Compound Monograph  ·  No. 16  ·  Research Use Only

PT-141 Bremelanotide — a melanocortin agonist that acts on the brain

A drug discovered by accident in a programme designed to make a sunless tan, abandoned once as a nasal spray, and approved in 2019 for a narrow female indication on the strength of two questionnaire scales — while the outcome most people would have counted did not move. This is what the evidence supports and what it does not.

Compiled by South Beach Longevity · 2 August 2026
Copyright 2026
Corpus 72 scientific full texts, ~892 printed pages
Metadata layer 121 indexed records · 368 full-text sweep · 70 without open full text
Source project 05 · Therapeutic Peptide Research Library
Constraint No human use, dose, route or schedule is recommended anywhere in this document

Section 01A hormone that did not last long enough

In the late 1970s a pair of researchers at the University of Arizona became interested in a hormone that almost nobody had a use for. Alpha-melanocyte stimulating hormone — α-MSH — is a short peptide released from the pituitary, and its most obvious job is to tell pigment cells to make pigment. In a frog it will darken the skin within minutes. In a human it does something similar and far less dramatic, and in the 1970s it was regarded as a curiosity of comparative endocrinology rather than as the starting point for a drug.

The two researchers were Mac E. Hadley, an endocrinologist who had spent his career on the pigment cells of frogs and lizards, and Victor J. Hruby, a peptide chemist. Their problem was practical. α-MSH is destroyed in the bloodstream within minutes, which makes it useless as a drug and awkward even as a laboratory tool. If you wanted to study what the hormone actually did, you first had to build a version of it that survived long enough to do it.

In 1980 they published one. The paper describes a molecule with two changes to the natural sequence: the methionine at position 4 replaced by norleucine, and the phenylalanine at position 7 flipped from its natural left-handed form to its mirror image, D-phenylalanine (Sawyer et al., 1980). Both changes attack the same weakness. Enzymes that chew up peptides in the blood recognise particular residues in particular orientations; a residue that is the wrong shape, or simply not the residue expected, is much harder to cut. The resulting compound, [Nle⁴, D-Phe⁷]-α-MSH, was twenty-six times more potent than the natural hormone in a melanoma cell assay and resisted degradation by serum enzymes. A companion study in frogs found skin darkening that persisted for six weeks after a single exposure (Hadley et al., 1981).

Section 02The reason for the tan

Once the idea did arrive, it was a good one, and Arizona was the right place to have it. Melanin is not decoration. Eumelanin, the darker of the two pigments human skin makes, absorbs ultraviolet light before it reaches DNA, and people who tan easily get less skin cancer than people who burn. A drug that produced the pigment without the radiation that normally induces it would, in principle, give the protection while skipping the damage that pays for it.

That logic has since been vindicated in a way the Arizona group would have recognised. Their first compound, under the names melanotan-I and then afamelanotide, was eventually approved — not as a cosmetic but as a treatment for erythropoietic protoporphyria, a rare disease in which sunlight causes severe pain. Human studies confirmed that it raises eumelanin specifically (Dorr et al., 2004). The first molecule out of the programme became a real drug, four decades later, for the reason the programme originally gave.

The second molecule is the one this document is about, and it got there by a stranger route.

Section 03Tying the molecule in a knot

Hruby's next move was structural rather than chemical. A short peptide in solution is floppy: it samples an enormous number of shapes, only one of which fits the receptor. If you could lock it into the shape that binds, you would get more potency for the same molecule, and fewer of the exposed backbone bonds that enzymes attack.

The way to lock it is to tie the ends together. In 1989 the group reported a new class of α-MSH analogues built by joining the side chain of an aspartate to the side chain of a lysine, forming a covalent bridge — a lactam — that closes part of the peptide into a ring (Al-Obeidi et al., 1989a). A companion paper the same year reported the biology: a twenty-three-membered ring was optimal, and the best compound was ninety times more potent than the natural hormone in a lizard-skin assay (Al-Obeidi et al., 1989b).

That compound was named melanotan II. It is a cyclic heptapeptide, and if you strip away the notation it is a small ring of six amino acids with a seventh residue hanging off it and an acetyl cap on the end. Inside the ring sit four residues — histidine, D-phenylalanine, arginine, tryptophan — which are conserved across the whole melanocortin family and which the receptors actually read. Everything else in the molecule exists to hold those four in position and to keep enzymes away from them.

Section 04The effect nobody was testing for

Melanotan II went into human testing as a tanning agent. What happened next is the part of this story that everyone has heard and that almost everyone gets wrong in at least one particular.

The version in general circulation is that a researcher on the project injected himself with twice the intended dose and experienced an erection lasting eight hours, and that the researcher was Mac Hadley. The first half of that is documented. The second half is not.

The incident appears in a peer-reviewed review of melanocortin receptors and penile erection, written by a group that included Hruby himself and Hunter Wessells, the urologist who ran the later clinical trials (King et al., 2007). It describes a self-described human pincushion who inadvertently took twice the expected concentration and, to his surprise, had an eight-hour erection along with nausea and vomiting. The review names nobody. Two of its authors were inside the programme and would have known who it was, and they still did not say. The attribution to Hadley by name traces to a single encyclopaedia sentence whose own editors flagged it as inadequately sourced, and which cites the review that names nobody. This document therefore reports the incident and not the name.

It is also worth noting that the anecdote is not load-bearing. The effect turned up in a formal clinical trial, and that trial is citable. In 1996 the Arizona group published a pilot phase 1 study of melanotan II in three male volunteers — a tanning study, single-blind, with alternating placebo days. Buried in its results is the observation that a complex of stretching and yawning appeared to correlate with the onset of spontaneous erections, which the volunteers experienced intermittently for one to five hours depending on the dose (Dorr et al., 1996). Nobody had administered a sexual stimulus. The drug was producing the response on its own.

Hadley later described the discovery in exactly those terms: the sexual actions of the peptide were found accidentally, while the group was studying its effects on human skin pigmentation (Hadley, 2005).

Section 05From a tanning peptide to a sex drug

The clinical follow-up was quick and, for a three-person pilot, unusually convincing. Wessells and colleagues ran a double-blind placebo-controlled crossover study in ten men with erectile dysfunction of psychological origin. Eight of the ten responded. Penile tip rigidity above eighty per cent lasted a mean of thirty-eight minutes on the drug against three minutes on placebo (Wessells et al., 1998). A second crossover study in twenty men found that seventeen responded to at least one of two injections, again without visual sexual stimulation, and — the more interesting finding — that increased sexual desire was reported after sixty-eight per cent of active doses against nineteen per cent of placebo doses (Wessells et al., 2000).

That last number is the hinge of the entire programme. A drug that produces erections is a drug in a crowded field; sildenafil had been approved in 1998 and worked on the plumbing. A drug that produces desire would be something nobody had. The melanocortin programme turned toward that.

But melanotan II itself was not a viable candidate. Its investigators doubted its clinical utility and no further trials were run. It was a potent pigment agonist — that was its original purpose — which meant that using it for sexual function required accepting tanning as a side effect, and it carried a narrow margin between the dose that worked and the dose that caused trouble.

The compound that Palatin Technologies took forward instead was the one this monograph is named for. Chemically it differs from melanotan II by a single functional group: where the parent ends in a primary amide, bremelanotide ends in a carboxylic acid. One nitrogen atom is replaced by an oxygen. Nothing else about the molecule changes.

Section 06What the accident was actually an accident of

It is tempting to file this under serendipity and move on, in the company of penicillin and the microwave oven. That framing is too generous, and it obscures the more interesting point.

The erection was not a random event. It was the predictable consequence of giving a potent, long-acting, brain-penetrant agonist for a receptor family that happens to be expressed all over the central nervous system, at a time when nobody had catalogued what those receptors did outside the skin. The melanocortin receptors were not even cloned until the early 1990s, several years after melanotan II was first made. The Arizona group built a superpotent ligand for a system whose extent was unknown, and then found out what the system controlled by watching what happened.

What makes the story worth telling is not the luck. It is that the same promiscuity which produced the discovery has followed the compound ever since. The receptor that made melanotan II a tanning agent is still there in bremelanotide, and in the approved drug it is no longer the therapeutic target — it is the reason a small number of patients develop patches of darkened skin. The programme changed its mind about which effect it wanted. The molecule did not change its mind about what it did.

Section 07One hormone family, five jobs

To understand what bremelanotide does, and why it does several things at once, you have to start with an unusual piece of biology. A single gene, proopiomelanocortin, produces a long precursor protein that is then cut into pieces — and the pieces are different hormones. Adrenocorticotropic hormone comes out of it. So do the melanocyte-stimulating hormones, alpha, beta and gamma. So does beta-endorphin. One transcript, several messages, depending on which enzymes are present in which tissue.

Those messages are read by five receptors, cloned in the early 1990s and named in order of discovery rather than of importance. All five are G-protein-coupled receptors of the same broad class, and all five conventionally signal by raising cyclic AMP inside the cell.

Two features of this system are worth holding on to. First, it is one of very few receptor families with endogenous antagonists — agouti-related peptide and agouti signalling protein actively switch these receptors off, rather than merely failing to switch them on. Melanocortin signalling is therefore a push-pull system with a resting tone that can be driven in either direction.

Second, MC4R is not a simple on-off switch. It couples not only to the cyclic-AMP pathway but also to at least four other intracellular routes, including one that bypasses G proteins entirely. That matters more than it sounds, and section 10 returns to it.

Section 08What bremelanotide binds, and why nobody agrees

Here is a fact that should be surprising about a drug approved in a major market: there is no published, single-laboratory, five-receptor potency table for bremelanotide. What exists instead is a scatter of measurements made by different groups, in different cell lines, mostly inside papers about other molecules that used bremelanotide as a reference compound. And they do not agree.

The approved labelling states a rank order — MC1R above MC4R, then MC3R, MC5R and MC2R — without printing the numbers behind it. Reviews reproduce that order (Böhm et al., 2024). But when independent laboratories have measured the compound directly, the top two positions swap depending on the system used.

The disagreement has a mundane explanation and an uncomfortable implication. The mundane explanation is that receptor pharmacology is sensitive to how many receptors a cell expresses and what else it expresses alongside them; the same compound can look like a different drug in BHK cells and HEK cells. The uncomfortable implication is that bremelanotide's own characterising paper — the 2003 account by the Palatin group that first described it as a melanocortin agonist for sexual dysfunction (Molinoff et al., 2003) — sits behind a paywall in a conference volume, so the numbers in general circulation have been reconstructed from other people's control experiments.

For the purposes of this monograph the practical position is this: the drug is a non-selective agonist that engages at least four of the five melanocortin receptors at concentrations achieved in use, and MC1R and MC4R are the two that matter clinically. Which of those two it prefers is not settled, and any source that states the rank order without naming an assay is repeating a label statement rather than reporting a measurement.

Section 09The pigment problem is the target problem

One consequence of that non-selectivity is worth stating plainly, because it is the neatest irony in the compound's history.

MC1R activation does exactly what the Arizona group designed it to do. The receptor sits on pigment cells; agonist binding raises cyclic AMP, which activates a transcription factor cascade ending at the enzymes that manufacture melanin, and the balance shifts toward eumelanin. This is the same pathway that ultraviolet light uses: UV damages DNA, the damage stabilises p53, p53 turns on proopiomelanocortin in skin cells, and the α-MSH produced locally activates MC1R on the neighbouring melanocyte. A melanocortin drug simply enters that circuit halfway along.

In the tanning programme this was the therapeutic effect. In the approved drug it is an adverse reaction: focal hyperpigmentation, reported in about one per cent of women receiving bremelanotide on the approved as-needed schedule (Böhm et al., 2024). Nothing about the molecule changed. The indication changed, and the same receptor moved from the benefit column to the harm column.

Section 10Why a brain receptor produces a nauseating drug

MC4R is not confined to whatever nucleus governs sexual desire. It is one of the most broadly distributed neuropeptide receptors in the brain, and the list of things it controls reads like a list of this drug's side effects: food intake, nausea and vomiting, blood pressure, heart rate.

Recent work has begun to explain why those effects can, in principle, be separated. MC4R couples to several intracellular pathways, and in mice they map onto different functions. Deleting the Gα_q/11 pathway from a specific hypothalamic nucleus abolishes an MC4R agonist's appetite-suppressing effect while leaving energy expenditure and glucose handling intact; deleting the Gα_s pathway from the same nucleus abolishes the blood pressure response. In principle a ligand that favoured one pathway over the other could suppress appetite without raising blood pressure — and that is the accepted explanation for why setmelanotide, a later melanocortin drug, has a cleaner cardiovascular profile.

Bremelanotide is classified as a balanced, unbiased agonist: it appears to engage the pathways roughly equally, which is consistent with its clinical profile of appetite effects and pressor effects and nausea and sexual effects arriving together. But that classification is inference rather than measurement. No pathway-profiling dataset for bremelanotide exists in the corpus assembled for this monograph — no bias factors, no transduction coefficients, no comparison of G-protein against arrestin recruitment. Everything said about this drug's signalling bias is extrapolated from its structural near-identity to melanotan II and back-inferred from which side effects patients report.

Section 11What the animals actually show

This is where the popular account of bremelanotide and the primary literature part company, and the gap is wide enough to be the most important finding in this Part.

The account usually given runs: the drug activates MC4R in the hypothalamus, which increases dopamine release in the brain's reward pathway — the projection from the ventral tegmental area to the nucleus accumbens — and that increase produces desire. It is a tidy story, it is repeated in clinical reviews, and it was drawn on one of the commissioned plates supplied for this monograph.

What the primary animal literature supports is considerably narrower.

What is demonstrated. Infusing bremelanotide directly into the medial preoptic area of the hypothalamus increases appetitive sexual behaviours in hormone-primed female rats — solicitations, hops and darts (Pfaus et al., 2007). Infusion into a neighbouring nucleus, the ventromedial hypothalamus, does not. In mice, MC4R is necessary for normal female sexual receptivity, and restoring the receptor only in one population of hypothalamic neurons fully normalises the reflex (Semple et al., 2023). Melanocortin agonists reliably activate the paraventricular nucleus in several rodent species. That is a real hypothalamic mechanism and it is well replicated.

What is not demonstrated, and is in fact contradicted. The dopamine arm of the story does not survive contact with the experiments. A 2025 study in female Syrian hamsters tested it directly and found that bremelanotide, at two doses, failed to enhance the rewarding value of sexual experience in a conditioned place preference paradigm — if anything the drug abolished the preference that saline-treated animals showed. Neither dose changed the sexual reflex itself. Mapping the receptors, the same group found MC4R messenger RNA to be rare in the nucleus accumbens and MC3R absent from it altogether; what MC4R there is sits mostly on fast-spiking inhibitory interneurons rather than on the projection neurons the reward model requires. Their conclusion is blunt: bremelanotide does not act on the ventral-tegmental-to-accumbens reward circuit and does not enhance sexual reward, and despite the drug's approval very little is known about its neurobiological mechanism (Borland et al., 2025).

An earlier study in female rats had already found no effect of bremelanotide on conditioned place preference for sexual experience (Pfaus et al., 2004), and the reviews that describe the drug's neurobiology concede that bremelanotide has no measured effect on dopamine release in the nucleus accumbens. The measurement that is often cited for the dopamine claim used α-MSH, not bremelanotide, injected directly into the ventral tegmental area, not given systemically.

Section 12An unexpected route: the social-context finding

One recent result deserves separating out, both because it is genuinely interesting and because it is routinely over-claimed.

Working in prairie voles, a 2024 study gave the parent compound melanotan II and mapped neuronal activation. In a non-social context the drug activated the paraventricular nucleus and nothing else. In a social context — thirty minutes with an unfamiliar animal of the opposite sex — it activated the paraventricular nucleus and the nucleus accumbens. In animals genetically lacking the oxytocin receptor, the paraventricular activation survived and the accumbens activation vanished entirely; blocking the oxytocin receptor pharmacologically reproduced that. A novel non-social object did not substitute for the social partner, so the requirement is sociality rather than novelty (Ford et al., 2024).

This is a real pharmacology-by-environment interaction: an effect the drug does not produce alone and the situation does not produce alone. It suggests that melanocortin agonism may work by recruiting endogenous oxytocin rather than by acting on reward circuitry directly — which would reconcile the accumbens findings with the negative reward results in section 11.

Three limits should travel with it. It used melanotan II and not bremelanotide. The activation marker used cannot distinguish which cell types responded, so an apparent increase could in principle reflect net inhibition. And the drug simultaneously reduced social investigation and increased self-grooming in both genotypes, so the neural effect happened despite less social interaction rather than because of more.

Section 13The one human mechanistic study

Everything above is animal work. There is exactly one study that looked inside the human brain while this drug was acting, and its results do not resemble the animal model at all.

Investigators at Imperial College London ran a randomised, double-blind, placebo-controlled crossover trial in premenopausal women with acquired generalised hypoactive sexual desire disorder, imaging each participant twice — once on bremelanotide 1.75 mg and once on placebo — while they watched erotic and control videos. Forty were randomised; thirty-one completed both visits (Thurston et al., 2022).

At twenty-four hours, more participants reported increased sexual desire on the drug than on placebo, twenty-one against eight. Inside the scanner, subjective arousal ratings did not differ between drug and placebo at all.

The brain changes were these: increased activity in the right cerebellum, decreased activity in the secondary somatosensory cortex on both sides, and increased activity in the supplementary motor area at the later scan. Connectivity between the amygdala and the insula was altered. That is the complete list.

There were no differences in the hypothalamus, the striatum, the nucleus accumbens, the ventral tegmental area, the amygdala considered alone, or the insula considered alone — and the hypothalamus and striatum were both included as regions of interest specified in advance, so their absence is a result and not an oversight.

The authors interpreted their findings in entirely cortical terms: reduced activity in secondary somatosensory cortex as diminished self-monitoring — the phenomenon clinicians call spectatoring, in which attention to one's own performance displaces attention to the experience — and supplementary motor activation as increased sexual motor imagery. They mapped this onto a model of the disorder as excessive top-down cortical inhibition, drawn from human imaging, and not onto the rodent melanocortin literature at all.

Set the human study beside the animal work and the position is uncomfortable but clear. The animal literature supports a hypothalamic action and refutes a reward-circuit action. The human imaging finds neither, and points instead at cortex and cerebellum. A drug approved in 2019 has, in 2026, no established mechanism in the species it is prescribed to.

Section 14Why it is not a Viagra for women

The comparison is unavoidable in popular coverage and it is wrong in a specific, instructive way.

Phosphodiesterase type 5 inhibitors act on smooth muscle in genital vasculature. They work at the end of the chain, on the physical response, and they require sexual stimulation to do anything: they amplify a signal that is already being sent. That is why they were developed from human tissue pharmacology explaining an effect already observed in the clinic.

Melanocortin agonists act in the brain, upstream of the physical response, and they do not require stimulation. In the early human studies erections occurred without any sexual stimulus at all, which is the pharmacological signature of a central rather than a peripheral drug — and it is why melanotan II can cause priapism, a complication no phosphodiesterase inhibitor produces at therapeutic dose. Direct experiments in rats confirm the point negatively: bremelanotide's precursor produces erection through brain and spinal melanocortin receptors and has no effect on intracavernosal pressure when its access to the central nervous system is bypassed, which excludes a penile site of action (Wessells et al., 2003).

The two classes therefore sit at opposite ends of the same sequence. They are not competitors and not substitutes. The distinction also explains the drug's adverse-effect profile: because the target is a brain receptor with many jobs, the unwanted effects are largely central too, and the most common of them — nausea — is a brainstem phenomenon rather than a stomach one.

Section 15The nasal spray years

Palatin's first serious attempt at a product was not the drug that exists today. It was a nasal spray, and the intended patients were men.

The logic was sound. A peptide cannot be swallowed, and men with erectile dysfunction were a proven market that sildenafil had just created. The nasal route avoided a needle and gave rapid absorption. Through the middle 2000s the programme published a coherent sequence of studies: a phase 1 in thirty-two healthy men, in which ten- and twenty-milligram intranasal doses significantly prolonged the time spent at eighty per cent base rigidity; a phase 2A confirming the twenty-milligram effect with a mean duration around twenty-four minutes and onset near thirty; and subcutaneous studies showing significant erectile responses at four and six milligrams (Diamond et al., 2004; Rosen et al., 2004). A combination study found that low-dose intranasal bremelanotide added to sildenafil produced more erectile response than sildenafil alone (Diamond et al., 2005).

Then the programme stopped, and the reason was blood pressure. The doses that worked by the nasal route were an order of magnitude larger than the dose that is approved today — ten to twenty milligrams against 1.75 — and at those doses the pressor effect that this receptor family produces became the limiting problem.

The lesson the programme drew is the one worth carrying forward, and it is stated well by the commissioned artwork later in this Part: the route and the dose, not the molecule, were what made the compound viable. Nothing about bremelanotide changed between the failed nasal spray and the approved autoinjector. The exposure changed.

Section 16Two trials that no longer count

Any reader going back to the older bremelanotide literature will encounter two substantial randomised trials that must now be set aside, and the reason is not scientific disagreement.

In 2008 the Journal of Sexual Medicine published a double-blind, placebo-controlled, fixed-dose randomised study of bremelanotide in women with arousal disorder. The same year the Journal of Urology published a randomised, double-blind, placebo-controlled trial of intranasal bremelanotide in 342 men who had not responded to sildenafil, reporting positive clinical results in 33.5 per cent of treated men against 8.5 per cent on placebo. Both were single-institution studies from the same investigator.

The first has been retracted. The retraction notice states that after retractions of the author's work by other journals, the journal re-reviewed all of his papers; that its reviewers raised multiple concerning questions about the methodology, results and statistical interpretation; that the author was asked to provide his original data and chose not to respond; and that the journal could therefore no longer verify the results or methods.

The second carries an expression of concern, issued in 2023.

Between them these were among the larger early human trials of the compound, and the male study is still occasionally cited for the proposition that bremelanotide works in sildenafil non-responders. It should not be. This monograph draws no conclusion from either, and readers encountering a favourable bremelanotide result from 2008 should check its provenance before relying on it.

Section 17Choosing a dose

The modern programme restarted with a subcutaneous phase 2b trial in premenopausal women, running from 2011 to 2012 across sixty-seven sites in the United States and Canada (Clayton et al., 2016). Three hundred and ninety-seven women were randomised in equal proportions to placebo or bremelanotide at 0.75, 1.25 or 1.75 milligrams, self-administered at home as desired for twelve weeks.

Two findings decided the shape of everything that followed.

First, the primary endpoint was the number of satisfying sexual events per month, and it was met: a gain of 0.7 events against 0.2 on placebo (p = 0.0180). Remember that, because it does not happen again.

Second, the lowest dose was declared ineffective and the highest was selected, on the stated grounds of optimal efficacy with an acceptable safety profile. The 1.75 milligram dose alone reached significance across the desire and distress instruments; the two higher doses pooled did so only on total and desire scores.

Two caveats travel with this trial and are rarely repeated. Its hierarchical statistical testing broke at the first secondary endpoint, which means the frequently quoted improvements in the desire and distress instruments are nominal p-values rather than protected ones. And the population was not the population that was eventually studied in phase 3 or described on the label: only 23 per cent had low desire alone, while 74 per cent had a mixed picture of low desire and impaired arousal.

Section 18What the phase 3 programme measured

The pivotal programme comprised two identical trials, known as studies 301 and 302 and together as RECONNECT, run between 2014 and 2016 (Kingsberg et al., 2019). Their design deserves attention because several of its features shape how the results should be read.

Women entered a four-week screening period, then a four-week single-blind period in which everyone received placebo and which established the baseline. Only then were they randomised, one to one, to bremelanotide 1.75 milligrams subcutaneously as needed or to placebo, for twenty-four weeks. A fifty-two-week open-label extension followed for those who completed.

The placebo run-in is not a neutral detail. Its stated purpose was to minimise the placebo effect, and it works by letting the initial response to being in a trial happen before the measurement starts. It is a legitimate and common design; it also means the reported treatment effect is measured from a baseline that has already absorbed some improvement.

That exclusion list is the single most important limitation on how these results generalise. Low sexual desire in the real world is frequently associated with depression, or is a side effect of the antidepressants used to treat it. RECONNECT screened out precisely that population. The trials tested the drug in women whose low desire had no identified psychiatric or pharmacological cause — which is what the label's word "generalised" is doing — and the result cannot be carried across to the much larger group in whom it does.

Section 19The two endpoints that were met

Both co-primary endpoints were met, in both trials, in the direction of benefit. There is no ambiguity about that and this document does not manufacture any.

The first co-primary endpoint was the desire domain of the Female Sexual Function Index. It consists of two questions, and it runs from 1.2 to 6.0. Integrated across the two trials, bremelanotide beat placebo by 0.35 points.

The second was a single item — item 13 — from a distress questionnaire, asking how often the woman feels bothered by low sexual desire. It runs from 0 to 4, where the categories are never, rarely, occasionally, frequently and always. Integrated, bremelanotide beat placebo by 0.33 points, in the direction of less distress.

Put the differences next to the rulers they were measured with and the picture changes without a single number being disputed. A gain of 0.35 on a 4.8-point range is about seven per cent of the instrument. A reduction of 0.33 on a four-point range is about a third of one response category: not enough, on average, to move a woman from "frequently" to "occasionally."

The trials also anticipated this objection and answered it with a responder analysis, which is the fairer way to look at a mean difference: a small average can conceal a meaningful benefit in a minority. An independent committee set the responder threshold before unblinding, anchored to a global question asking women how much they thought they had benefited. On that anchor, 58.3 per cent of bremelanotide recipients were responders against 36.1 per cent on placebo in study 301, and 58.2 against 35.4 per cent in study 302.

That is a real difference of about twenty-two percentage points, and it is the strongest number in the programme. It also carries its own caution: more than a third of women on placebo judged themselves to have benefited.

Section 20The endpoint that failed

Beneath the two co-primary endpoints sat a ranked list of secondary endpoints, and at the top of that list was the outcome most people would consider the point of the exercise: the change in the number of satisfying sexual events.

It did not work.

Two consequences follow, and the second is larger than the first.

The immediate consequence is that a drug for low sexual desire, taken as needed before anticipated sexual activity, did not increase the number of satisfying sexual encounters women had. This is the endpoint that carried the phase 2b trial, where it was primary and where it succeeded. It was demoted to secondary for phase 3, and then it failed.

The structural consequence is that the trials' own statistical plan specified hierarchical testing, in which each endpoint is tested only if the one above it succeeded. Because the first ranked secondary failed, every endpoint below it lost inferential status and became, in the paper's own words, supportive and exploratory. The forest plot of secondary outcomes that appears in the primary publication, and the favourable numbers quoted from it in later reviews, all sit below that break.

What was reported in its place was a post-hoc analysis of a different quantity: not the number of satisfying events but the percentage of sexual encounters that were satisfying, which rose 25.0 per cent on bremelanotide against 9.8 per cent on placebo. That is a legitimate observation and it was labelled as post-hoc. It is not the endpoint the trial specified.

Section 21What it costs to take

The adverse-effect profile of this drug is not subtle and was never hidden. Nausea occurred in 40.0 per cent of women on bremelanotide against 1.3 per cent on placebo. Its character is well documented: median onset thirty minutes after the injection, median duration 2.4 hours, roughly 98 per cent of episodes mild or moderate, sporadic rather than after every dose. It caused 8.1 per cent of participants to discontinue. Flushing occurred in 20.3 per cent against 0.3 per cent, and headache in 11.3 per cent against 1.9 per cent.

The cardiovascular effect is the one the programme was built around, and it was characterised properly — with twenty-four-hour ambulatory monitoring at fifteen-minute intervals in the phase 2b trial, each participant compared against her own preceding placebo dose, with an independent blinded hypertension specialist adjudicating (White et al., 2017).

The third characteristic effect is the one the molecule was originally designed to produce. Focal hyperpigmentation was reported in about 1 per cent of the 627 women receiving bremelanotide in the placebo-controlled phase, with a further 1 per cent during open-label treatment. Two participants developed melanocytic naevi, both mild and both resolved, and no melanoma or other skin cancer occurred anywhere in the programme.

Section 22Pharmacokinetics: a deliberately short-acting drug

Almost every peptide drug of the last decade has been engineered to last longer — acylated so that it binds albumin and circulates for days. Bremelanotide went the other way, and that is a design decision rather than an accident.

It is only 21 per cent bound to plasma protein, which is remarkably low for a peptide. Its terminal half-life is 2.7 hours, with a range of 1.9 to 4.0. Peak concentration arrives about half an hour to an hour after injection. Absolute bioavailability by the subcutaneous route is essentially complete, and it makes no difference whether the injection goes into the thigh or the abdomen. It is cleared by hydrolysis of the ring's own amide bonds, with no involvement of the liver's cytochrome enzymes, and eliminated about two-thirds in urine and a quarter in faeces.

For an as-needed drug taken before anticipated activity, short is the correct answer. A compound that produced a 3 mmHg pressor effect and 40 per cent nausea would be intolerable if it lasted a week.

Section 23Interactions, and one that matters

Bremelanotide slows the stomach. That single pharmacological fact generates its whole interaction profile: anything swallowed around the same time is absorbed more slowly and less completely. Two drugs are named specifically in the literature, indomethacin and naltrexone.

Naltrexone is the one worth pausing on. It is used to maintain abstinence in alcohol and opioid dependence, and it works by occupying opioid receptors. A drug that materially reduces its absorption could allow that blockade to lapse. The published guidance is to avoid the combination. Quantitative exposure ratios for either interaction were not found in the corpus assembled here and are not printed.

Against that, the interaction that everyone expects is absent: a phase 1 crossover study in twenty-four healthy volunteers found no pharmacokinetic interaction with alcohol and no clinically relevant blood-pressure change when the two were combined, and the phase 3 trials found no difference in adverse events by alcohol consumption. That is a genuine point of difference from flibanserin, the other approved drug for this indication, which carries a marked alcohol interaction. One qualification: the alcohol study used the intranasal route, not the subcutaneous one.

Section 24Who stayed

The most revealing numbers in the whole programme are not about efficacy. They are about who was still there at the end.

Nearly a third of participants — 411 of 1,267 — did not complete the twenty-four-week core phase, and withdrawal was strongly asymmetric between arms. In study 301, 41.9 per cent of women on bremelanotide withdrew against 16.0 per cent on placebo; in study 302, 43.8 against 28.4 per cent.

Then came the optional open-label extension, in which everyone who entered received active drug. Of those who had completed on placebo, 87 per cent chose to continue. Of those who had completed on bremelanotide, 70 per cent did. The women who had actually taken the drug for six months were markedly less willing to take it for another year than the women who had taken placebo and did not yet know what it was like.

Of the 684 who entered the extension, 272 — about forty per cent — completed the fifty-two weeks. The leading reason for leaving was adverse events, at 21.2 per cent pooled. Among those who had been on placebo and were now getting active drug for the first time, adverse-event discontinuation ran at around a quarter.

One further number puts the whole treatment in proportion. Women in the extension took a median of about twenty-five to twenty-seven injections over some four hundred days — roughly one dose a fortnight — despite being permitted many times that. Whatever this drug is, in practice it was not something most participants reached for often.

Section 25What the label actually says

Bremelanotide was approved on 21 June 2019 under new drug application 210557, held at the time by AMAG Pharmaceuticals. Two features of that approval are worth noting before the wording itself.

There was no advisory committee. The approval letter states that the application was not referred to one because outside expertise was not necessary and there were no controversial or significant safety or efficacy issues that would benefit from committee discussion. That is a striking sentence to find attached to this compound, given that the preceding drug for the same indication reached the market only at its third attempt and after a contested advisory committee, and given the published dispute described in Part Four. There is also no risk evaluation and mitigation strategy — again unlike its predecessor.

The indication is drawn narrowly and the label works hard to keep it there. It covers premenopausal women with acquired, generalised hypoactive sexual desire disorder, defined as low desire causing marked distress or interpersonal difficulty and not due to a co-existing medical or psychiatric condition, problems with the relationship, or the effects of a medication. A separate Limitations of Use section states that the drug is not indicated in postmenopausal women or in men, and not indicated to enhance sexual performance.

Almost everyone using PT-141 today falls outside every one of those clauses.

The label also records the duration of effect as unknown, and states that the optimal window for administration has not been fully characterised. For a drug taken before anticipated activity, that is an unusual admission.

The three warnings

Blood pressure and heart rate. The label reports maximal increases of 6 mmHg systolic and 3 mmHg diastolic, peaking two to four hours after the dose, with a corresponding fall in heart rate of up to 5 beats per minute, returning to baseline usually within twelve hours. Those are maximal figures and sit above the mean placebo-subtracted values from the ambulatory study in section 21 — the two describe different things and should not be conflated. Sixteen days of daily dosing produced no additive effect.

Focal hyperpigmentation. Here the label supplies the dose-response the trial reports do not. At up to eight doses a month, 1 per cent of patients were affected, involving the face, gingiva and breasts. In a separate study, 38 per cent developed focal hyperpigmentation after eight consecutive daily doses, and among those who continued for eight more days a further 14 per cent developed new pigmentary changes. Patients with dark skin were more likely to be affected. And on the question Part Three left open, the label is explicit rather than silent, though not reassuring: resolution was not confirmed in all patients after discontinuation.

Nausea. Forty per cent of treated patients, requiring anti-emetic therapy in 13 per cent and leading to premature discontinuation in 8 per cent. Incidence was 21 per cent after the first dose and fell to about 3 per cent afterwards.

The adverse reaction table, and the number that is not in it

The label's own pooled phase 3 table, covering 627 treated and 620 placebo patients, is the authoritative version of the safety profile and differs in several places from figures in circulation.

The figure that matters most is not in that table but in the sentence beneath it. The discontinuation rate due to adverse reactions was 18 per cent on bremelanotide against 2 per cent on placebo. That is a ninefold difference on the label's own numbers, and it is the regulatory-record version of the odds ratio that the independent re-analysis in Part Four was criticised for surfacing.

One further item appears in the label and nowhere in the trial publications discussed above. A single case of acute hepatitis occurred in the open-label extension after ten doses over a year, with transaminases more than forty times the upper limit of normal and bilirubin six times it, resolving four months after the drug was stopped. The label states that the drug's role could not definitively be excluded, and the agency required three years of quarterly liver-injury safety reporting.

What was required afterwards

Three postmarketing requirements were imposed, and all three concern pregnancy and lactation rather than efficacy: a prospective pregnancy registry study and a retrospective claims-based cohort study, both due to complete in 2030, and a trial in lactating women to measure bremelanotide in breast milk.

That third study has been done. It ran in 2025 under the current owner, enrolled ten women, and completed in November of that year. Its results have not been posted, and the approved labelling still states that there is no information on the presence of bremelanotide in human milk or its effects on a breastfed infant. The measurement exists; the label does not yet reflect it.

Section 26A dispute conducted in print

Most drugs are argued about in conference corridors. This one has been argued about in journals, by name, with replies and counter-replies, and the exchange is unusually informative because both sides published their reasoning.

The critical case was made most forcefully by Glen Spielmans, a psychologist, in two papers four years apart. The first re-analysed the pivotal trials against the regulatory submission (Spielmans, 2021). Its findings were procedural rather than pharmacological, and they were these: that 72.7 per cent of the outcomes listed in the trial protocols were not reported in the primary publication; that fifteen secondary measures which were reported did not appear in the protocols; that no secondary outcome came with a stated rationale or evidence of validity; and that two findings unfavourable to the drug went unmentioned.

Those two findings are worth stating exactly, because they are the sharpest numbers in the debate. Discontinuation caused by adverse events was substantially higher on bremelanotide, with an odds ratio of 11.98 (95% CI 3.74–38.37) — a number needed to harm of six. And on a combined measure of completing the trial and then electing to enter the open-label study, participants preferred placebo: odds ratio 0.30 (95% CI 0.24–0.38), a number needed to harm of four.

The trial investigators replied under the title "Failure of a Meta-analysis" (Kingsberg et al., 2021). Spielmans returned in 2024 with a second paper concentrating on measurement: it reported that eight of the eleven efficacy outcomes specified on the trial registry had never been published, analysed them, and found effect sizes ranging from nil to small; and it found no validity evidence for the categorical response outcomes that had been published (Spielmans & Ellefson, 2024).

A separate critique came from a drug bulletin, framing the approval as a regulatory rather than a scientific event: it observed that in clinical trials flibanserin produced about one additional enjoyable sexual experience every two months and bremelanotide produced none, and argued that the second drug capitalised on the precedent set by the first (Mintzes et al., 2021).

Section 27The instruments measured the drug that validated them

There is a structural problem underneath the statistical one, and it has had less attention than it deserves.

The distress instrument used for the second co-primary endpoint — the Female Sexual Distress Scale for Desire, Arousal and Orgasm — was validated on the bremelanotide phase 2b dataset, by the same development programme that was using it to test the drug. The instrument and the compound were developed in the same exercise.

That is not automatically improper; new indications often require new instruments, and the validation was published. But the validation's own numbers are not reassuring. Test–retest reliability came out at 0.61, below the conventional threshold of 0.70. A confirmatory factor analysis returned a root-mean-square error of approximation of 0.142, which the authors concede did not meet recommended fit criteria. The responsiveness statistic was 0.13, and the authors state the reason plainly: the change in the treatment group was not much bigger than the change in the placebo group.

Nor is the first co-primary above criticism. The desire domain of the Female Sexual Function Index is two questions. The programme itself evidently regarded that as thin: it developed a further instrument, the Elements of Desire Questionnaire, explicitly because researchers and regulatory agencies had expressed concern that those two items might not capture the whole experience of women's desire. That instrument was carried in the pivotal trials as an exploratory endpoint, and its own daily-recall reliability came out at 0.33.

So the position is this: a drug whose approval rests on two questionnaire scales, one of them two items long and one of them a single item, where the distress scale was co-developed with the drug and falls short on its own psychometric criteria. Spielmans's 2024 conclusion — that the benefits are statistically modest and limited to outcomes for which scant validity evidence exists in this population — follows from the programme's own published validation work rather than from any hostile reanalysis.

Section 28The placebo problem

Female sexual dysfunction has one of the largest placebo responses in medicine, and any honest reading of this drug has to sit inside that fact.

A meta-analysis of eight randomised trials using the same functional index found that 1,723 women given placebo improved by 3.62 points on it (Weinberger et al., 2018). For scale, the entire index runs from 2 to 36, and the integrated bremelanotide effect on the desire subdomain was 0.35. Placebo, in this field, is a powerful intervention.

A second analysis, this one by investigators sympathetic to drug treatment, put the effect sizes side by side. Medications for low desire averaged a Cohen's d of 1.0; cognitive behavioural therapy and mindfulness training also averaged 1.0; placebo averaged 0.55 and a waiting list averaged 0.05 (Pyke & Clayton, 2018). Their argument was that recommending psychotherapy over medication is premature because the two perform similarly. The same numbers support a less comfortable reading: a bit more than half of what a medication delivers is delivered by placebo, and neither medication nor psychotherapy is much better than active engagement of any kind.

The pivotal trials acknowledged a placebo responder rate around 35 per cent and attributed it to expectation, anxiety and reward — which is a fair description and also a description of the mechanism the drug is supposed to be acting on.

Section 29Three measures, and only one of them moved

The most economical way to weigh this drug is to notice that the evidence base contains three different kinds of measurement of the same thing, and that they disagree in a consistent direction.

There is the self-report measurement: two questionnaire scales, on which the drug produced small, statistically robust, repeatable improvements. That is the basis of the approval.

There is the behavioural count: how many satisfying sexual events actually occurred. Zero change, in both trials, integrated P = 0.630.

And there is the physiological measurement. Here the finding is easy to miss because it sits in an older paper: bremelanotide was found to lack effect on vaginal responses to sexual stimuli in women with arousal disorder, in contrast to the same programme's finding that it improved erection in response to visual sexual stimuli in men. That contrast is drawn explicitly in a 2025 review of how sexual desire is evaluated, whose author concludes that bremelanotide enhances sexual motivation in men whereas there is no such effect in women (Ågmo, 2025).

Section 30The diagnosis itself is contested

There is one more layer, and it is the strangest.

Hypoactive sexual desire disorder, the condition bremelanotide is approved to treat, no longer exists as a diagnosis in American psychiatry. The fifth edition of the Diagnostic and Statistical Manual, published in 2013, merged it with female sexual arousal disorder into a single category called female sexual interest and arousal disorder. The International Classification of Diseases retained hypoactive sexual desire disorder, for both sexes. A specialist society maintains a third, broader definition. Three authorities, three incompatible constructs, one drug label.

The merger is itself contested among sexual-medicine specialists, largely because the new category lacks validity and clinical-utility data, and because its criteria would exclude from diagnosis a substantial number of women with moderate rather than severe symptoms. The consequence is that the entire evidence base for this drug — including its pivotal trials — rests on the older, superseded criteria, and reviews of the field say so explicitly because a literature on the replacement category barely exists (Clayton et al., 2018).

Underneath the nosology sits a longer-running objection: that low sexual desire in women is substantially a product of circumstance, relationship and social context, and that turning it into a receptor problem misplaces the cause. The strongest empirical version of that objection is not rhetorical. Women diagnosed with the interest and arousal disorder have been found to show genital responses to sexual stimuli in the laboratory that are indistinguishable from those of healthy controls — and in one study, normal in the laboratory but markedly reduced at home, in the same women. If the response is intact in one setting and absent in another, the disorder is situational rather than generalised, and "generalised" is the word in the approved indication.

Section 31Weighing it

The evidence supports the following, and this document commits to it.

The drug does something. Two pre-specified co-primary endpoints were met in two adequately powered, well-conducted, placebo-controlled trials, with consistent direction and magnitude across both, and a prespecified responder analysis with an anchor fixed before unblinding showed a twenty-two-point advantage in the proportion of women who judged themselves better. Nothing in the critical literature disputes those results as results. The dispute is about what they mean.

The effect is small, and smaller than the programme expected. The trials were powered on an assumed desire-domain difference of 1.0 and delivered 0.35. Both integrated differences fall below the thresholds the programme's own phase 2b work identified as minimally important, though the investigators do not apply them that way.

The countable outcome did not move. Satisfying sexual events did not increase, the trial's own hierarchy therefore collapsed below that point, and the physiological measure in women is also null. A reader who wants a single sentence should have this one: the drug reliably changes how women rate their desire and distress, and has not been shown to change what happens.

The tolerability cost is real and is paid by everyone. Forty per cent nausea, an eightfold-odds excess of adverse-event discontinuation, and a revealed preference in which women who had taken the drug were markedly less willing to continue it than women who had taken placebo.

And the mechanism is unknown. Part Two established that the animal model most often cited for this drug is contradicted by the primary animal literature, and that the only human mechanistic study found changes in cortex and cerebellum rather than in any of the regions the model predicts.

None of that makes the approval indefensible. The condition is real, the distress is real, the alternatives are few, and a modest average can contain a meaningful benefit for a minority — which the responder analysis suggests it does. What it makes indefensible is the confident language in which the compound is usually described, both by those selling it and by those dismissing it. The honest summary is that this is a drug of small and uncertain benefit, appreciable cost in tolerability, unknown mechanism, and a narrow licensed population — and that almost everyone taking it today, as Part Five shows, falls outside that population entirely.

Section 32The compound most people take is not the one that was approved

Everything in the preceding three Parts concerns a prescription medicine licensed for a narrow population: premenopausal women with acquired, generalised hypoactive sexual desire disorder, using a fixed 1.75 milligram dose from an autoinjector, no more than once in twenty-four hours and no more than eight times a month.

That is not how most people encounter PT-141. The compound is sold widely online as a research chemical, in vials, to be reconstituted and injected by the buyer, at doses the buyer chooses, overwhelmingly by men. This is the largest practical fact about the compound in 2026, and no monograph that omits it is describing the world.

The scale of the divergence showed up in the assembly of this document before any of its content did. A sweep of the project's literature stores for files naming the compound returned 466 documents. Fewer than twenty were science.

Section 33How the grey market works, and what it borrows

A 2026 review of unregulated peptide use describes the machinery in detail: direct-to-consumer websites, research-chemical vendors, telehealth and longevity clinics, informal sellers, international suppliers, and vendor communities on messaging platforms and forums (Hailu et al., 2026). Products carry "research use only" and "not for human consumption" labelling which, as the review puts it, functions as a liability shield rather than a safety warning — while the same vendors discuss human dosing openly. Certificates of analysis, purity percentages and third-party testing claims are deployed to manufacture an impression of rigour, and the review notes those certificates may not be independent, batch-specific, or sufficient to establish sterility, degradation, solvent quality or even that the tested sample matches the product shipped.

The review's central argument is the one worth carrying, and it is about bremelanotide specifically. It calls the phenomenon legitimacy spillover: approved melanocortin and peptide drugs lend borrowed credibility to unapproved ones. Bremelanotide is named as one of the drugs doing the lending. Its FDA approval — earned in a narrow female indication, on two questionnaire scales, with the countable endpoint null — is used as evidence that PT-141 "works," for people and purposes the trials never examined.

A separate 2026 analysis measured the legitimate side of access across all fifty states and the District of Columbia, and found it uneven in a way that has nothing to do with medicine (Quesada et al., 2026). Thirty-two states permit telehealth prescribing to be initiated remotely and nineteen restrict it; twenty-six states have no compounding outsourcing facility registered within their borders at all. The regional differences are statistically significant, with the Northeast largely high-access and the South largely low-access. The same paper notes that bremelanotide is the only FDA-approved peptide for any sexual medicine indication, and that being approved, it can be prescribed off-label — including for low desire in men.

Where the material comes from, and what regulators have done about it

The supply side is not hidden. It is visible in the American drug regulator's own establishment listings, where alongside the finished product there are nine active bulk active-ingredient listings for bremelanotide, several of them explicitly categorised as bulk ingredient for human prescription compounding, registered to establishments in Shanghai, Qingdao, Nanjing, Shenzhen, Hangzhou and the United States, with entries appearing from 2020 through 2026. This is the documented channel that feeds compounded and grey-market product, and it is a matter of public record.

What is more surprising is what is not on record. Bremelanotide appears on none of the agency's bulk-substance lists for compounding — not in any category of the 503A nominated list, not on the 503B list, and it has never appeared on a pharmacy compounding advisory committee agenda. That absence is real rather than a gap in coverage: the same documents do name the other peptides sold through the same channels, among them BPC-157, TB-500, ipamorelin, kisspeptin-10, epitalon and thymosin alpha-1.

Enforcement has nonetheless touched the compound three times, and the three actions are instructively different. A Florida pharmacy received a warning letter in 2019 for distributing injectable bremelanotide assayed at 114.3 per cent of its labelled content — a superpotency and adulteration finding, not a licensing one. A Kentucky compounder received one in 2020 naming "Bremelanotide (PT-141)" among twenty-two bulk substances the agency considered ineligible for compounding. And an import alert first published in 2020 and still live carries a red-list entry for a Montreal firm shipping "PT 141 Peptide," detained without physical examination on the ground that no approved application exists for PT-141 — the alert naming the firm's own social-media dosing advice as the evidence of intended human use.

Who owns it, and where it is not sold

The commercial history is short and tells its own story. AMAG Pharmaceuticals licensed North American rights from Palatin in early 2017 and held the application at approval in 2019. The two companies mutually terminated that licence in July 2020 — thirteen months after launch — with rights reverting to Palatin. In December 2023 Cosette Pharmaceuticals acquired worldwide rights. Cosette is the current holder of record, the product is currently marketed in the United States, and there are no generics: a search of the regulator's application database on this active ingredient returns exactly one application.

The geography is the more telling fact. Bremelanotide is not approved anywhere in Europe, and the reason is not that it was refused — it was never filed. The European regulator's own medicines dataset, which lists refused and withdrawn applications as well as authorised ones, contains no entry for the compound in any status. It is likewise absent from the British, Canadian and Australian registers, and Canada is notable because AMAG's licence explicitly covered it. A Korean phase 3 bridging trial of 193 women completed in 2023; no bremelanotide product appears on the Korean register in any status.

So a drug approved in the United States in 2019, and described in its own approval letter as raising no controversial or significant safety or efficacy issues, has in seven years not been brought to market by any other major regulator — and in most cases was never put to them.

Section 34What the parent compound does to people

The clearest evidence of what can go wrong with unsupervised melanocortin use does not concern bremelanotide. It concerns melanotan II, and it is reported here for that reason and with that boundary explicit.

A 55-year-old man injected 2 milligrams of melanotan II subcutaneously to induce a tan — the same dose he had used each year for six years — and developed a painful sustained erection lasting thirty hours. Bedside management failed: intracavernosal phenylephrine to a total of 4,500 micrograms with repeated irrigation produced detumescence that reversed within minutes each time. He required operative decompression under general anaesthesia. At fifteen weeks he had new-onset erectile dysfunction, rigidity he rated four out of ten, fibrosis of the erectile tissue on examination, and no response to phosphodiesterase inhibitors (Mallory et al., 2021).

Two prior cases exist in the literature. One involved 6 milligrams and produced systemic toxicity with renal dysfunction and rhabdomyolysis, requiring three days of intensive care. At least five melanomas have been reported in melanotan II users, every one in a person with independent risk factors — fair skin, freckling, family history, or in two cases genetic variants that raise melanoma risk on their own.

None of this is bremelanotide. The doses are higher, the compound is a far more potent pigment agonist, and the priapism cases arise from a molecule whose sexual effect was strong enough that its own investigators doubted its clinical utility. But two features do transfer. The pigmentation effect is the same receptor and is steeply dose-dependent, reaching 63 per cent at supratherapeutic exposure in a controlled study of bremelanotide itself. And melanocortin analogues belong in the differential diagnosis for anyone presenting with priapism — which the treating urologists in the case above argue explicitly, because nobody asks.

Section 35The second act: appetite

The most active current research on bremelanotide is not about sex. It is about weight, and it arrives from the opposite direction to every other drug in that field — an approved sexual-medicine compound being re-examined metabolically, rather than a metabolic drug that happened to cause arousal.

The pharmacology is sound and unsurprising: MC4R is one of the central regulators of food intake, and mutations in it are the commonest single-gene cause of severe human obesity. Two phase 1 inpatient studies in premenopausal women with obesity gave bremelanotide at roughly three to four times the approved dose, one to three times daily, for four to fifteen days (Spana et al., 2022). Caloric intake fell by about 400 kilocalories a day, consistently, from the first treatment day. Weight at day 16 fell 2.1 kilograms against 0.7 on placebo, a difference of 1.3 kilograms (95% CI −1.9 to −0.8).

Four qualifications should travel with those numbers. The doses were supratherapeutic and given daily, where the approved product is as-needed. The longest study ran sixteen days with no follow-up after stopping, so nothing is known about rebound. Waist circumference did not change. And the twenty-four-week phase 3 trials at the approved dose found no clinically significant effect on weight at all. This is a real short-term anorectic effect at high exposure, not evidence of durable weight loss.

The current programme therefore uses it as an adjunct rather than a monotherapy. A phase 2 trial combined bremelanotide with tirzepatide in 108 adults with obesity, running tirzepatide alone for four weeks and then adding bremelanotide or placebo (NCT06565611). It completed in early 2025.

Section 36The melanocortin field's record in obesity

Bremelanotide is entering a field with a long history of failure, and the failures are instructive because they were not all the same failure.

Three different liabilities of the same receptor family — pressor effects, pigmentation, and simple inefficacy — and bremelanotide has demonstrably two of the three, at smaller magnitude, because it is given rarely and at low dose.

The one success is instructive in the opposite direction. Setmelanotide was approved in 2020 for obesity caused by specific genetic defects upstream of the receptor, and later for Bardet–Biedl syndrome. Its advantage is not that it is a better appetite drug in general: in general obesity it produced weight loss under three per cent in most trials and none at all in one. Its advantage is that it works where the endogenous signal is missing, which is a much smaller population and a much stronger effect. It is also not free of the class liabilities — pigmentation occurs in most patients who take it.

The deeper explanation now being offered for the whole record is that signalling matters more than potency: MC4R couples to several intracellular pathways, appetite suppression and blood-pressure elevation run through different ones, and a ligand biased toward the appetite pathway could separate them. That is a coherent hypothesis with mouse genetics behind it. As section 10 recorded, no such profiling has been published for bremelanotide, which is classified as balanced by structural analogy rather than by measurement.

Section 37The indication it is not approved for

There is an awkward coherence running through this monograph that deserves to be stated at the end rather than buried.

The erectile effect is the one melanocortin sexual finding that turns up unbidden and repeatedly, in programme after unrelated programme. It appeared in the 1996 tanning pilot, in the melanotan II crossover studies, in the intranasal male trials, and then as an incidental observation in obesity programmes run by three different companies for entirely different reasons — LY2112688 produced increased libido and erections in phase 1, MC4-NN2-0453 produced sexual arousal in 27 per cent of participants, AZD2820 produced erections at its higher doses, and setmelanotide produces sexual events in about a quarter of male patients.

It is arguably the best-replicated melanocortin sexual effect in the entire literature. It is not the approved indication. The approved indication rests on two questionnaire scales in women, where the physiological measure was null.

Development in men has not stopped. A phase 2 trial of a co-formulated single injection combining bremelanotide with a phosphodiesterase 5 inhibitor, for men who do not respond to phosphodiesterase inhibitors alone, is described as under way in a 2026 review. That is the combination the 2005 intranasal study first suggested, returning twenty years later at a survivable dose and by a survivable route — which has been this compound's pattern throughout.

Section 38What would settle it

This monograph has been more critical than complimentary, so it should be specific about what would change the assessment. Five studies would, and none of them is exotic.

A trial with a behavioural primary endpoint. The countable outcome failed in phase 3 as a secondary. A trial powered on satisfying sexual events as the primary endpoint, in the licensed population, would either vindicate or retire the drug. The phase 2b trial did exactly this and succeeded; nobody has repeated it at scale.

A blinding-integrity assessment. Ask participants at the end which arm they believe they were in. With 40 per cent nausea against 1.3, this is the single cheapest study that would materially change how the existing results are read, and it has not been reported.

A single-laboratory, five-receptor potency panel. The published binding data disagree about which receptor this drug prefers. One competent pharmacology paper would resolve a question that has been open since 2003.

Signalling-pathway profiling. If the appetite and pressor effects run through different intracellular routes, then measuring where bremelanotide sits between them would explain its side-effect profile and predict whether the obesity programme can work.

Laboratory analysis of what is actually being sold. Given that the overwhelming majority of human exposure to this compound now occurs outside the licensed indication, through vials bought online, the purity, potency and sterility of that product is a more urgent public-health question than any remaining efficacy question — and, as section 33 records, nobody has yet published an answer.

Section 39References

Generated from verified NCBI records rather than from recall. Author lists, journal names, volumes, pages and identifiers are taken from the PubMed record for each citation, and the build refuses to run if any identifier fails to resolve.

That gate earned its place during this build. Six of the identifiers first drafted for this list resolved to real but entirely unrelated papers — a multidrug-resistant Enterococcus genome, a microRNA study, a paper on acoustic comfort — because they had been recalled rather than derived. Every one was caught by reading the author and title line the generator prints beside each identifier, and replaced with the record found in the harvest. A populated citation field is not a correct citation field.

Section 40How this document was assembled

The corpus was built in two layers and they are reported separately, because they are not the same kind of evidence.

The external layer queried PubMed and PubMed Central. The PubMed query returned 121 records and a full-text sweep of PubMed Central returned 368 articles naming the compound anywhere in their body. Every candidate then passed through two screens in a fixed order.

The identity screen ran first. "PT-141" is a weak retrieval key: PT is one of the most overloaded abbreviations in the biomedical literature, and a boundaried search for the designation matches prothrombin time, physical therapy, photothermal series, patient numbering and — as this build found — Salmonella typhimurium phage type 141. The matcher therefore accepts an unambiguous name on its own but requires a bare designation to be corroborated by melanocortin subject matter nearby. It rejected a 1992 paper on Salmonella epidemiology, a Cas9 tag-switching method, and a clinical genomics platform, none of which concern this compound.

The substantive-use screen ran second, deciding by document structure rather than by title. Most articles that name a marketed peptide do so once, in a table of agents; that is evidence of the field's attention rather than evidence about the molecule, and pooling it inflates a corpus with nothing readable in it.

The local layer swept the project 05 literature stores and found 466 files naming the compound and passing the identity gate, 435 after removing duplicate copies held in more than one store. That number is not reported as a corpus, because it is not one: 272 were archived web snapshots, 139 were affiliate or commercial blog posts, and 17 were scientific full texts, of which all but one had already been retrieved from PubMed Central. Section 32 gives the breakdown, because the ratio is itself a finding about this compound.

Section 41Evidence handling

Study type is stated in the sentence that reports the finding. On this compound that is not a formality. The animal literature and the human literature point in different directions about mechanism, and the two rodent species most used disagree with each other; a finding phrased without its species attached would be actively misleading.

Two compounds are held apart throughout. Bremelanotide and melanotan II differ by one atom and are routinely conflated, including by two of the papers in this corpus. Harms documented for the parent — priapism requiring surgery, rhabdomyolysis, reported melanomas — are reported in this document as parent-compound events and are never attributed to the approved drug.

Retracted work is excluded from every conclusion. Two randomised bremelanotide trials published in 2008 are affected: one has been retracted, the other carries an expression of concern issued in 2023. Both are listed in the references so that a reader can identify them, and section 16 states what the retraction notice says. No claim anywhere in this document rests on either.

Where the corpus cannot answer a question, the document says so. The volume of distribution, the quantitative magnitude of the naltrexone interaction, the reversibility of hyperpigmentation, the anatomical sites it affects, any signalling-bias profile for this compound, and any laboratory analysis of grey-market product are all absent from the material read for this monograph, and each is named as absent where it arises rather than filled from a neighbouring drug.

70 indexed records carry no open full text. Several are important — the compound's own 2003 characterising paper, the intranasal-era trials, and much of the critical commentary on effect size. Where this document relies on one of those, it relies on the structured abstract and the registry record, and the corpus figure above should not be read as implying that every cited work was read in full.

Commissioned artwork was treated as evidence to check, not as decoration. Six plates were supplied for this monograph and every value printed on them was verified against the corpus. Three were admitted and three were withheld. A receptor-potency plate carried a quantitative axis no published source supports, and whose rank order one of the two measured binding panels reverses. A pharmacokinetic plate drew a concentration curve peaking near ten times the measured value, in contradiction with a parameter printed on its own adjacent panel. A mechanism plate drew, as the drug's mode of action, the precise neural pathway that the most recent primary study set out to test and reported as absent. In each case the erroneous content was the plate's subject rather than an incidental label, and a caption cannot repair a figure whose subject is wrong, because the reader takes the picture. The three admitted plates carry caption qualifications where a printed value differs from the published record.