What Is Gender Dysphoria, and What Does the Evidence Actually Show?

The DSM-5-TR diagnosis is gender dysphoria. It requires a marked incongruence between experienced gender and assigned sex, lasting at least six months, and clinically significant distress or impairment. Adults need two of six listed features, including a strong desire to be rid of one's sex characteristics or to have those of another gender, and a strong conviction of having the typical feelings and reactions of that gender.

The ICD-11 diagnosis is gender incongruence. It requires the marked and persistent incongruence and does not require distress at all. The World Health Organization also moved it out of the mental and behavioural disorders chapter into conditions related to sexual health, on the stated ground that trans identities are not conditions of mental ill health, while keeping a diagnostic code so that care stays accessible.

So one system requires distress and the other does not. Neither is the same as being transgender, and the American Psychiatric Association says plainly that not all transgender people experience dysphoria. Dysphoria is the distress. Incongruence is the mismatch. Identity is who a person is. A person can have the third without the second, or the second without the first.

That leaves at least five different things a study might be counting: self-reported trans identity, a diagnostic code, referral to a service, incongruence without distress, and dysphoria with it. Most disagreements in this subject come from quoting a figure from one of those and applying it to another.

Three sources of prevalence figures differ by about an order of magnitude, and they differ because they count different things. Population surveys, which ask about identity, find the highest rates. Specialist clinic caseloads, which count people who sought treatment, find far lower ones. Administrative health records, which capture only people with a recorded diagnosis, find the lowest. A systematic review found no whole-population study of clinically diagnosed dysphoria exists, so its prevalence has never been directly estimated.

Two further distinctions run through every section below. Evidence from people who sought and received care describes those people, not everyone with the condition. And a comparison between treated adults and the general population can show that outcomes differ; it cannot show that treatment caused the difference, because the two groups differ in many ways besides treatment. The studies that make these comparisons say so in their own text.

No single cause has been established, and the reviews do not claim one. The working picture is a complex interplay of biological, psychological and social factors, with several lines of evidence each pointing somewhere without any of them reaching a mechanism.

Genetics. Twin studies find higher concordance in identical than fraternal pairs, which is consistent with some heritable contribution. A pooled analysis of 463 twin pairs found 21.2% concordance in identical twins against 8.7% in fraternal. The samples are small and drawn from clinics, and concordant pairs are more likely to get published than discordant ones. A Swedish register study found no concordant same-sex pairs and read that as pointing to familial or intrauterine factors rather than genes specifically. No candidate gene has replicated. No adequately sized genome-wide study exists, and the field's own consensus review calls for one.

Prenatal hormones. Girls exposed to high prenatal androgen through congenital adrenal hyperplasia show more cross-gender identification than controls, and a minority later identify as male. People with complete androgen insensitivity, raised female, almost always identify as female. These are natural experiments in populations with atypical hormone exposure and atypical medical histories, and they support prenatal hormones as one plausible influence without establishing them as the cause in the wider population.

Brain structure. Imaging studies report differences between trans and cisgender participants in some regions and not others. The largest pooled analysis, of 803 scans from participants who had not yet taken hormones, found subcortical and surface-area differences and described a distinct brain phenotype rather than a simple shift toward the other sex. Regional findings do not replicate consistently across studies, most earlier work scanned people already on hormones, and the reviews name that as the central confound.

Autism. The co-occurrence is well established, with meta-analytic estimates around 11% against a general population rate of 1% to 2%, and the largest study, of over 640,000 people, found trans and gender-diverse people three to six times more likely to be autistic. The direction is unknown, and the authors say so.

The honest bottom line is the field's own: several mechanisms are plausible, none is demonstrated, and a study capable of demonstrating one would need to follow people from before any of these factors could be measured, which no study has yet done.

Adults who have been through treatment report, in study after study, that it helped them. The report is consistent across qualitative interviews, validated questionnaires and the one randomised trial.

A qualitative meta-synthesis of adult experience of gender-affirming hormones found an overarching theme its authors summarised as a roller-coaster ride, but in the end, worth it, with a minority describing hormones as redirecting rather than relieving their dysphoria. A meta-synthesis of the experience of dysphoria itself, from 20 studies, identified four broad themes: distress around assigned versus experienced gender, the interaction of gender and society, social consequences, and the internal processing of rejection.

The clearest numerical picture comes from a validated measure developed for this population, the Gender Congruence and Life Satisfaction Scale, in a UK sample of 451 trans and 338 cisgender adults. On a five-point scale, trans adults who had received no medical intervention scored 3.03. Those who had received hormones and genital surgery scored 3.88. Cisgender adults scored 4.45. Among trans men, chest satisfaction was 1.54 in the 46 who had had no intervention and 4.24 in the 17 who had taken hormones and had chest surgery. These are cross-sectional comparisons from a single dataset rather than the same people followed over time, and the authors say so. The direction is consistent, but not every subscale moves: in that same trans men comparison, the genitalia and the physical and emotional intimacy subscales showed no significant difference, which the authors read as distress about the genitals persisting, or becoming more prominent, once chest distress is resolved.

What adults criticise is the process rather than the treatment. The systematic reviews of adult experience report gatekeeping, inadequate procedural information, provider competence and long waits as the recurring complaints. A Dutch study found 44% of respondents waiting for care, with median waits over a year for half the services measured, and respondents describing distress and worsening health during the wait. In England, an NHS review of adult services published in December 2025 found a majority of clinics with exceptionally long waiting times and patients often waiting years to be assessed.

The same literature also records adults who valued a thorough assessment, or wished theirs had been more so. Both views are present in peer-reviewed qualitative work. The more common criticism is of unnecessary barriers, but the other view exists and is not a fringe one.

One thing to hold onto about all of this: most of it comes from people recruited through clinics, community organisations or online, not from population samples. It describes the experience of people who answered, and it describes it well. It is not a prevalence figure for every adult with the condition.

Hormones. There is one randomised trial, and it is favourable. In Australia, 64 adults seeking testosterone were randomised either to start immediately or to wait the standard three months. Over those three months, the immediate group showed a decrease in gender dysphoria of 7.2 points against the waiting group, a clinically significant decrease in depression of 5.6 points on the PHQ-9, and a fall in suicidality: it resolved in 11 people in the immediate group, 52% of those who had it at the start, against 1 person in the waiting group, 5%. Adverse events were injection-site pain in seven people and one transient headache. The trial was open-label, short and covered masculinising treatment only, which is why it raises the certainty of the wider evidence without settling it.

Around that trial sit the systematic reviews. A 2025 review that required a treated group and a group who wanted but had not yet received care found 28 outcome studies and graded them moderate to very low certainty. A 2021 review found hormone therapy associated with better quality of life and lower depression and anxiety. Several of its twenty studies found no change, most of all on anxiety, and none found that hormones made any of the three worse. It graded the certainty low because the studies were uncontrolled, small and confounded with other care. The pattern across every review is the same: the direction is consistent, and the designs cannot yet prove it.

One review is frequently misread. The Cochrane review of oestrogen and antiandrogen treatment for trans women found no completed study that met its criteria. That is an absence of qualifying studies, not a finding that the treatment fails, and the review's own conclusion is that there is insufficient evidence to determine efficacy or safety either way. No Cochrane review of testosterone in trans men exists at all.

Surgery. Outcomes differ so much by procedure that any single figure for "surgery" is misleading, and this is where a lot of circulating claims go wrong.

Chest surgery has high satisfaction and low complication rates. Vaginoplasty has high satisfaction with a real reoperation rate, most of it elective. Phalloplasty is the outlier, with urethral complication rates that are high by the standard of any surgery, alongside satisfaction that remains high in the people who have it, because the outcomes they are rating are the ones the procedure was for. Merging phalloplasty and metoidioplasty into one figure, which is common, produces a number that describes neither.

Across all procedures, the evidence that people are satisfied afterwards is stronger than the evidence that surgery causes a change in mental health, because the mental-health studies are the ones without comparison groups.

Why the comparison groups are missing. The strongest test of a treatment gives it to one group and withholds it from another, with the two chosen at random. That cannot be done here. You would have to tell an adult who needs hormones to go without them for a year while someone measured what happened, and no ethics committee would approve it. The Australian trial came closer than anything else on this page by randomising the wait instead of the treatment: everyone received testosterone, and some received it three months later than the rest. Three months is about the limit of what that design can ask of people, which is why it is the only trial of its kind here.

The alternative is to record what happens to people in ordinary care. England did not. The same December 2025 review that found those waiting times found no patient outcomes data at all across the commissioned clinics, meaning no record of how many patients were diagnosed, how many went on to treatment, or what became of them afterwards. The review states that this makes it impossible to properly understand patient outcomes or the safety of these services, and that the gaps place the clinics outside standard NHS quality assurance expectations. Whether those years of waiting would have answered anything about treatment cannot now be established, because the records that would show it were never kept.

Adults accumulate exposure years that shorter studies cannot see, and the Dutch clinic cohort, which has followed people since 1972, supplies most of what is known.

Cardiovascular. Oestrogen raises the risk of venous thromboembolism. That is established, and the size of it depends heavily on which oestrogen. The often-quoted twentyfold figure comes from an earlier era when trans women were prescribed ethinyl oestradiol, a formulation no longer used for this purpose. In the modern cohort the standardised incidence ratio for VTE in trans women is about 4.5 against reference men and 5.5 against reference women; stroke is elevated at about 1.8 and 2.4; myocardial infarction is elevated against women and not significantly different from men. An American cohort of nearly 5,000 adults put the absolute VTE excess at about 4 extra cases per 1,000 trans women over two years and 17 per 1,000 over eight, against cisgender men. Route and formulation change these figures materially, and applying the old numbers to current regimens overstates the risk several times over. Testosterone in trans men shows no consistent elevation in stroke or heart attack, with a raised MI ratio against reference women in one cohort.

Bone. The Dutch cohort found that hormone treatment did not negatively affect bone density over ten years in either trans women or trans men. Fractures came out differently. A nationwide study of patients from the same clinic found that trans women aged 50 and over had more fractures than men their age, 4.4% against 2.4% over three years, and about as many as women their age, 4.2%. Younger trans women and trans men showed no significant increase against either group. Where bone density declines have appeared, they concentrate in people who had gonadectomy and were then under-dosed, which makes continued adequate hormone dosing after surgery the thing that matters.

Cancer. The breast-cancer result is a clean example of a finding that reads two ways depending on the comparison. Among 2,260 trans women on hormones, 15 developed invasive breast cancer, after a median 18 years of treatment. That was about 47 times the rate in cisgender men and about a third of the rate in cisgender women. Both comparisons are correct. Prostate cancer in trans women runs at about a fifth the rate in reference men. Screening is a practical problem: systems keyed to registered sex can fail to invite people who still have a cervix, breast tissue or a prostate, and the studies note it.

Fertility. Hormones impair fertility, and uptake of preservation before starting is low, at 3% to 16% in countries where patients pay for it themselves, with cost the reason most often given. Where counselling is thorough and cost is not a barrier, uptake in one cohort reached 86% among trans women. What happens to stored gametes later is poorly characterised.

Mortality. The Amsterdam cohort of 2,927 trans women and 1,641 trans men, all of whom started hormones at 17 or older, found mortality higher than the general population: for trans women a standardised ratio of 1.8 against reference men and 2.8 against reference women; for trans men 1.8 against reference women and 1.2 against reference men, the second of those not a significant difference. That excess did not decline across five decades. Among the trans women, the causes of death were cardiovascular disease, lung cancer, HIV-related disease and suicide; among the trans men, the raised risk against women in the general population came from suicide and other non-natural causes. The authors' interpretation of that pattern is worth quoting in full: it "gives no indication to a specific effect of hormone treatment, but indicates that monitoring, optimising, and, if necessary, treating medical morbidities and lifestyle factors remain important in transgender health care." A separate analysis of the same cohort found suicide deaths among trans women decreased over time.

The 2011 Swedish study of 324 people who had surgery between 1973 and 2003 found higher mortality, suicide and psychiatric admission than population controls, and is the most-quoted paper in this area. Its authors state in the paper that it is only informative about health after reassignment and that no inferences can be drawn about the effectiveness of reassignment as a treatment, because it compares treated people to the general population rather than to untreated people with the same condition. The lead author has objected publicly to its use as evidence that transition causes suicide.

These are three measures, and a great deal of confusion comes from treating them as one.

Regret after surgery, in adult cohorts with long follow-up, is low. A meta-analysis of 27 studies pooling 7,928 patients found 1%. The Amsterdam cohort found 0.6% among trans women and 0.3% among trans men who had gonadectomy, measured as starting the opposite hormones afterwards. Sweden, counting formal applications to reverse legal gender across fifty years, found 2.2%, declining over the period. The real caveat on all three is loss to follow-up: some studies in the meta-analysis lost over 40% of participants, and regret, when it comes, tends to come late. The honest range is around 1% to 2% with the acknowledgement that the true figure could sit somewhat above it.

Discontinuation is stopping treatment, and it is not the same as regret. A Finnish register of 1,359 people found 7.9% discontinued hormones over a mean 8.5 years. An American military-system cohort of 952 people, average age 19, found about 30% had stopped refilling within four years: 36% of those who started as adults, against 26% of those who started younger; its authors note that a lapsed prescription cannot tell you why, and that prescriptions filled outside the military system were not counted.

Detransition is where the figures diverge most, and they diverge because of who gets asked. A consecutive series of 175 adults discharged from a UK clinic found that 12 of them, 6.9%, had reverted to their original gender role, and that regret was documented for 2 of those 12, which is 1.1% of the 175. The gap between those two numbers, in the same people, is the point, though both are small counts read off clinical notes rather than measured, and 21.7% of that cohort disengaged from the service altogether, so nothing is recorded about what happened to them. A large American survey of 27,715 trans adults recruited through community organisations found that 13.1% of the 17,151 who had ever pursued gender affirmation had detransitioned at some time, 2,242 people, which is 8.1% of everyone surveyed. Among those 2,242, 82.5% cited at least one external reason: pressure from a parent, stigma, difficulty finding work. A smaller 15.9% cited at least one internal reason, including fluctuation or uncertainty about their gender identity, and the two overlap. Every respondent still identified as trans at the time of the survey, so what that figure mostly measures is temporary detransition under pressure. Surveys recruited through detransition communities find the opposite reason profile, with most citing internal reasons and many reporting regret.

Each of those is true of its own sample. None is a population rate, because no population sample of detransition exists. The number a person quotes tells you which study they read.

"Surgery reduces suicide." The study that established this in public argument, in the American Journal of Psychiatry, was formally corrected in 2020 after independent statistical review. The corrected finding is that the reanalysis showed no advantage of surgery on later mental-health treatment use, and that the original conclusion was too strong. The correction also says the comparison that produced the null result was retrospective and confounded, so it withdraws the positive finding without establishing a negative one.

"Transition makes suicide nineteen times more likely." That figure is real and it comes from the Swedish study above, where it is an adjusted hazard ratio of 19.1 with a confidence interval running from 5.8 to 62.9, a range that wide because the study counted few suicide deaths. It compares treated people to the general population and its authors say in the paper that it cannot evaluate whether the treatment worked. Reading it as a treatment effect is the misuse its own authors have warned against.

"41% of trans people attempt suicide." A lifetime figure from a non-probability survey. It describes the people who answered, over their whole lives, and is not a population rate, an annual rate or a post-treatment rate.

"Regret is under 1%." Close to what the adult surgical cohorts show, with the loss-to-follow-up caveat attached. Those are adult figures and do not transfer to other populations.

"Cochrane found hormones don't work." Cochrane found no qualifying studies to review, which is a different finding.

"Oestrogen carries a twentyfold clot risk." True of a formulation that is no longer prescribed for this purpose. The modern figure is about a fifth of that.

"Low-certainty evidence means no benefit." The error underneath most of the others. Low certainty is a statement about study design, not about the treatment, and every review that assigned that grade also reported that the studies pointed in a favourable direction. Evidence of no benefit and evidence of harm are separate findings that the adult literature does not contain.

1. Gender dysphoria requires distress; ICD-11 gender incongruence does not; neither is the same as being transgender. Figures from one construct do not transfer to another, and most cross-source errors start here.

2. No cause is established. Twin data suggest some familial contribution, prenatal hormones are one plausible influence, imaging finds a distinct pattern without a mechanism, and autism co-occurs at several times the general rate with unknown direction.

3. Adults who receive treatment consistently report that it helped, on validated measures as well as in interviews. The one randomised trial is favourable on dysphoria, depression and suicidality. Every systematic review finds the same direction and grades the certainty low because the studies are uncontrolled, which is a statement about design rather than effect.

4. Surgical outcomes differ sharply by procedure. Chest and vaginoplasty have high satisfaction and manageable complication profiles; phalloplasty has high satisfaction alongside urethral complication rates that are high by any surgical standard.

5. Long-term risks are real, quantified, and mostly attributable to formulation, dosing and general health rather than to treatment as such: modern oestrogen carries a VTE risk a fifth of the older figure, bone density is preserved with adequate dosing, and the cohort authors state that the mortality pattern gives no indication of a specific hormone effect.

6. Regret after surgery is around 1% to 2% in adult cohorts. Discontinuation and detransition are different measures with different numbers, and detransition figures depend entirely on who was recruited.