will lose anyway if it never reaches the follicle in quantity. This is why all three of these are topicals applied directly to the scalp, and why the serious ones are dosed twice a day.

The second is that how long a molecule stays put may matter more than how tightly it grabs. Affinity is really a ratio of two speeds, how fast a molecule binds and how fast it lets go. A well-known argument in pharmacology holds that the length of time the drug stays attached, its residence time, often predicts the real-world effect better than equilibrium affinity does. There is a related effect in tissue, where a molecule that falls off a receptor frequently rebinds to that receptor or a neighbouring one before it can drift away, so occupancy in a real scalp outlasts what the raw numbers suggest.³⁴

This is also what these drugs are up against. DHT binds the androgen receptor around twice as tightly as testosterone does, and it lets go three to five times more slowly, which is what makes it the more powerful signal even though there is much less of it around. And balding scalp is a harder target than it looks: dermal papilla cells taken from balding scalp carry measurably more androgen receptor than cells from unaffected scalp, so there is more to occupy exactly where you need to occupy it.⁵⁶⁷

You will see a comparison like this in every thread on the subject. RU58841 binds at about 1.8 nanomolar. KX-826 comes in at 0.28 nanomolar, so it must be the stronger one. Clascoterone is somewhere behind both.

Every part of that is unsafe, and it is worth walking through why, because the same trick appears constantly in supplement and research-chemical marketing.

Start with RU58841. The figure quoted is a set of dissociation constants, 1.8 nanomolar for the human receptor and 1.1 for the rat, and it is always attributed to a 1994 paper by Battmann and colleagues. That paper exists. Its abstract does not contain the numbers. It contains no quantitative affinity data at all, only the sentence that the compound displays high affinity for the hamster prostate and flank organ androgen receptors. The full five-value panel that circulates online, covering human, rat, mouse and two hamster tissues, is retrievable from the datasheets of chemical suppliers, which cite the paper but publish figures the paper’s abstract does not carry. Different vendors also disagree with one another, and at least one reports an association constant where another reports a dissociation constant, which are different quantities.⁸⁹¹⁰

Now pyrilutamide. Its 0.28 nanomolar figure is not a dissociation constant at all, it is a half-maximal inhibitory concentration, an IC50, and it traces to a patent filed by the company that makes the drug. No assay format, no tracer, no concentrations, nothing that would let anyone reproduce or interpret it has ever been published.¹¹

And clascoterone has no publicly retrievable affinity number whatsoever. Not in its FDA label, not in its mechanism papers, not in the review literature. Its original characterization quantified potency only as a ranking in a hamster test: roughly three times more potent than flutamide, about twice as effective as finasteride, about as active as cyproterone acetate.¹²¹³⁴⁷

So the famous three-way comparison is a vendor catalog, a company patent, and a blank space.

Even if all three numbers were solid, they still could not be lined up, because a Kd and an IC50 are not the same measurement. A dissociation constant describes how tightly a molecule sits in a receptor at equilibrium. An IC50 describes the concentration needed to knock down half of some measured response, and it depends heavily on how the experiment was set up. The 1973 paper everyone cites for converting between them, by Cheng and Prusoff, actually concludes that the two are not equal when the inhibition is competitive, which is exactly the case here. Converting requires knowing the concentration of the tracer molecule used in the assay and that tracer’s own binding constant, neither of which is published for any of these three.¹⁴¹⁵¹⁶

How much can the setup alone change the answer? There is a clean illustration in the androgen receptor literature. When researchers measured how fast testosterone and DHT let go of an isolated fragment of the receptor, the two looked nearly identical, 25 minutes against 32. When they measured the same two hormones against the full-length receptor, the gap opened to 55 minutes against 188. Same molecules, same question, and the answer moved by a factor of four depending on what they were tested against.⁵

RU58841 is genuinely old. The RU prefix comes from Roussel Uclaf, the French pharmaceutical company where it originated. As the program changed hands the molecule was renamed twice, first to HMR-3841 and then to PSK-3841. All three codes refer to the same compound, which matters because the human trials were registered under a name almost nobody searches for.¹⁷¹⁸

The animal evidence is real and reasonably good. In stumptailed macaques, a species that goes bald in a pattern resembling ours, topical RU58841 produced what the researchers described as a potent increase in the density, thickness and length of hair. In a separate study, human scalp grafts taken from balding men were kept alive on mice and treated for six months. Among the grafts treated with RU58841, 8 of 29 active follicles started a new growth cycle, against 2 of 28 in the control grafts, and hair grew measurably faster.¹⁹²⁰

It then went into humans, and this is the part that gets misdescribed. Two trials were registered, both retrospectively and both on the same day in October 2005, years after they had actually finished.²¹²²

The first ran in France for four weeks in 2002, targeting 60 men. Its purpose was not to grow hair. Its primary outcome was safety, measured through hormone profiles on days 1, 15 and 28, and one of its secondary questions was whether the men’s female partners were being exposed to the drug in ordinary daily life, which is why the study enrolled the partners too. Hair count was not an outcome of that trial at all.²¹

The second is the one that matters. It ran in the United Kingdom and Belgium from October 2002 to August 2003, targeting 120 men, comparing 2.5% and 5% solutions applied once daily against vehicle over six months, with total and anagen hair counts as a primary outcome. That is a properly designed efficacy trial.²²

It has never reported a result. In both registry records the fields for results, publications and adverse events are empty, and have stayed empty through updates as late as 2014. Neither trial ever appeared on ClinicalTrials.gov. No human trial of RU58841 has ever been published in a medical journal under any of its three names. A search of the entire indexed medical literature returns 17 records for the molecule, and not one of them is a clinical trial.²²²³²⁴

What people quote instead is a marketing page. The archived ProStrakan website said the molecule had completed a proof-of-concept phase 2 study in alopecia and had demonstrated similar efficacy after six months of treatment as that observed with current oral therapy for alopecia after twelve months, based on the increase in net hair count, with no systemic antiandrogenic effect observed, in 90 patients.²⁵

Read that carefully, because it is thinner than it looks. The page never names finasteride, it says current oral therapy. It gives no hair count, no comparison group result, no statistics, and no adverse event data. Its figure of 90 patients matches neither registered trial, whose targets were 60 and 120. And the sentence directly after the efficacy claim is the giveaway: this product is available for licensing. It is a sales page, written to attract a buyer.²⁵

No buyer came. The drug databases record the ending without explaining it: two entries in May 2007 reading, simply, no development reported. ProStrakan was later bought by Kyowa Hakko Kirin and no successor has ever revived the program. No company statement, filing or report explaining the decision could be located.¹⁷

Pyrilutamide, KX-826, comes from Kintor Pharmaceutical of Suzhou, China, the same company behind the androgen receptor degrader GT20029. It has by far the most complete public trial record of the three, and it is the only one of them whose late-stage numbers have been disclosed in full.²⁶

One naming point first, because it causes real confusion. Koshine, styled KOSHINÉ, is not the drug’s brand name. It is Kintor’s cosmetics brand, and KOSHINE 826 is the trade name used when the compound is sold to other companies as a cosmetic ingredient. The investigational drug has no brand name, for the straightforward reason that it is not approved anywhere.²⁷

The program began with two American phase 1 studies in Miami, one single-dose and one running fourteen days, both about safety and how much drug enters the blood. A Chinese phase 2 in 120 men followed, and the company reported that the 0.5% twice-daily arm gained 15.34 more hairs per square centimeter than placebo at 24 weeks, at p equal to 0.024. That dose was carried into phase 3. A separate phase 2 in 160 Chinese women reported 11.39 hairs over placebo at p equal to 0.0087.^2829303132

An American phase 2 in 123 men also ran, finishing in February 2023. Its registered primary endpoint was explicitly a comparison against placebo. Kintor headlined the result Successful Completion, and the only figure it gave was that the 0.5% twice-daily group increased by about 10 hairs per square centimeter compared with baseline, statistically significant at p equal to 0.0088. On the placebo comparison, the one the trial was designed around, it said only that the drug indicated an improvement. No placebo-adjusted difference and no placebo-adjusted p-value has been released in the three and a half years since, and no results have been posted to the registry.³³³⁴

Then the first phase 3 failed. It was a large trial, 740 Chinese men, 0.5% twice daily against vehicle for 24 weeks. In November 2023 Kintor reported that hair counts rose against baseline with high significance, at p below 0.0001, and that the drug beat placebo at every visit but that none of those differences reached statistical significance. The trial missed the endpoint it was designed to hit.³⁵³⁶

What happened next is the most informative thing in the whole program. Kintor went back and doubled the concentration. The rerun was an adaptive trial testing both 0.5% and a new 1.0% strength, and the company’s stated reason for the higher strength was that it significantly increased how much drug stayed on scalp cells. A 90-man first stage reported in July 2025. The main stage, 666 men, reported on 18 March 2026.³⁷³⁸³⁹

Those results are the clearest numbers any of these three drugs has produced. Hair count rose 15.33 per square centimeter on 1.0% twice daily, 14.46 on 0.5%, and 4.68 on vehicle. That gives vehicle-adjusted gains of 10.65 and 9.78 hairs per square centimeter, both at p below 0.0001. Kintor reported excellent safety with no drug-related serious adverse events, though it disclosed no adverse event rates, no secondary endpoint results and no confidence intervals.⁴⁰

Two things about that deserve flagging. The 0.5% arm, the exact strength that had failed in 740 men, worked this time, which points away from dose as the reason for the first failure and toward how that first trial was run. The next section takes that up, because there is a specific claim about it. And the vehicle arm behaved very differently between the two stages of the same adaptive trial, gaining 8.73 hairs in the 90-man stage against 4.68 in the 666-man stage. Control arms are not stable things.³⁸⁴⁰

A separate 52-week open-label study reported good long-term tolerability with no drug-related sexual dysfunction, which is the reassurance people most want from an antiandrogen. It had no control arm, so its efficacy figures, that 46% of patients gained at least 10 hairs per square centimeter, should be read as description rather than proof.⁴¹

Kintor’s medical lead put numbers to that in a 2024 interview, and they are worth having because the company has never published an adverse event table. Across the 740-man phase 3 and the roughly 200-man long-term study, he counted two reported sexual adverse events in total. One was erectile dysfunction, in a man on placebo. The other was mild, self-reported, needed no treatment, and occurred in a man who was having a shingles outbreak at the same time.⁴⁴

A specific explanation circulates for why pyrilutamide’s first phase 3 failed: that the trial ran through China’s COVID lockdowns and patients could not take the drug properly. It is easy to dismiss as fan defense of a favored molecule. It is not. Kintor said it, in some detail, and the reason people struggle to find it is that the record is a video rather than a document.

Nothing in Kintor’s English or Chinese announcement archive, in its exchange filings or in the Chinese financial press attributes the failure to the pandemic. The company’s own announcement of 27 November 2023 gives the numbers and offers no explanation for them at all.⁴²³⁶

The explanation was given verbally, in English, on the YouTube channel KwRx, run by the Reddit user noeyys, in an interview with two Kintor scientists published in December 2024. Wilson Lou, a physician trained at Beijing Medical School with a PhD from the University of Pennsylvania who runs the medical side of the program, was asked directly what had changed between the trial that failed and the ones that worked, and whether compliance was better the second time.⁴⁴

His answer was blunt, and it started with the timeline. A six-month study took two years to run from beginning to end, and recruitment was not the reason, because in his words a patient is not difficult to recruit at all. Then the cause: as you probably know, the COVID situation in China had a huge impact, and we were forced to stop the study.⁴⁴

On compliance he was more specific than the secondhand version of this story usually is. Compliance is out of the window, he said, you really cannot check. Even for the people who needed the drug, the company could not ship it, because everybody was quarantined at home. He added that when people are worried about food, putting a solution on your hair is not a priority, and that the stress and the fever of the illness itself can change hair on their own.⁴⁴

Then he gave the comparison that makes the claim checkable. The later trial he contrasted it with began at the end of December 2023, after China had dropped its quarantine policy, so nearly all of its recruiting happened in normal conditions. The product, he said, was the same. What differed was compliance and patient follow-up. He also noted that his was not the only Chinese company whose long trials came out of that period looking strange.⁴⁴

Two qualifications are worth keeping. This came thirteen months after the failure, from the company whose drug missed, spoken rather than filed. And the most contemporaneous account emphasizes something else: two weeks after the failure, Kintor founder and chief executive Tong Youzhi told a Chinese interviewer that a larger trial across more centers introduces variables that erode the signal seen in phase 2, that the result surprised everyone at the company, and that they would consider raising the dose.⁴³

Those two accounts are not in conflict, and the trial data ended up favoring the compliance one more than you would expect. Kintor did raise the dose. But the 0.5% arm, the exact strength that had failed in 740 men, hit its endpoint in the rerun at 9.78 hairs per square centimeter over vehicle, at p below 0.0001. If the first failure had been a dose problem, that arm had no business succeeding. It succeeded.⁴⁰

Clascoterone is the odd one out, and the first thing to understand is that it is a steroid. It is cortexolone 17-alpha-propionate, a modified version of a natural precursor in your own cortisol pathway. Look at the three structures side by side and it is obvious: RU58841 and pyrilutamide are built on the same kind of compact non-steroidal frame, while clascoterone has the four fused rings of a steroid hormone. It is routinely misfiled as non-steroidal online, including by people who should know better.⁴⁵

That matters because the four-ring shape is what your other steroid receptors recognize too, so a steroidal antiandrogen carries a cross-reactivity risk a non-steroidal one does not. Clascoterone’s answer is not a different shape but a different strategy: it is designed to be broken down quickly once it leaves the skin, so the effect stays local.

It is the only one of the three that is an approved medicine, though not for hair. The FDA approved Winlevi, a 1% cream, for acne in August 2020. The hair-loss product is a different thing: five times the concentration, in an alcohol solution rather than a cream, developed under the name Breezula.⁴⁶⁴⁷

The local-only strategy works less completely than the marketing suggests, and the evidence for that is not from critics but from the company’s own peer-reviewed paper. A published phase 1 study states plainly that when used in a cream formulation little of the drug is absorbed, but that the solution formulation developed for androgenetic alopecia leads to a measurable systemic concentration and accumulation of the antiandrogen. The acne cream’s own FDA label carries a warning for suppression of the adrenal stress-hormone axis, seen in 5% of adults and 9% of adolescents at day 14, and records a potassium signal as well.⁴⁸⁴⁷

European regulators took this seriously enough to refuse the acne cream outright in April 2025, on the grounds that there is a risk of the medicine being absorbed into the blood and suppressing the hormonal axis, which could impair growth and sexual maturation in adolescents. Approval came only on re-examination in August 2025, with a restricted indication, dose limiting, monitoring, and adolescent use confined to facial acne. That is the regulatory history of the 1% cream. The hair product is five times stronger.⁴⁹

The mid-stage hair data are the best-documented part of the story and they look good. A 404-man dose-ranging trial in Germany reported absolute six-month gains of 13.01, 12.21 and 20.79 hairs per square centimeter for the 2.5%, 5% and 7.5% twice-daily arms, against minus 0.11 for vehicle. At twelve months the placebo-adjusted differences were 10.2, 13.8, 14.3 and 12.7 hairs. Notably, phase 3 went ahead at 5%, not at the numerically best 7.5%.⁵⁰⁵¹

Then the phase 3 program, and here the reporting changes character. SCALP1 and SCALP2 enrolled 1,465 men between them, the largest program any of these three drugs has run. In December 2025 Cosmo announced that both trials met their hair-count endpoint, and reported the result as a 5.39-fold relative improvement over vehicle in one study and 1.68-fold in the other.⁵²⁵³⁵⁴

No absolute hair counts were given. Not for the treated arms, not for the vehicle arms, not in that release, not in the April 2026 twelve-month release, not in the July 2026 half-year report, and not on either trial registry. Every publicly available efficacy figure for this program is a ratio.⁵⁴⁵⁵⁵⁶

That choice makes the result impossible to place. A ratio is a fraction, and its size is driven by its denominator, which here is however much the placebo group happened to change. The company’s own phase 2 shows exactly how unstable that denominator is: its vehicle arm moved by minus 0.11 hairs per square centimeter. Divide by a number that close to zero and the ratio can be made enormous while the actual gain stays ordinary. It also explains the otherwise baffling gap between the two phase 3 results, 5.39-fold against 1.68-fold in trials the company itself describes as identical in design. The likeliest explanation is not that the drug worked three times better in one of them but that the two vehicle arms behaved differently.⁵⁰

The same release also asserts that clascoterone blocks DHT at the follicle without systemic absorption, which contradicts the company’s own published phase 1 paper quoted above. And its arithmetic is internally inconsistent: it equates 5.39-fold with 539%, but a 5.39-fold ratio is a 439% increase. The company uses both phrasings in one document.⁵⁴⁴⁸

For contrast, when the same company reported the acne trials that won FDA approval, it published absolute numbers: success rates of 18.8% and 20.8% against 8.9% and 6.5% on vehicle, and absolute lesion count reductions. It chose ratios only for hair.⁵⁷

The three moleculesOpen full size

You do not need a statistics background to catch the two moves that do most of the work in hair-loss marketing. Both appear in the material above.

The first is comparing a group against its own starting point instead of against the control group. It sounds reasonable and it is close to meaningless, because the control group improves too. Bland and Altman showed in 2011 that testing each randomized group separately against its own baseline is invalid and biased, and quantified how badly: the real false-positive rate of that procedure can reach 50% for two groups and 75% for three, against the 5% everyone assumes. Their worked example is a skin-cream trial where the active group had a significant within-group change, the vehicle group did not, and the actual difference between the two groups was not significant at all.⁵⁸

Kintor’s November 2023 announcement of its failed phase 3 contains both framings in adjacent sentences, which makes it an unusually clean specimen. The drug promoted hair growth compared to baseline with statistical significance at p below 0.0001. Compared with placebo, there was improvement at all visit points, with no statistical significance. The first sentence is the impressive one and the second is the trial result.³⁶

The second move is reporting a relative change instead of an absolute one. A 5.39-fold improvement sounds like a different category of result from 10 extra hairs per square centimeter, even when the two describe the same finding. The relative version also hides its own denominator, so you cannot tell whether the drug did well or the placebo group simply did badly.

There is a third thing worth knowing, which is not a trick but a genuine feature of this field: control arms are wildly inconsistent. Documented changes in vehicle groups over comparable periods include minus 10 hairs per square centimeter, plus 3.9, plus 4.68, plus 6.7, plus 8.73, and minus 0.11. Part of that is regression to the mean, the statistical tendency of people selected for having a problem to look better on remeasurement. Part is physical: at least one trial suggested that massaging a vehicle into the scalp twice a day increases local circulation, and that starting a study in autumn and finishing in spring flatters both arms.⁶⁴⁵⁹⁶⁰

That variability is why a drug’s effect must always be read as the gap between the arms, and why a ratio built on top of an unstable denominator tells you very little.

The only pivotal number available in absolute terms belongs to pyrilutamide, so it is the only one that can be placed against the drugs people already take. It compares respectably.

Finasteride 1 mg, the most studied drug in this field, produced 107 more hairs than placebo at twelve months in its pivotal trials, in a one-inch circle covering 5.1 square centimeters, which works out to roughly 21 hairs per square centimeter. A separate 48-week study using an actual one-square-centimeter window found a 17-hair difference. Topical 5% minoxidil produced about 14.7 hairs over vehicle at 48 weeks. The closest methodological match to pyrilutamide’s trial is the phase 3 of topical finasteride spray, which used the same 24-week window and the same tattooed one-centimeter target area, and produced vehicle-adjusted differences of about 13.5 for the topical and 14.4 for oral finasteride.^61626364

So pyrilutamide’s 10.65 hairs at 24 weeks sits slightly below topical finasteride measured the same way over the same period. That is a real result, in a believable range, and not a revolution.

Now the caveats, and there are enough of them that this should not be read as a league table.

• The counted area is not standardThe finasteride pivotal trials counted hairs in a 5.1 square centimeter circle. Modern trials use one square centimeter. Even within one company the definition moves: Kintor’s GT20029 trial counted in 0.903 square centimeters while its KX-826 trial used a full one.
• The threshold for a countable hair is not standard eitherSome trials count hairs at least 40 micrometers thick, others at least 30. Lowering the bar moves hairs from uncounted to counted without anything growing.
• The counting software has known errorA published critique found automated digital analysis underestimated mid-scalp density by roughly 27%, and argued the claim that it is suitable for treatment-response trials is overstated.
• The populations differBaseline scalp density runs around 226 hairs per square centimeter at the vertex in people of European descent and about 171 in healthy young Chinese men. The pyrilutamide trials were Chinese; the finasteride and clascoterone trials were not.
• The durations differTwenty-four weeks against forty-eight weeks against twelve months. Hair counts keep moving, and the drug’s own gain and its gap from placebo move in different directions over years.
• The comparison has never been runEvery sentence above compares separate trials done by different people at different times. The one study that ever put a topical antiandrogen head to head against minoxidil, a 95-man clascoterone trial that finished in May 2016, has never reported a result.

By strength of public evidence, the order is clear, and it is not the order the binding numbers would give you.

1. 01
   Pyrilutamide is first, narrowly, and on transparency rather than potency

   It is the only one of the three that has stated a late-stage result the way results should be stated: 666 men, 15.33 hairs against 4.68 on vehicle, a gap of 10.65, p below 0.0001. It also has the most complete registered trial history, a 52-week safety study, and a failure it disclosed rather than buried. Its weakness is that all of this is still company reporting, none of it is published, and its US registry entries sit unposted years after completion.

2. 02
   Clascoterone is a very close second, and could easily be first

   It has the largest program, 1,465 men, the only regulatory approval of the three for anything, and by far the deepest peer-reviewed pharmacology. Its phase 2 numbers, which were reported in absolute terms, are strong. It sits second here only because its phase 3 was reported as a ratio, which makes the size of the win unknowable from outside. Publish the absolute hair counts and this ranking may well invert.

3. 03
   RU58841 is a distant third, on evidence rather than on chemistry

   The molecule may well work. The animal data are decent and it reached a properly designed six-month efficacy trial. But that trial reported nothing, no human trial of it has ever been published, development stopped without explanation nearly twenty years ago, and what is sold today has no manufacturing standard behind it. Third place here reflects an absence of evidence, not evidence of absence.

This is a snapshot of a moving situation, and the specific things that would move it are worth naming so you can watch for them rather than for noise.

• Cosmo publishes absolute hair countsIf clascoterone’s phase 3 gap turns out to be comfortably above 10.65 hairs per square centimeter, it takes first place outright. If it is below, the ratio reporting starts to look like a choice rather than a convention.
• Either phase 3 appears in a peer-reviewed journalConfidence intervals, adverse event tables, dropout handling and prespecified analyses are all invisible right now for both drugs.
• A regulator actually approves oneKintor targets a Chinese filing in the second half of 2026 with approval hoped for 2027. Cosmo targets a US filing in the first quarter of 2027 and Europe in the second. Neither has filed anything yet.
• The unpublished trials reportThe 2016 clascoterone study against minoxidil, the 2003 RU58841 efficacy trial, and the American pyrilutamide phase 2 all have results somewhere. Any of them would change the picture.
• Someone runs a head-to-headNo study has ever compared any two of these three, in humans or in a test tube. Until one does, every comparison here included is arithmetic across separate experiments.
• Long-term safety data accumulatesA 2026 analysis of the FDA’s adverse event database found 1,085 reports involving clascoterone and flagged adrenal axis suppression among the signals, alongside a reproductive and breast disorder signal. That is the acne cream at 1%. The hair product is five times stronger and taken indefinitely.

If you want the one-line answer: pyrilutamide and clascoterone are both real, both have late-stage evidence that they beat placebo, and both are close enough that the ordering depends on a number one of them has not released. RU58841 is not in the same conversation, not because the molecule is bad but because twenty-three years after its last trial finished, nobody has ever said what happened in it.

The more durable takeaway is about how to read the next one of these, and there will be a next one.

Ignore binding constants unless the assay is described. Ask for the gap between the treated group and the placebo group, in hairs per square centimeter, over a stated number of weeks. Treat a change from baseline as no evidence at all. Treat a ratio as a number with something missing underneath it. Notice whether a company that publishes absolute results for one product switches to relative ones for another. And when a company explains away a failure, do not accept or reject the explanation on the spot, wait for the next dataset to agree or disagree with it, because that part is not in the company’s gift.

Apply those checks and most of the noise around hair-loss drugs resolves. Two of these three pass them. One never got the chance.