Anti-androgens Compared

Evidence Review · ER-009

Part I — Comparing drugs without a head-to-head trial

1. What the comparative evidence can — and can’t — say

The natural question is “which anti-androgen works best?” For years the honest answer was that the trial to settle it had not been done. The most thorough attempt, a 2021 PRISMA systematic review, searched for anti-androgens in trans women assessed on feminising outcomes — breast size, body composition, facial and body hair — or on change in total testosterone, and concluded from low-quality evidence that cyproterone acetate, medroxyprogesterone acetate and leuprolide may suppress total testosterone more than spironolactone, but that it is unclear which anti-androgen is most effective at feminisation.^[1]

Then one comparison was actually run, and it is the most instructive result in this review. A double-blind randomised trial at one outpatient clinic gave people starting oestradiol either spironolactone 100 mg or cyproterone acetate 12.5 mg daily for six months. Cyproterone suppressed total testosterone significantly harder — 1.48 versus 4.29 nmol/L. And it made no detectable difference to the breasts: no significant difference in breast–chest distance, estimated breast volume, or breast satisfaction.^[8] The agent that clearly won on the testosterone scoreboard showed no advantage on the outcome the scoreboard is meant to stand in for. In that trial, breast-volume change was associated with age and serum oestradiol rather than the anti-androgen — a hint that the oestrogen regimen, more than the choice of blocker, may be the larger lever, though associations are not predictors: the largest prospective cohort found no evaluated clinical or laboratory parameter that reliably predicted breast development.^[8,10]

That single trial is the cleanest illustration of the thesis, though not proof of it across every outcome. Total-testosterone suppression is easy to measure, so it becomes the scoreboard — but it scores only one of the two ways these drugs act. A production-suppressor (cyproterone, a GnRH agonist) drives the total down and looks decisive; a receptor-blocker (spironolactone, bicalutamide) can leave the total barely moved while still blunting the androgen signal at the tissue. Rank the drugs by how far they push the number down and you will overrate the production-suppressors and underrate the blockers — not as a documented pattern of clinician error, but as a property of the wrong measure. Serum testosterone primarily reflects circulating androgen production and metabolism, not androgen-receptor occupancy; even the free or bioavailable fraction, the fix urged in the monitoring review, is still an exposure measure, not a readout of receptor signalling — nor of the local conversion of testosterone to the more potent DHT at the skin and hair follicle (Section 6). The linked questions of which assay to trust and how that free fraction is calculated are set out in the companion Hormone Monitoring & Target Ranges → review.

Three cautions keep this honest. First, one small six-month trial on a single outcome is not proof that these drugs feminise equally: it was not powered to demonstrate equivalence, its confidence intervals still permit a real difference, and — with both arms at fixed low doses (spironolactone 100 mg, cyproterone acetate 12.5 mg) — it says nothing about higher doses, nor about hair, body composition, dysphoria, or the other two agents.^[8] Its own authors drew the modest conclusion — that anti-androgen choice should rest on clinician and patient preference and side-effects, with further research needed — and this review follows them.^[8] Second, on timing: six months captures the phase in which most breast growth is thought to occur — the largest prospective cohort found it modest and primarily in the first six months before plateauing — but not necessarily final development, since guidelines still cite maximal growth at two to three years, so a later divergence between the drugs cannot be excluded.^[8,10] Third, the outcome itself is slippery: there is no validated, universally agreed objective metric of feminisation, and breast growth in that trial varied more than thirty-fold between individuals on the same drug (roughly 20 to 790 mL).^[8] Both ends of the comparison — the surrogate and the outcome it stands in for — are soft, which is precisely why a clean ranking has been so hard to produce.

So the comparison worth making is not “who lowers testosterone most.” It is: what does each drug do, how good is the evidence that it delivers feminisation, and — since that evidence is thin for all of them — what does each cost in safety, monitoring, reversibility and access. Those axes, not a feminising-potency league table, are where the real and decidable differences live. Confidence ratings below use the series scheme — 🟢 reasonably robust, 🟡 moderate, 🔴 limited — as editorial judgements, not formal GRADE.

Confidence: 🟢 High (the one head-to-head feminising-outcome trial found no breast-growth difference despite differential testosterone suppression) / 🔴 Low (any general ranking of these drugs by feminising efficacy).

Part II — The agents, graded

2. Spironolactone

Mechanism — a receptor story, mostly. Spironolactone is an aldosterone (mineralocorticoid-receptor) antagonist, repurposed as an anti-androgen since the 1980s. It acts against androgens by two routes: it is a moderate androgen-receptor antagonist, and it partially inhibits the steroidogenic enzymes 17α-hydroxylase/17,20-lyase.^[1] Its receptor affinity is modest — weaker than the nonsteroidal antagonists discussed later — and how much each route contributes has not been quantified; the mechanistic evidence is largely pharmacological. What matters for interpretation is that the receptor route does its work without necessarily moving the number. In a small pharmacodynamic study of five healthy men, even high-dose spironolactone did not significantly reduce serum total testosterone and produced a transient rise in luteinising hormone; added to oestradiol in trans women it does assist suppression of testosterone toward female concentrations, but its antiandrogen effect is substantially at the receptor rather than through production.^[1] The clearest mechanistic data here are from small studies in cisgender men. The practical consequence follows directly from Section 1: a still-detectable total testosterone on spironolactone does not by itself mean the drug is failing — it can be acting at a receptor the blood test cannot see.

Efficacy evidence. Weak, like all of them, on the outcomes that count. Spironolactone suppresses total testosterone less than cyproterone acetate^[1,2,8] — a genuine finding that, on its own, says little about feminisation for the receptor reason above. And when the two were tested head-to-head, six months of either produced no significant difference in breast development despite that gap in suppression — evidence that, over six months and for breast growth at least, spironolactone’s weaker effect on the total number did not translate into detectably weaker breasts.^[8] That trial used spironolactone 100 mg daily and so does not speak to the higher doses often used in practice. Spironolactone is inexpensive and familiar, and is used against androgen-driven skin and hair effects, though that benefit is extrapolated largely from acne and hirsutism data in cisgender women rather than from trans-specific trials. Its predominance in United States practice owes much to the fact that cyproterone acetate is not FDA-approved there.^[3,9]

Safety — the potassium question, right-sized. Because it antagonises the mineralocorticoid receptor, spironolactone can raise serum potassium. The signal is real but, in otherwise healthy adults, modest: the largest cohort of trans, gender-diverse and nonbinary people on spironolactone found hyperkalaemia — defined here at the screening threshold of >5.0 mmol/L, below the ~5.5 mmol/L usually taken as clinically significant — in 2.5% (8/318), significantly more common over the age of 45 than under it (8.9% versus 1.5%), and rising with age and renal impairment.^[4] A retrospective chart review found no potassium above 5.5 mmol/L in patients with creatinine below 2 mg/dL, and a paediatric-service study likewise found clinically significant hyperkalaemia rare.^[5,6] The Endocrine Society recommends checking potassium every three months in the first year and annually thereafter;^[7] the cohort data have prompted a reasonable argument that this over-monitors healthy people under 45 without renal or cardiac comorbidity, while remaining prudent with older age, renal impairment, or other potassium-raising drugs.^[4,6] Standard spironolactone interactions compound that risk — ACE inhibitors, angiotensin-receptor blockers, other potassium-sparing agents, NSAIDs and trimethoprim — and significant renal impairment is the main setting where the caution is real; in the head-to-head trial the drug’s characteristic biochemical signal was in fact a rise in serum urea and creatinine rather than potassium.^[8] Beyond that, the mineralocorticoid action brings dose-limiting nuisance effects — polyuria, thirst and orthostatic symptoms.

Confidence: 🟢 High (mechanism; that CPA lowers total testosterone more, including in a randomised trial) / 🟡 Moderate (hyperkalaemia risk and its age dependence; no breast-growth difference versus CPA in one small RCT) / 🔴 Low (feminising efficacy on hair, body composition and the longer term).

3. Cyproterone acetate

Mechanism — the effective suppressor, and the most mixed. Cyproterone acetate is a highly reliable testosterone-suppressor — clearly stronger than spironolactone in the comparative data — and mechanistically the least tidy of the four. It is a progestogen that suppresses the hypothalamic–pituitary axis — lowering luteinising hormone, and so testicular testosterone production — and it also antagonises the androgen receptor and dampens steroidogenesis.^[1] With oestradiol it reliably drives total testosterone into the female range, which is why the head-to-head trial found it out-suppressed spironolactone.^[8] The catch, from Section 1, is that the suppression advantage bought no extra breast development in that trial^[8] — so CPA’s case rests not on superior feminisation but on being a dependable suppressor, set against one of the most consequential safety profiles of the four.

The dose has collapsed — and that is the safety story. The most important shift in CPA practice is that the effective dose turned out to be a fraction of the traditional one. In an ENIGI cohort of 882 trans women, 10 mg daily suppressed testosterone as completely as 100 mg — every dose reached below the 2 nmol/L threshold — while prolactin rose and HDL fell as the dose went up.^[16] The 10 mg stratum was small, so that conclusion rests on the overall dose–response pattern and on later low-dose studies rather than on a large 10 mg group alone: a separate cohort found low-dose (10–20 mg) as effective as high-dose (50–100 mg) and probably safer,^[17] and titration work shows even alternate-day or twice-weekly dosing can hold target.^[19] Because CPA’s serious harms are dose- and cumulative-dose-dependent, using the lowest effective dose is the single biggest lever on its safety. Contemporary trans-care practice has converged on 10–12.5 mg/day or less where CPA is used, for as short a duration as the goal allows; regulators and dose-finding studies support that broader principle rather than any single number.

Meningioma — the risk that reshaped practice. The signal that transformed CPA prescribing is intracranial meningioma. In a French cohort of 253,777 women, high cumulative exposure (≥3 g over six months) carried a meningioma incidence of 23.8 versus 4.5 per 100,000 person-years against low exposure (adjusted hazard ratio 6.6), with a steep dose–response — above 60 g cumulative, the hazard ratio reached 21.7.^[15] Meningiomas carry progesterone receptors, a plausible mechanism for a progestogen, and — importantly — the risk recedes after stopping: one year off treatment, the excess had fallen to 1.8-fold.^[15] European regulators responded in 2020: the EMA restricted cyproterone at daily doses of 10 mg or more — the range used in trans care included — to situations where other options have failed, then reduced to the lowest effective dose, and contraindicated it in anyone with a history of meningioma; the reported risk is concentrated at 25 mg/day and above and after prolonged exposure.^[20] This evidence is overwhelmingly from higher-dose use in cisgender women; the risk at the 10–12.5 mg doses now standard in trans care is expected to be much lower but is not well quantified, and long duration still accumulates dose. Trans women are now a large and growing share of CPA users, which makes the signal directly relevant here.

Hepatotoxicity — more common at higher doses, but occasionally severe. CPA’s other serious harm is liver injury. Asymptomatic liver-enzyme rises are common at higher doses, and rare but severe reactions — cholestatic and hepatocellular hepatitis, autoimmune hepatitis, and fulminant liver failure — are documented, mostly at the 100–300 mg doses used for prostate cancer, though the severe idiosyncratic reactions are not strictly dose-predictable.^[18] Liver-function monitoring is therefore standard, the drug withdrawn if enzymes climb, and CPA avoided in pre-existing liver disease.^[18] The ENIGI dosing cohort found no liver-enzyme difference across CPA doses,^[16] so at low trans-care doses the hepatic risk appears smaller — but it is why the enzymes are watched.

Prolactin, lipids and clotting. Three further dose-dependent effects round out the profile. CPA raises prolactin — mild hyperprolactinaemia in the head-to-head trial,^[8] and a clear dose–response in cohorts (roughly doubled on high versus low dose)^[17] — though frank prolactinoma is rare and the alarming early cases were on regimens no longer used. HDL falls as the dose rises,^[16] and the cardiovascular and thrombotic profile is less favourable than some alternatives, consistent with CPA’s progestogenic nature; high-dose cyproterone is among the agents guidance suggests avoiding where thrombotic risk is a concern. Bone loss is a concern with prolonged hypogonadism if oestradiol replacement is inadequate — which is why these anti-androgens are given with oestrogen, not alone. The venous-thrombosis and route evidence, and prolactin monitoring, are graded in the companion Hormone Monitoring & Target Ranges → review.

Fertility — a second signal, and one that has not travelled. This review tells the reader, repeatedly, that anti-androgen choice should be led partly by fertility goals. Until now it has offered no fertility evidence to lead with. There is some, for one of the four agents, and it points at cyproterone.

De Nie et al. examined orchiectomy specimens from 214 trans women who underwent genital gender-affirming surgery in Amsterdam between 2006 and 2018.^[28] Mature spermatozoa were present in 4.7%, all of whom had started treatment at Tanner stage 4 or later. In 88.3% only immature germ cells remained. And in 7.0% — fifteen specimens — there were no germ cells at all. Every one of those fifteen came from someone who had begun medical treatment in adulthood.

The confound is total, and it is a protocol fact. In that programme, testosterone suppression was triptorelin for those who started as adolescents, and cyproterone acetate, 25–100 mg daily, for those who started as adults.^[28] There is no overlap. Age at initiation and anti-androgen class cannot be separated in this cohort, and no analysis can separate them. The authors name four candidate explanations for the adult–adolescent difference — “age, lifestyle (a higher percentage of smokers and alcohol drinkers), higher dosages of estradiol or the use of cyproterone acetate instead of GnRHa” — and age is the first of them.^[28]

But they elaborate a mechanism for one alone, and they act on that one. Whereas a GnRH agonist works by shutting off gonadotrophin secretion, cyproterone acetate is also a direct androgen-receptor antagonist — it blocks androgen action in androgen-dependent organs, the testis among them — and so, they write, “might have more profound and irreversible effects on testicular tissue.” Their conclusion is a prescribing one: “The potential consequence of irreversible infertility might be an extra reason to not prescribe cyproterone acetate anymore.”^[28]

What this evidence is not. It is one cohort. It is cross-sectional histology captured at a single moment, in tissue removed at surgery. The sample is selected several times over before any tissue is examined — 788 people were identified and 214 analysed; oestrogen-monotherapy and spironolactone users were excluded by design; the groups were capped and randomly sampled. It is not a sample of people who start cyproterone. It is a sample of those who later reached, and chose, gonadectomy. The finding rests on fifteen events. And the dose is the historical one: 25–100 mg daily, not the 10–12.5 mg on which contemporary practice has converged (above) — so, exactly as with meningioma, the magnitude of any fertility risk at present-day doses is not quantified. A brief pre-operative interruption of hormones did not change what was found, but a four-week gap is not a test of recovery, and the study does not offer one.

Why it belongs in the comparison anyway — and it is not because the evidence is strong. It is because of an asymmetry this review has already drawn for the same drug. The meningioma excess recedes after stopping: 1.8-fold one year off treatment, against a hazard ratio of 6.6 at high cumulative exposure.^[15] Germ cells that are gone do not come back. A weak signal about an irreversible harm is decision-relevant at a level of certainty at which a weak signal about a reversible one is not — and that is not special pleading, it is what happens when causal confidence and clinical importance are graded separately rather than folded into one mark. Two further things make the case: this is the only histological evidence bearing on fertility for any of the four agents, and the comparator in that cohort — the GnRH agonist — is the arm whose germ cells were retained. Nobody here is being asked to give up a treatment for which there is no alternative.

Graded on four axes — the scheme introduced in ER-015, which separates the four questions a single traffic light conflates. The rest of this review still uses the older combined mark and will be converted at its next version.
Causal attribution — 🔴 limited. One cross-sectional histological cohort; age and anti-androgen class perfectly confounded by protocol; fifteen events; four candidate explanations, of which the drug is one.
Observed effect size — 🟡 estimable once, unreplicated. Complete germ-cell absence in 7.0% (15/214), entirely within the cyproterone-treated adult subgroup; none in the triptorelin-treated adolescent subgroups.
Clinical importance — 🟢 high, if real. Irreversible loss of the tissue on which any later fertility preservation depends. Unlike meningioma, it does not recede on stopping.
Population directness — indirect. Adult surgical specimens from people who reached and chose gonadectomy, on 25–100 mg daily — not the doses now used.

What follows is a conversation, not a contraindication. This does not establish that cyproterone acetate causes germ-cell loss, and this review does not say that it does. What it establishes is that the only histological evidence in existence is consistent with it, that the investigators who produced it thought so, and that they changed their own prescribing as a result — and that none of this has reached a guideline, a summary, or a consent discussion. For a person for whom future fertility matters, that is worth knowing before starting, alongside sperm banking, and alongside the fact that an alternative with comparable suppression exists.

The honest balance. CPA is a highly reliable testosterone-suppressor, and at the low doses now standard most of its dose-dependent harms recede. But it still carries one of the weightiest safety profiles of the four, headed by a real, dose- and duration-dependent meningioma risk whose excess incidence falls substantially after stopping. The defensible posture is the lowest effective dose, the shortest reasonable duration, liver-enzyme and symptom-directed prolactin monitoring, and prompt imaging for new or progressive neurological or visual symptoms.

Confidence: 🟢 High (that CPA suppresses testosterone effectively) / 🟡 Moderate-to-high (that low-dose CPA matches higher doses — small low-dose strata, but consistent later evidence) / 🟢 High (the dose-dependent meningioma association) / 🟡 Moderate (the magnitude of meningioma and hepatic risk at the low doses now used; the cardiovascular/thrombotic profile).

4. GnRH agonists

Mechanism — suppression by pituitary desensitisation. GnRH agonists (leuprolide, triptorelin, goserelin, histrelin) work by a paradox. Given continuously rather than in the body’s natural pulses, they first over-stimulate the pituitary GnRH receptor — a transient testosterone “flare” — then desensitise and downregulate it, shutting off luteinising and follicle-stimulating hormone and, with them, testicular testosterone production.^[7,25] The result is profound, reliable, source-level suppression: in a conference-abstract cohort of trans women on leuprolide, testosterone fell by 89% (432 to 47 ng/dL) within three to six months and stayed there for years.^[21] This is suppression at the source rather than at the receptor or in the periphery — comparable with or greater than CPA depending on regimen — which is why, in Section 3, CPA could not be called the strongest suppressor outright.

Efficacy. On the one thing that testosterone-suppression measures, GnRH agonists are about as good as it gets: a retrospective comparison of leuprolide against CPA, both with oestradiol, found testosterone similarly suppressed in both groups over a year.^[23] No head-to-head trial establishes any feminisation-outcome advantage over the cheaper alternatives — the same evidence gap as everywhere in this review — so the case for GnRH agonists rests on completeness of suppression and a clean safety profile, not on proven better feminisation.

Safety — favourable, with flare, route and bone caveats. This is the agent that carries none of the others’ signature organ risks: no meningioma, no hepatotoxicity, no hyperkalaemia. Its downsides are of a different kind. The initial flare can transiently raise testosterone before suppression takes hold, and warrants thought at initiation. Menopausal-type effects — hot flushes, low mood, reduced libido — appear if the induced hypogonadism is not adequately covered by oestrogen. And bone: sustained suppression of sex steroids lowers bone density, but in adult feminising care the GnRH agonist is given with oestradiol, so bone is protected as long as that replacement is adequate — the bone concern is really a concern about inadequate oestrogen, not the GnRH agonist itself. The distinct question of bone accrual during adolescent puberty suppression is out of scope here, per the boundary set in the Purpose.

The catch is cost and access. What limits GnRH agonists is not safety but price and delivery. They are given by depot injection or implant rather than a daily tablet, and they are substantially more expensive than the oral anti-androgens — often enough to require prior authorisation or specific funding to obtain. Cost and funding — not evidence — are the main reason GnRH agonists are a first-line anti-androgen in some health systems and a rarely-reached option in others. It is the clearest case in this review of access, rather than efficacy or safety, deciding which drug a person actually gets.

The honest balance. GnRH agonists give profound, reliable, source-level suppression with the fewest drug-specific organ toxicities of the four — their main burdens are the flare, the injection or implant, hypo-oestrogenic symptoms if oestradiol is inadequate, and cost. Were cost and route no object they would be an obvious first choice for suppression; in practice, price and access relegate them to settings where they are funded, or where other agents have failed or are contraindicated — for instance to sidestep CPA’s meningioma risk, or spironolactone in renal impairment.

Confidence: 🟢 High (profound, reliable testosterone suppression, and a favourable organ-safety profile) / 🟡 Moderate (bone protection contingent on adequate oestradiol replacement) / 🔴 Low (any feminising-outcome advantage over cheaper agents).

5. Bicalutamide

Mechanism — receptor blockade without testosterone suppression. Bicalutamide is a potent non-steroidal androgen-receptor antagonist: it competitively blocks testosterone and dihydrotestosterone at the receptor. It does not suppress testosterone production — and, by blocking feedback, it may raise luteinising hormone and testosterone; some of that higher testosterone may aromatise to oestradiol, which may contribute to breast development, especially in monotherapy.^[24,25,26] Bicalutamide therefore does the one thing that breaks the testosterone scoreboard completely: breast development has been reported while serum testosterone sits high — above the female range and the usual trans-care targets — because nothing is lowering it. This is the purest illustration in the review of Section 1’s argument: the number does not fall at all, yet the drug is working. Monitoring against a testosterone target is meaningless here; bicalutamide is judged by effect and by safety bloods, not by a suppression number.

Efficacy — promising, thinly evidenced. The data that exist are striking on breast development: in a case series of transfeminine adolescents given bicalutamide after a GnRH analogue was refused by insurers, most developed breasts within six months, many to Tanner stage III or beyond.^[24] But that is close to the whole evidence base — small, uncontrolled, heavily adolescent case series — and major guidelines do not recommend it, noting the data in transfeminine populations are very sparse and safety information lacking.^[25] Its appeal, especially in the United States, is real: it feminises, it is a daily tablet, and it fills the gap left by cyproterone’s unavailability and the cost of GnRH agonists — but the evidence is not yet there to place it alongside the established agents.

Safety — the liver, and the uncertainty. The concern that has held bicalutamide back is hepatotoxicity. It can cause liver injury, and rare fulminant hepatic failure — including fatal cases — has been reported, mostly from its prostate-cancer use;^[25] a hepatotoxicity case in a transfeminine adolescent has been published, though with significant confounders (immunocompromise, recent COVID-19, another hepatotoxic drug).^[27] Against that, a published comparative cohort of transfeminine adolescents and young adults found that low-dose bicalutamide did not raise transaminases clinically significantly versus other anti-androgen regimens^[26] — suggesting the serious harm is idiosyncratic and uncommon at the doses used, rather than a routine dose-related rise. The honest reading: severe liver injury is rare but real and unpredictable, liver-function monitoring is essential, and the thin safety base is itself a reason for caution. Because bicalutamide does not induce gonadal suppression, it does not create hypogonadism in the way CPA or a GnRH agonist does — though skeletal protection still depends on adequate sex-steroid exposure, especially oestradiol.

The honest balance. Bicalutamide is the most mechanistically interesting of the four and the least established. It feminises by receptor blockade rather than by suppression — the clearest demonstration in this review that suppression and feminisation are not the same thing — and it is convenient and increasingly used. But its evidence base is thin and largely adolescent, and its signature risk is a rare, unpredictable, occasionally fatal liver injury. Until better efficacy and safety data in adults exist, it sits as a second- or later-line option, used with liver-function monitoring and clear-eyed consent about how little is settled.

Confidence: 🟡 Moderate (that bicalutamide feminises via receptor blockade and aromatisation — mechanism clear, outcome data thin and mostly adolescent) / 🟡 Moderate (hepatotoxicity real but rare and idiosyncratic; comparative transaminase data reassuring but limited) / 🔴 Low (efficacy and safety in adults, where the evidence is sparsest).

Part III — Skin and hair: the dermatological dimension

6. What the anti-androgens do to skin and hair

Clinically, the changes an anti-androgen produces soonest are often on the skin, not the breasts. Androgens drive the sebaceous gland and the hair follicle, and much of that drive is local: within the skin, 5α-reductase converts testosterone to dihydrotestosterone (DHT), a more potent androgen that binds the receptor more tightly and pushes sebocytes to make more sebum, with acne-prone sebaceous units showing increased androgen metabolism, including 5α-reductase activity, consistent with local DHT amplification.^[13,14] Blunt that signal — by lowering the substrate (the production-suppressors) or blocking the receptor (spironolactone, bicalutamide) — and androgen-driven oily skin and acne often improve. It is a real, quickly-noticed benefit, and for many people a large part of why the drug is worth taking. It is also the bridge back to the monitoring problem: DHT amplification is a tissue event no serum testosterone captures.

What improves. Beyond sebum, androgen blockade acts on the hair follicle. A systematic review of hair outcomes found that feminising therapy — oestradiol with or without an anti-androgen — may reduce facial and body hair and may improve androgenetic alopecia, but that the evidence is mostly subjective grading rather than objective measurement, and the change often falls short of a person’s goals.^[11] Body hair responds more than facial hair, and scalp-hair loss may arrest or partly reverse.^[11,12] Much of the underlying dermatological evidence — spironolactone for acne, hirsutism and female-pattern hair loss — is drawn from cisgender women; the trans-specific data are thin and largely subjective.

The limitation that matters most. It is as important as the benefit, and routinely misunderstood: established terminal facial hair is usually not eliminated by anti-androgens alone. A male-pattern beard is made of follicles already transformed from fine vellus to coarse terminal hairs during puberty; withdrawing the androgen signal thins and slows new growth but does not turn those follicles back — one systematic review found facial hair decreased in only around 28% of adults on feminising therapy, against a much larger effect on body hair.^[12] Durably removing or reducing terminal facial hair means physically targeting the follicle: electrolysis — described regulatorily as permanent hair removal, and effective on hair of any colour — or laser, usually framed as permanent hair reduction, whose results depend on melanin contrast (hair colour and skin phototype) as well as device, fluence and operator skill. The practical point is one of expectation — an anti-androgen is not a substitute for hair removal, and saying so plainly prevents real disappointment.

By agent. The skin and hair effect follows the mechanism. The receptor-blockers act directly at the sebaceous-gland and follicle receptor — spironolactone has the longest dermatological track record, in acne, hirsutism and female-pattern hair loss, almost all of it in cisgender women^[13] — while the production-suppressors (cyproterone acetate, GnRH agonists) work by lowering the DHT substrate. 5α-reductase inhibitors, which block the testosterone-to-DHT step itself and are sometimes added for scalp hair or persistent virilised features, are out of scope here and covered separately. The oestrogen side of the skin story — barrier, collagen, hydration and pigment — is graded in the companion Skin Effects of Oestrogen → review.

Confidence: 🟢 High (the androgen/DHT–sebaceous mechanism) / 🟡 Moderate (that anti-androgen therapy improves acne and oily skin in transfeminine people specifically — strong mechanistically and in cisgender data, thinner in trans-specific outcomes; and the magnitude of hair change and AGA improvement, mostly subjective data) / 🟢 High (that established terminal facial hair is usually not eliminated by anti-androgens alone and needs physical hair removal).

Part IV — Choosing between them

7. The safety-led comparison

Pull the four sections together and the shape of the decision is clear. On feminising efficacy the agents barely separate — the one head-to-head trial found no breast-growth difference, and no adequately powered trial ranks all of them across the outcomes that matter (Section 1). What separates them is everything else: how completely they lower testosterone, what each costs in safety, and what each demands in route, price and monitoring. So the practical question is not “which is best” but “which risk can this person least afford, and which practical constraints bind?” — and the answer falls out of the comparison below.

How the choice actually gets made. Read the table from the person, not the drug.

• Future fertility that matters and is not yet secured gives a reason to weigh cyproterone against the alternatives, and to have the fertility-preservation conversation first. The only histological evidence in this field found complete germ-cell absence exclusively in the cyproterone-treated arm — but drug and age at initiation are perfectly confounded there, the doses were the historical 25–100 mg, and the finding rests on fifteen specimens. It does not establish causation. It is also the one signal in this review that does not reverse on stopping, and the comparator whose germ cells were retained is a GnRH agonist.^[28] (Section 3)
• A meningioma history, or reluctance to carry that risk, steers away from cyproterone — toward spironolactone, a GnRH agonist, or, with its own caveats, bicalutamide.
• Renal impairment, older age, or other potassium-raising drugs steer away from spironolactone.
• Liver disease steers away from both cyproterone and bicalutamide — the two with hepatic signals.
• Cost or formulary limits — the commonest real-world constraint — steer away from GnRH agonists however clean, toward the cheap oral agents; conversely, where GnRH agonists are funded they are often first-line on the strength of their safety.
• A need for the most complete suppression, or failure or contraindication of the oral agents, points to a GnRH agonist.
• A wish to avoid injections rules out GnRH agonists in practice.

Two threads run under all of it. First, the monitoring a drug needs is part of its cost: spironolactone wants potassium (context-dependent), cyproterone wants liver enzymes and dose-and-duration discipline, bicalutamide wants liver enzymes — and none is well served by chasing a testosterone target, least of all bicalutamide, which is designed to leave the number high. Second, whichever agent is chosen, bone and much of feminisation depend on adequate oestradiol, not on the anti-androgen: the blocker manages testosterone, and oestrogen drives many of the feminising changes, with the anti-androgen contributing differently by tissue.

The honest synthesis of the whole review: with comparative efficacy a near-blank, anti-androgen choice is a safety-and-practicality decision matched to the individual — not a ranking. The most defensible choice is individualised — by contraindications, safety profile, monitoring burden, availability, cost, route and the person’s own priorities — and the best anti-androgen is simply the one whose risks and demands fit the person in front of you.

Confidence: 🟡 Moderate. The individual safety signals and practical facts are well established (Sections 2–5); the comparative weighing of them here is a reasoned framework, not a trial-derived algorithm.

Part V — Evidence and take-homes

8. Evidence quality

The review’s grading, stated plainly. Two things sit on firm ground and one does not. What is firm: the mechanisms, the individual safety signals, and the completeness of suppression — the meningioma dose–response, spironolactone’s quantified potassium risk, GnRH agonists’ castrate-level suppression, and the fact that bicalutamide leaves testosterone high are all well established. What is weak: the comparative feminising efficacy that would let anyone rank the agents — one small six-month trial, and no more (Section 1). And what is intermediate: bicalutamide’s adult efficacy and safety, where the evidence is sparse and skews adolescent, and the trans-specific dermatological outcomes, strong in mechanism but thin in quantified data. Ratings use the series scheme — 🟢 reasonably robust, 🟡 moderate, 🔴 limited — as editorial judgements, not formal GRADE. See Reading the Evidence.

9. Frequently asked questions

Which anti-androgen is best? No trial ranks them on feminisation, and the one head-to-head comparison found no difference in breast development. The choice is made on safety and practicality, not a league table. (Sections 1, 7)

My testosterone is still detectable — is the drug failing? It depends on the drug. Spironolactone and especially bicalutamide act at the receptor and can leave the number high while working — bicalutamide is designed to. On cyproterone or a GnRH agonist, a stubbornly high level more often means under-dosing or missed doses. (Sections 2, 5)

Does cyproterone affect fertility permanently? Possibly, and this is the least-discussed thing about the drug. In the only study to look at testicular tissue directly, complete absence of germ cells was found in 7% of specimens — and every one came from someone who had started treatment as an adult, when the protocol used cyproterone rather than a GnRH agonist. Drug and age cannot be separated in that study, the doses were the old high ones, and it rests on fifteen specimens, so it does not prove cyproterone causes it. But the investigators thought it likely enough to say so in print and to change their own prescribing. Unlike the meningioma risk, this one would not reverse if you stopped. If future fertility matters to you, it is worth raising before you start — and worth asking about sperm banking. (Section 3)^[28]

Is cyproterone too dangerous because of meningioma? The risk is real but dose- and duration-dependent, and it largely reverses after stopping; the low doses now standard carry far less cumulative exposure. The sensible course is the lowest effective dose, the shortest reasonable duration, and prompt imaging for new neurological symptoms. (Section 3)

Why can’t I get cyproterone in the United States? It has never been FDA-approved there; spironolactone or bicalutamide are used instead. (Sections 3, 5)

Will an anti-androgen remove my facial hair? It can thin and slow growth, but established terminal facial hair is not eliminated — that needs electrolysis or laser. (Section 6)

Is the anti-androgen what feminises me? Largely the oestrogen: it drives many feminising changes and protects bone, while the anti-androgen reduces androgenic signalling and may contribute differently by tissue — skin, hair, libido. Adequate oestradiol matters more than the choice of blocker. (Sections 6, 7)

10. Clinical bottom line

1. There is no evidence-based “best” anti-androgen; the one head-to-head trial found no feminisation difference. Choose on safety and practicality.^[1,8]
2. Judge a receptor-level blocker by effect, not by a testosterone target — bicalutamide leaves the number high by design.^[8,24]
3. Cyproterone: lowest effective dose (10–12.5 mg), shortest reasonable duration; watch liver enzymes; image promptly for new neurological or visual symptoms; avoid with a meningioma history.^[15,16,20]
4. Spironolactone: cheap and safe in healthy younger people; check potassium with age, renal impairment, or other potassium-raising drugs.^[4]
5. GnRH agonists: the most complete suppression and cleanest safety, but cost and injection limit access.^[21]
6. Bicalutamide: mechanistically appealing and convenient, but thin adult evidence and a rare, unpredictable liver risk — second-line, with liver monitoring.^[24,25]
7. Whichever agent is chosen, bone and most feminisation depend on adequate oestradiol, not on the blocker.