Feminising Hormone Therapy

Evidence Note

What is the question?

Feminising hormone therapy (FHT) uses oestrogen — almost always 17β-oestradiol, the same hormone the ovary makes — usually alongside a medication that lowers or blocks testosterone. Guidelines usually frame the targets as oestradiol and testosterone levels typical of premenopausal women, adjusted to a person’s goals and safety, though treatment does not reproduce the monthly cycle of hormones. The question this note summarises is a practical one: what does that reliably change, what can it not change, and how good is the evidence for each?

What does the evidence say?

Oestradiol produces feminising changes in most adults who take it: breast development, softer and less oily skin, more body fat laid down in a more feminine pattern, less lean mass and muscle size (strength changes less consistently), and finer body hair. The changes are gradual, unfolding over months to years in a fairly consistent order. What hormones do not do is just as clear: they do not undo the voice changes or the skeletal proportions — jaw, brow, height, the bony frame — set by an androgen-driven puberty, although bone turnover is affected, and they reduce established facial hair only modestly. Acne often settles as androgen activity falls, though new acne after starting oestradiol is not rare.

Breast development is the best-measured outcome. Prospective studies show that most measured growth happens in the first year, with slower growth continuing after that, and that measured size over the first three years is modest on average; three years of follow-up does not establish ultimate size. In a large prospective cohort, none of the variables examined — including the oestradiol level in the blood — predicted how much a given person had developed at one year; a later randomised trial found, in an exploratory analysis, greater estimated breast volume at six months with higher oestradiol levels and older age. So the evidence on oestradiol level is mixed, and neither study tested deliberately higher targets. One guideline, from 2016, states that once testosterone is adequately suppressed there is no evidence that pushing oestradiol higher buys more feminisation, and a 2025 systematic review found that the commonly cited target range lacks a primary evidence base — which does not make levels above or below it interchangeable or equally safe.

The choice of testosterone blocker has now been tested against breast development in one small, short double-blind trial. Cyproterone acetate lowered testosterone below 2 nmol/L more often than spironolactone, but no statistically significant difference in breast development was detected over six months. The trial may have been too small to detect a difference in breast volume, and it does not show the two drugs to be equivalent, but it does not show that the more suppressive drug produces more breast development over six months. Two people in the cyproterone group stopped treatment because their mental state worsened, although mean depression scores did not differ between groups and two cases cannot show that cyproterone caused it.

On safety, professional guidance supports individualised treatment with 17β-oestradiol for adults after assessment and informed consent; the balance of benefit and risk depends on the regimen, other health conditions and a person’s goals, and monitoring does not remove risk. Venous thromboembolism — blood clots — is a prominent serious vascular safety signal, although how often it occurs, and how much of it a particular modern regimen causes, remain uncertain. The alarming historical figures were seen chiefly with an older, far more thrombogenic synthetic oestrogen, ethinylestradiol, which is no longer recommended and has largely been abandoned. With modern oestradiol, clot risk remains a concern on a low absolute base; age, a personal or family history of clots, surgery and immobility affect individual risk, and studies with older participants or longer treatment report more clots, though those study-level findings do not show an independent effect of treatment duration, and the available studies mix regimens and treatment histories and do not give a single figure for any particular modern regimen. Compared with cisgender men, heart-attack risk has not been shown to be higher; one recent Dutch study found it lower, which does not show that hormones protect against heart attack. The evidence on stroke is mixed: a US cohort found raised ischaemic stroke among people who started oestrogen, with the statistically clear excess after six years of follow-up (which does not mark a biological threshold), while the most recent Dutch cohort found no statistically detectable increase after adjusting for socioeconomic status. Overall mortality in the longest-running cohort is higher than in the general population. The study’s authors judged that the pattern of causes did not point to a specific effect of the hormones, but the study could neither establish nor exclude a contribution from hormones and could not compare regimens, so that remains an inference rather than a finding.

On mental health, observational studies commonly report reductions in depression and psychological distress during hormone therapy; confidence in that association is greater than confidence in how much of the benefit feminising hormones themselves cause, or how long it lasts: a 2025 systematic review included no randomised trial of starting feminising therapy and judged the uncontrolled studies to be at serious to critical risk of bias. Within the mainly psychosocial outcomes it covered it found no substantive harms, which does not rule out uncommon or long-term harms.

How strong is the evidence?

It is uneven, and worth reading with that in mind. Breast development rests on objective, prospective cohort data — moderate confidence, and the best-measured of the feminising changes. Testosterone suppression by different anti-androgens now rests on two randomised trials, both small and short. Most of the other physical timelines come largely from expert-consensus and clinical-observation tables rather than careful measurement, so they are best treated as indicative. The safety evidence comes from a small number of observational, mostly European cohorts that cannot fully separate the hormones from confounders such as smoking and HIV, and many of them draw on patients of the same Amsterdam clinic, so their agreement is weaker evidence than it looks; and the review identified no randomised trial of the core long-term outcomes. The overall picture is mixed: strongest where outcomes have actually been measured, weaker where figures are extrapolated, and still evolving in the contested areas.

Where are the uncertainties?

Several important questions remain genuinely open. Why breast development varies so widely is unknown — no consistent predictor has been found among the variables examined. Whether progesterone adds meaningful benefit is not established: an open-label randomised trial from Amsterdam, in people on long-term therapy after gonadectomy, has reported its results only in conference reports, with its largest estimates in the arms combining doubled oestradiol and progesterone; the abstract does not establish that the two interact or fully specify how the figures were calculated, and no peer-reviewed paper had been identified as of September 2026, or in a targeted web search on 12 October 2026 — and a separate placebo-controlled trial was terminated for lack of recruitment after enrolling five participants and posted no breast outcomes. Progestogens may also add clot risk, possibly more so with medroxyprogesterone acetate, but the signal rests on very few events. Whether oestrogen raises stroke risk is itself an open question. Which anti-androgen best balances effect against risk over the long term is not settled, although the one trial on breast development detected no statistically significant difference between cyproterone acetate and spironolactone over six months. Whether, and for whom, oestradiol should be continued in later life after gonadectomy has only begun to be studied. How large the clot risk is on a particular modern regimen and route is not known. And robust long-term data on bone health in people taking modern regimens are thin. Much of the standard caution about clot risk and route is extrapolated from studies of cisgender women rather than demonstrated directly in trans women.

Oestrogen reliably produces feminising changes in most adults. How far it will — for breasts, at least — nothing yet examined has consistently foretold.

That single fact — dependable direction, unpredictable degree — is the one most worth carrying into any conversation about starting.

What does this mean for practice?

This note is a summary, not a prescribing guide; decisions about starting, and with which medications, belong with a person and their own clinician. But a few points are worth stating, with the limits of their evidence.

Once testosterone is suppressed, pushing oestradiol higher has not been shown to add feminisation, and in the one trial that compared them on that outcome, the more testosterone-suppressive anti-androgen was not shown to produce more breast development over six months. Route may matter for clot risk: transdermal oestradiol (patch or gel) is associated with lower clot risk than tablets in cisgender women, and with smaller rises in clotting markers in trans women, and is often preferred for those at higher risk, including smokers and older people — though in trans women no adequately powered comparison of clot outcomes by route has been identified: the one cohort that compared routes, counting clots of all kinds, had too few events to tell. Tobacco use with oestrogen is associated with increased clot risk, so stopping is worth raising at every visit — though smoking alone is not an absolute bar to treatment; that does not extend to continued smoking together with a previous clot or other substantial clot risk, where one guideline advises that oestrogen should not be given. Some conditions call for specialist discussion rather than automatic exclusion; an active oestrogen-sensitive cancer is a contraindication, and a guideline and a later haematology review differ on whether oestrogen must stop during treatment of an acute clot — the review supports continuing it only alongside full-dose anticoagulation, on evidence from cisgender women. Each anti-androgen has its own contraindications, assessed separately: cyproterone, for example, must not be used by anyone with a current or previous meningioma, and its product information also excludes liver disease and existing blood clots; spironolactone’s product information excludes, among other things, a raised potassium, Addison’s disease, significant kidney impairment and use with other potassium-sparing diuretics. Cyproterone acetate’s harms — liver toxicity, blood clots, low mood and fatigue, raised prolactin, cholesterol changes and meningioma — need separate assessment, and where it is used, regulators advise the lowest effective dose; in one cohort, average testosterone suppression at 10 mg daily was similar to that at higher doses, though few people started on 10 mg and not everyone was suppressed. Finasteride, sometimes added, now carries UK and European warnings about depression, suicidal thoughts and sexual side effects; dutasteride’s information was updated as a precaution, without an established causal link, and how often these effects occur is not known. Baseline assessment, a discussion of fertility preservation before starting — recovery of sperm production after stopping has been documented, but its likelihood, timing and completeness are uncertain, and the discussion is worth revisiting during treatment and before any removal of the testes — and ongoing clinical and laboratory monitoring are all recommended, though evidence that a particular monitoring schedule reduces major events is limited, and normal blood results do not rule out serious complications: sudden focal neurological symptoms, severe unexplained breathlessness, concerning sudden chest pain or collapse need emergency care, and new swelling or pain in one leg needs urgent assessment. Hormone therapy is not reliable contraception. Screening follows the organs a person has and their hormone-exposure history, and a new breast lump or other suspicious breast change needs prompt assessment whatever a person’s screening eligibility; in England, people registered with their GP as male are not invited for breast screening but can be referred by their GP. In 19 selected older volunteers after gonadectomy, stopping oestradiol for twelve weeks produced menopause-like symptoms in some, which eased on restarting, while average mood and body-image scores did not change apart from a slight fall in happiness; that cannot predict how an individual would tolerate stopping permanently, and there is no established age at which oestradiol must be stopped, and continuation is a decision worth discussing, with a plan for bone health, rather than assuming either way.

Finally, expectations deserve as much care as prescriptions. Breast growth is usually modest and has not been consistently predicted in advance by any factor studied; for some, surgery later is a reasonable part of the plan rather than a sign that treatment failed. The body’s response is real, gradual and individual — and that individuality is not a problem to be solved so much as a fact to be planned around.

Preparation. This note was prepared with the assistance of a large language model (Claude, Anthropic); how its companion review was checked is set out in that review’s Declarations. Licensed CC BY 4.0.