Denifanstat: The Acne Pill That Turns Off the Oil

Forty Years, Four Levers, One New Idea

If you have ever been prescribed something for acne, you have been handed a drug that was already old. Isotretinoin was approved in 1982. Doxycycline and minocycline predate it. Spironolactone was designed for blood pressure, works for acne only in women, and is still off-label for it. The combined oral contraceptive is, again, for women only. The most recent oral acne drug to reach the US market, sarecycline in 2018, was a narrower-spectrum tetracycline. A better antibiotic, in other words, not a new idea.^5,15

Acne has four accepted drivers: too much sebum, sticky follicular keratin, the bacterium Cutibacterium acnes, and inflammation. The one lever no gender-neutral oral drug has pulled cleanly since 1982 is the first. Isotretinoin pulls it hard, and pays for that with teratogenicity, iPLEDGE paperwork, lipid monitoring, and a tolerability profile that ends a meaningful share of courses early.⁵

That is the gap denifanstat is aimed at. It is an oral, once-daily inhibitor of fatty acid synthase, the enzyme sebaceous glands use to make oil from scratch. Sagimet Biosciences, the US company that owns it, has called it "the first innovative oral treatment for acne in more than forty years" in at least four separate press releases.^9,10 Dermatologists quoted in the trade press have been warmer still. Neal Bhatia, a past vice president of the American Academy of Dermatology, described the mechanism as "removing the fuel from the car" and suggested the drug could replace a large share of the roughly five million oral antibiotic prescriptions written for acne each year.¹³

I understand the enthusiasm. The mechanism is the most interesting thing to happen in acne pharmacology in my professional lifetime. But I have read every trial this molecule has been through, in three different diseases, and the story is more complicated than the press releases. The pivotal acne trial is fifteen months old and still unpublished. The safety profile in acne patients is strikingly cleaner than the same dose produced in liver patients, and nobody has explained why. And the phrase "isotretinoin without the baggage," which you will hear a great deal in the next two years, currently rests on a sebum number generated in cancer patients at several times the acne dose.^4,7

An Enzyme, a Fatty Acid Found Nowhere Else, and a Cancer Drug That Failed

Your sebaceous glands are among the most lipogenically active tissues in your body, and they are unusual in another way: they build most of their lipid from scratch. The process is called de novo lipogenesis, and its final committed step is run by fatty acid synthase, or FASN, which stitches acetyl-CoA and malonyl-CoA into palmitate. Palmitate is then elongated and desaturated into the fatty acids that make up sebum. Company materials, citing prior mechanistic work, put the share of sebum lipid that comes from de novo synthesis at roughly 80 percent.⁷

One of those downstream products deserves a name. Sapienic acid is a 16-carbon fatty acid found essentially nowhere in the human body except sebum. It cannot be obtained from diet; it exists only because the gland made it. That makes it a nearly perfect biomarker of target engagement. Block FASN, and sapienic acid in sebum should fall. It does. In an early oncology study, the drug reduced de novo sebum lipids by more than 90 percent by day 15, dose-dependently.^4,7

Here is the first caveat, and it is the sponsor's own. That 90 percent figure came from cancer patients dosed at 100 milligrams per square metre of body surface, several times the 50 milligram flat dose used in acne. Sagimet's investor slides carry a footnote saying exactly that.⁷ There is no published sebumeter or Sebutape measurement at the 50 milligram acne dose in acne patients. When you see "over 90 percent sebum reduction" in a headline, it is describing a different dose in a different population. The number that matters, sebum suppression at 50 milligrams in someone with acne, has not been released.

The molecule's origins explain why an oncology cohort exists at all. Denifanstat began life as TVB-2640 at a company called 3-V Biosciences, later renamed Sagimet, and it was designed to starve tumours of lipid. The first-in-human trial enrolled 136 patients with advanced solid tumours across eleven sites in the US and UK between 2013 and 2017.⁴ Efficacy was thin: no complete or partial responses on monotherapy. But the side effects were instructive, because they were the side effects of a drug doing exactly what it was designed to do in every lipid-hungry tissue at once. Alopecia in 61 percent of patients. Hand-foot syndrome in 46 percent. Dry skin in 22 percent. The dose-limiting toxicities were reversible effects on skin and eyes.⁴ Read that list again with acne in mind. Dry skin and dry eye are exactly what you would expect from a sebum-suppressing drug. Hair loss is the one nobody in dermatology wants to talk about, and I will come back to it.

The sponsor also claims a second mechanism: FASN inhibition reduces cytokine secretion and Th17 differentiation in preclinical models, which would make the drug anti-inflammatory as well as anti-sebum.⁷ That is plausible and it is mechanistically tidy. It has not been clinically dissected in any acne trial, so I treat it as a hypothesis with a marketing budget.

The Liver Trial That Tells You What 52 Weeks Look Like

After oncology stalled, Sagimet pivoted the molecule to fatty liver disease, on the theory that if you can switch off lipogenesis in a sebaceous gland you can switch it off in a hepatocyte. That produced the only long, blinded, placebo-controlled dataset this drug has at the acne dose, and it is worth understanding in detail because it is the closest thing we have to a one-year safety trial.

Two things from that trial matter enormously for acne. First, the drug works in a second organ system through the same mechanism, which is real evidence that FASN inhibition at 50 milligrams does something biologically meaningful in humans. Second, one in five patients lost hair on it, against one in twenty-five on placebo, in a blinded trial where nobody knew who was on drug.³ Hold that number.

And then, in April 2026, Sagimet announced it was freezing further MASH development unless it could find outside money, and redirecting its capital to dermatology.¹² A drug with Breakthrough designation in a disease with enormous unmet need was set aside so the company could bet the enterprise on acne. That is not a criticism, exactly. Acne is a bigger market with shorter trials and no liver biopsies. But it means every dollar of the $257.6 million Sagimet held at the end of June 2026 is riding on the acne data, and you should read the acne press releases with that in mind.¹¹

The Phase 2 Hit Lesion Counts and Missed the Endpoint That Matters

The Chinese rights to denifanstat belong to Ascletis Pharma, a Hangzhou company listed in Hong Kong, under a license from Sagimet. Every acne trial to date has been run by Ascletis, in China, in adults. The first was a Phase 2 dose-ranging study whose topline results were announced in May 2023 and whose peer-reviewed publication arrived three years later, in the July 2026 issue of the Journal of the European Academy of Dermatology and Venereology.¹

I want to be fair about this. A 45-patient arm is not powered to detect a difference in a yes-or-no responder endpoint, and the trial was not designed to. Lesion counts are a continuous measure, and continuous measures find signals that binary ones miss in small samples. The 50 milligram arm's lesion reduction was real and clinically meaningful.¹ An accompanying commentary in the same journal, reasonably, called the drug "a promising step forward."²

But the thing that later "succeeded" in Phase 3, the IGA responder rate, did not work in Phase 2. That is not disqualifying. It does mean the pivotal IGA result has never been independently replicated, and that the only replication attempt on record failed. Also note the entry for "urine protein positive" in the adverse event list. It appears in the peer-reviewed Phase 2 paper and in none of the promotional material since. I do not know what it means. Neither, apparently, does anyone else, because nobody has followed it up in print.

480 Patients, Three Endpoints Met, Zero Pages in a Journal

The pivotal trial, study ASC40-303, enrolled 480 Chinese adults aged 18 to 40 with moderate-to-severe acne, randomized one-to-one to denifanstat 50 milligrams or placebo once daily for 12 weeks.^6,16 Baseline characteristics were well balanced: about 102 total lesions per patient in each arm, roughly 86 percent graded moderate and 14 percent severe.⁷ Topline results were announced by both companies in early June 2025, presented as a late-breaker at the European Academy of Dermatology and Venereology congress in Paris that September, and shown again at Fall Clinical in Las Vegas.^6,13

These are good numbers. Let me say that plainly before I take them apart. A placebo-adjusted IGA success difference of 18.6 points is larger than the roughly 9 to 14 points clascoterone cream achieved in its pivotal trials and the 7 to 11 points sarecycline achieved in its own.¹⁵ The 63.5 percent reduction in inflammatory lesions exceeds sarecycline's roughly 50 percent. The trial was properly randomized, blinded, and adequately sized. If this were published, with a CONSORT diagram and a supplementary appendix I could read, I would be closer to a stronger rating than I am.

It is not published. Fifteen months after topline, the pivotal trial of a drug whose Chinese marketing application was accepted in December 2025 exists in public only as slides.⁸ The Phase 2 took three years to reach print, so perhaps this is just Ascletis's pace. But the consequences of press-release-only data are not hypothetical, and this trial gives me a concrete example.

Sagimet's June 2025 release and the trade coverage that followed reported dry skin in 6.3 percent of denifanstat patients versus 2.9 percent on placebo, and dry eye in 5.9 versus 3.8 percent.^6,14 Sagimet's own April 2026 investor deck, describing the same trial, reports dry eye at 10.9 percent on drug versus 9.2 percent on placebo, with a footnote about classification.⁷ Same trial, same sponsor, two different dry-eye numbers, and I cannot tell you which one is right because there is no paper. The second pairing, incidentally, shows almost no separation from placebo at all. Peer review exists to catch precisely this kind of thing. Also worth flagging: a dermatologist who presented at Sagimet's investor event described the placebo IGA rate as "roughly 17 percent" in a subsequent interview.¹⁴ The primary source says 14.58. Numbers drift when they live in slide decks.

Fifty-Two Weeks of Safety Data and Not One Efficacy Number

Patients who finished the 12-week trial could roll into an open-label extension, study ASC40-304, in which 240 of them took denifanstat 50 milligrams for a further 40 weeks. Those originally randomized to drug therefore accumulated a full year of exposure.^8,16 Ascletis announced topline results in January 2026.

Zero discontinuations across a year of treatment is a genuinely encouraging tolerability signal, and I do not want to bury it. But two things bother me. The first is that a company sitting on 52-week efficacy data chose to describe it with an adjective. If lesion counts kept falling, that is a number, and it is the number that would tell us whether this is a drug you take for three months or a drug you take forever. The second is the hair.

What "Isotretinoin Without the Baggage" Would Actually Require

The largest network meta-analysis of acne treatments, covering 221 randomized trials, 37 interventions, and 65,601 patients, ranks oral isotretinoin first for total lesion reduction, ahead of every topical combination and every antibiotic.⁵ It reduces sebum production by roughly 80 to 90 percent at standard doses and keeps it suppressed for months after the course ends, which is why a finite course can produce durable remission. That durability is the entire value proposition. You take it for five months and, for most patients, the acne does not come back.

Denifanstat has no head-to-head trial against isotretinoin, or against anything else. Every comparison you will read, including the ones I made above, is cross-trial and therefore soft. Different populations, different baseline severities, and different placebo response rates. The Chinese placebo arm's lesion reduction of 35 percent was unusually high, which makes cross-trial reading even more hazardous than usual.⁶

More fundamentally, denifanstat and isotretinoin may not be the same kind of product. Nobody has published what happens when you stop denifanstat. There is no relapse data, no off-treatment follow-up, no sebum measurement after withdrawal. The pharmacology suggests a suppression drug rather than a remission drug: block the enzyme, sebum falls; stop blocking it, the gland presumably resumes. If that is right, the comparison is not with a five-month isotretinoin course but with an open-ended oral antibiotic, or with spironolactone, which women take for years. That is a perfectly respectable product. It is a different product from the one the phrase "isotretinoin without the baggage" implies.

Nineteen Percent in the Liver, One Patient in the Face

Here is the discrepancy I keep returning to. In FASCINATE-2, denifanstat 50 milligrams for 52 weeks produced alopecia in 19 percent of patients against 4 percent on placebo, in a double-blind trial.³ In the acne extension, denifanstat 50 milligrams for up to 52 weeks produced one case of grade 1 hair thinning among 240 patients, in an open-label study.⁸ Same molecule, same dose, same duration, and roughly a 45-fold difference in the rate of the one adverse event that would matter most to a 22-year-old choosing an acne pill.

There are honest explanations. MASH patients are older, sicker, and more likely to have baseline hair changes from metabolic disease. Ascertainment differs: an open-label acne study relies on patients volunteering a complaint, whereas a blinded liver trial actively solicits adverse events at every visit. But that second explanation cuts the wrong way for the sponsor, because it means the acne study is precisely the design that under-counts diffuse thinning. And the mechanism does not care which disease you have. Hair follicles carry sebaceous glands and depend on local lipid synthesis; the oncology trial at higher doses caused alopecia in three out of five patients.⁴ A dose-dependent effect on hair is exactly what the biology predicts. The US Phase 3 will enrol about 450 adolescents. I would like someone to be counting their hair.

AURORA, the NMPA, and a Company With One Bet

The trial that would answer most of my questions is the one Sagimet just cleared to run. On 13 August 2026 the FDA issued a "study may proceed" letter for AURORA, a US Phase 3 in roughly 800 patients aged 12 and up, of whom about 450 will be adolescents, randomized two-to-one to denifanstat or placebo for 12 weeks and followed by a 40-week open-label extension.^9,10 The co-primary endpoints are IGA success plus absolute change in inflammatory and non-inflammatory lesion counts, the FDA's preferred convention. Screening begins in October, first patient in the fourth quarter, with enrolment expected to take about six months.¹⁰ If it reads out in late 2027 with numbers like the Chinese trial, in an American population that includes teenagers and darker skin, I will revisit this issue and I expect to move the rating.

In China, the marketing application was accepted by the NMPA in December 2025 and a decision is pending.⁸ It is entirely possible denifanstat is approved for Chinese patients before its pivotal trial is published anywhere. Sagimet, meanwhile, raised $175 million in April 2026, holds $257.6 million in cash, and says it is funded through the AURORA readout and a US filing.¹¹ A second FASN inhibitor, TVB-3567, is in Phase 1 with sebum measurements built into the design, and a topical version is in early development.¹⁰ The company is not being coy about its strategy: it is an acne company now.

Julie Harper, founding director of the American Acne and Rosacea Society, presented at Sagimet's investor event in April and discloses consulting for the company among fifteen others.^7,14 Bhatia, Hilary Baldwin, and Sonia Batra have all spoken favourably in the trade press.¹³ These are serious clinicians and their enthusiasm is not manufactured. But I note that the loudest voices on this drug are the ones closest to it, which is the usual arrangement in dermatology and the usual reason I read the primary sources instead.

A Real Drug, Rated on the Evidence That Exists