Eden Openly
Our research
Independent, open and checkable research on gender-affirming care: what the evidence shows, how the guidance that governs care measures up to it, and where the errors in between are.
Eden Openly publishes evidence reviews, appraisals of guidance and policy, and comparative studies of how care is provided. Every finished piece is free to read, licensed for reuse, and deposited in a public archive with a permanent identifier. Our method, our evidence standards and our corrections are published in full, so that the work can be checked rather than trusted.
Our contribution
What Eden Openly has added, as distinct from the research we synthesise.
What we have found
In medical evidence, most errors are introduced after the research, not in it.
In building one review, we traced the evidence chain behind a national policy consultation and verified each link against its source. Roughly one error in five was in the primary papers. The rest were introduced downstream: in an appraisal table, in a systematic review’s one-line summary, in an uncorrected author manuscript, in a news report, and in our own reading. None was detectable by checking that a citation resolves. All were visible to someone who opened the source.
NHS England’s March 2026 evidence reviews contain errors that change what they appear to show.
Three non-significant comparisons are reported as significant, and one significant result favouring treatment is reported as non-significant. A cardiometabolic harm signal in adolescents takes no account of the treated group being tested about three times as often as controls, which would weaken the signal rather than strengthen it. We reported 21 items to NHS England in July 2026. At our last check, on 1 September, none had been resolved. Correcting every item would not change the reviews’ overall very-low-certainty conclusions; it would change what several of their individual findings say.
Reporting errors at source can work.
After our letter, the European Journal of Endocrinology directed a corrigendum to a 2024 paper whose abstract and results disagree on bone density. It is the first correction outcome from our rule of reporting errors to the people who can fix them before publishing anything about them. Until the corrigendum appears, the register records it as directed, not corrected.
What we have built
A register of how long errors in evidence survive
Every defect we find in published evidence, our own included, recorded with what was reported, to whom and when, and what has changed since. The measure is the interval between a report and a correction.
A deposit standard enforced in code
From ER-015 onward, every review states how well each source was read; sources are checked weekly for retractions, and a review cannot be built if a check fails. Every review carries its own correction record.
A certainty scheme that asks four questions, not one
Causal attribution, effect size, clinical importance and population directness are graded separately. “The evidence is weak” usually conflates them.
One method for every guideline
85 guidelines, policies and rulings from 29 countries, appraised in an order fixed in advance, with every statement graded and every appraisal independently verified.
What we have published
Fifteen evidence reviews and appraisals, each deposited with a DOI, versioned, and free to read and reuse under CC BY 4.0. One pre-registered study is under way, the method for appraising the guidelines is registered, and the Cass Review appraisal is being registered.
What we are trying to do
- Make the evidence readable by anyone. Understanding should not be limited to specialists. Every review comes with a short Evidence Note, and everything we publish is free, under CC BY 4.0.
- Describe uncertainty honestly. Our aim is not to remove uncertainty from complex subjects. It is to describe it clearly enough that readers can reason with it.
- Hold every guideline to the same standard. Guidance on gender-affirming care points in different directions. We are appraising all of it, by one method, whichever way it points.
- Get errors in the evidence corrected at source. When we find a defect in someone else’s work, we report it to the people who can correct it before we publish anything about it, and we record what happens next.
Last updated 6 October 2026
What the evidence shows
These findings come from the research our reviews synthesise. They are other researchers’ results, which we have read, graded and set out plainly, with the certainty each deserves and including those that cut against positions we hold.
More oestrogen doesn’t mean more breast growth
In the largest prospective study, blood oestradiol level did not predict breast growth. Given adequate testosterone suppression, higher oestradiol did not enhance breast development.
Graded robust
Stronger suppression, same result
In a randomised trial, cyproterone suppressed testosterone far harder than spironolactone (1.48 against 4.29 nmol/L) and made no detectable difference to breast development. One small, six-month trial: no detected difference is not proof of equivalence.
Graded robust (testosterone as a poor surrogate)
→ ER-009
Testosterone is not contraception, and does not appear to cause permanent infertility
Ovulation can continue without periods, and 80% resumed menstruation within six months of stopping. Two opposite myths, both wrong. Conception and live-birth rates after stopping remain uncertain.
Graded robust (not contraception); moderate (fertility recovery)
→ ER-007
Clot risk rises; heart-attack risk does not
Feminising hormone therapy is associated with roughly a doubling of venous clot risk. Yet in the largest registry, heart attacks were recorded at half the general-population male rate and stroke at the same rate; another large cohort found no difference. This is observed incidence, not proof of protection.
Clot risk: direction established, size less so · Heart attack: lower or no different
→ ER-018
A chosen name, used everywhere
Trans young people who could use their chosen name at home, at school, at work and with friends reported 71% fewer symptoms of severe depression and 65% fewer suicide attempts. One observational study of 129 young people in the United States.
Graded moderate
→ ER-001
Hormone therapy and depression
A 2023 review of 46 studies found gender-affirming hormone therapy consistently reduced depressive symptoms and psychological distress, and an open-label randomised trial found a clear fall in depression scores within three months. Most studies are observational, and quality-of-life and suicide outcomes are less certain.
Graded moderate (depression and distress)
Legal protection did not make patients feel safe
State laws protecting trans patients did not reduce avoidance of healthcare in any subgroup studied. Felt safety appears to come from how people are treated, not from what the law says. A single study, in the United States.
Graded moderate
→ ER-004
Puberty blockers: benefit unproven, bone cost clear
A psychological benefit from puberty suppression specifically is not established: the evidence is mixed and low in certainty. That suppression reduces bone density accrual during treatment is graded robust. Absence of high-quality evidence of benefit is not evidence of no benefit.
Benefit: limited · Bone: robust
→ ER-014
Our own errors in our favour, corrected
An external reviewer caught a draft of ER-018 saying therapy “substantially reduces depression and suicidality”, an overclaim in the direction of our own view. In ER-015, a correction restored adverse cardiometabolic associations with testosterone in adolescents that an earlier draft had wrongly removed.
Both logged in the reviews’ correction records
The most-cited oestradiol target has no outcome data behind it
The 100–200 pg/mL target is one of the most confidently cited numbers in the field and one of the least validated: a systematic review found no supporting outcome data, and guidelines disagree on the number. This is an appraisal of the target, not an argument against monitoring.
Graded limited: an absence of evidence
→ ER-008
In progress
Appraising the guidelines
Guidance appraisal programme · 85 documents
Every major guideline, policy statement, government evidence review and court ruling on gender-affirming care, from 29 countries and international bodies, appraised by one method and each given a published finding. The order was fixed in advance. First come WPATH Standards of Care 8, the American Academy of Pediatrics 2018 policy statement and the Endocrine Society’s 2017 guideline.
- Method registeredThe programme standard and the appraisal protocol, registered publicly on the Open Science Framework. 7 October 2026.
- First appraisals set upWPATH SOC-8 resumes under the registered method; the next two documents are registered before anyone opens them.We are here
- AppraisalOne document at a time, in the published order.
- Verification and depositEach appraisal checked by an independent human verifier, then deposited with a DOI.
→ The programme and the full list of 85 · Method registration 10.17605/OSF.IO/SY4GP
Appraising the Cass Review
Guidance appraisal · Meta-appraisal
A structured appraisal of the Cass Review’s final report and the seven systematic reviews and survey it commissioned from the University of York: were they soundly done, and do the Review’s conclusions follow from them? No findings will be published until the whole appraisal is complete, and then all of them will be released together.
- Methods registrationThe full protocol, frozen publicly before any scoring.We are here
- Claim registrationThe Review’s claims to be appraised, frozen before any is scored.
- AppraisalScoring in a random order, under the registered rules.
- Single releaseEvery result, record and correction, published together.
Models of Gender-Affirming Care in Asia
Comparative policy study · Registered
What the health systems of South Korea, Japan, Taiwan, Hong Kong, Thailand, Malaysia and Singapore formally require, provide and charge for gender-affirming hormones and surgery, read from the rules themselves in their original languages. We have found nothing comparable for the region.
- Protocol registrationFrozen on the Open Science Framework before any extraction. 30 September 2026.
- South Korea pilotTesting the method on real sources. Recruiting the Korean extractor and verifier.We are here
- Extraction and verificationEvery non-English source read in the original by a paid native-language extractor, then independently checked.
- Correction round and releaseChecked for factual errors by clinicians and community organisations, then released openly with a DOI.
→ The study · Call for extractors and verifiers · Registration 10.17605/OSF.IO/UVHDA · Archive 10.5281/zenodo.23057857
Erythrocytosis and Raised Haematocrit in Testosterone Therapy
Evidence review · ER-023 · Working draft
How much testosterone raises haematocrit, where the usual 54% action threshold comes from, and what is known about clotting risk in trans-masculine people. A working draft is published and deposited, with its correction record. Verification of the remaining sources is under way.
- Working draft depositedVersion 1.0, with its correction record. 19 September 2026.
- VerificationThe remaining sources read and annotated against the primary papers.We are here
- Verified versionA new version deposited under the same concept DOI, with every change logged.
→ Review · Evidence Note · Correction record · DOI 10.5281/zenodo.22830379
Evidence Transmission Register
Ongoing register
Defects we have found in published evidence, from national evidence reviews and journal articles to our own work, recorded alongside what we reported, to whom, on what date, and what has changed since. Our own errors sit in the same register as everyone else’s. The measurement is not the error. It is the interval between reporting a defect and anything changing, and whether anything changes at all.
- VerifyThe defect is checked against the document itself. Nothing unverified leaves the register.
- ReportOne notice, once, to the people who can correct it.
- RecordWhat happened, and when, kept public whether or not anything changes.Ongoing
Published
Each Evidence Review comes with a short Evidence Note that summarises it in about five minutes. Each DOI below resolves to every version of the work. Earlier versions stay available.
Hormone therapy
| Work | Permanent record | Read | |
|---|---|---|---|
| ER-005 | Feminising Hormone Therapy | 10.5281/zenodo.22868536 Version 2.2 |
Review · Note |
| ER-007 | Masculinising Hormone Therapy | 10.5281/zenodo.21268487 Version 1.0 |
Review · Note |
| ER-008 | Hormone Monitoring & Target Ranges | 10.5281/zenodo.21274735 Version 1.0 |
Review · Note |
| ER-009 | Anti-androgens Compared | 10.5281/zenodo.21284343 Version 1.2 |
Review · Note |
| ER-013 | 5α-reductase Inhibitors: Finasteride and Dutasteride in Feminising Therapy | 10.5281/zenodo.21295653 Version 1.0 |
Review · Note |
| ER-018 | Venous Thromboembolism, Myocardial Infarction and Ischaemic Stroke in Feminising Hormone Therapy | 10.5281/zenodo.21417059 Version 1.0 |
Review · Note |
| ER-023 | Erythrocytosis and Raised Haematocrit in Testosterone Therapy (working draft) | 10.5281/zenodo.22830379 Version 1.0 |
Review · Note |
Adolescents
| Work | Permanent record | Read | |
|---|---|---|---|
| ER-014 | Puberty Suppression in Adolescents | 10.5281/zenodo.21306155 Version 1.0 |
Review · Note |
| ER-015 | Masculinising and Feminising Hormones in Adolescents | 10.5281/zenodo.21324879 Version 1.4 |
Review · Note · What it does not conclude |
| GA-001 | Grading the Guidance: KCSIE 2026 on Children Questioning Their Gender | 10.5281/zenodo.22285831 Version 2.0 |
Appraisal · Resources for schools |
The experience of care
| Work | Permanent record | Read | |
|---|---|---|---|
| ER-001 | The First Five Minutes of Care | 10.5281/zenodo.21115059 Version 1.0 |
Review · Note |
| ER-002 | Readiness, not Diagnosis | 10.5281/zenodo.21155889 Version 1.0 |
Review · Note |
| ER-003 | Chosen Names and Pronouns | 10.5281/zenodo.21155995 Version 1.0 |
Review · Note |
| ER-004 | Healthcare Avoidance | 10.5281/zenodo.21156001 Version 1.0 |
Review · Note |
Skin and hair
| Work | Permanent record | Read | |
|---|---|---|---|
| ER-006 | Skin Effects of Oestrogen in Transfeminine People | 10.5281/zenodo.21211028 Version 1.1 |
Review · Note |
| Clinical review | Acne in Gender-Affirming Care | Not deposited | Read |
| Clinical review | Rosacea in Gender-Affirming Care | Not deposited | Read |
| Guide | Hair Removal in Gender-Affirming Care | Not deposited | Read |
Evidence reviews are updated when significant new evidence appears. A version number means a review has been revised, not that an earlier version was wrong. A change to any certainty grade or finding makes a new major version. Every change is logged on our corrections page.
How we keep ourselves honest
We publish under a first name only, so we can’t ask anyone to trust a name. (Corrected, 10 October 2026: this said “a pseudonym”. Eden is the author’s legal first name; her surname is not published. Why.) We publish the method instead. Every review is built to a published deposit standard, and its checks are enforced in code: if one fails, the build stops.
Standards for every piece of work
Primary sources only
No claim rests on news coverage, press releases or someone else’s summary. Figures come from a study’s results, not its abstract. Where the two disagree, the results govern.
We say how well each source was read
From ER-015 onward, each source is marked as read in full, read in abstract only, held but with limited access, or not obtained, so readers can see what each claim rests on.
Evidence kept apart from interpretation
What a study demonstrates, what its authors think it means, what we think it means, and what follows for practice are kept separate on the page.
Overlapping studies mapped
Where several figures come from the same underlying data, we say so, so that repetition is never mistaken for independent confirmation.
A watched source archive
Every source is held in an archive that is checked weekly for corrections, errata and retractions. A review cannot be built without it.
Permanent, versioned records
Every version of a review is deposited with its own DOI under a single record. Superseded versions stay available.
No silent corrections
Every correction, clarification, update or retraction is logged publicly. We keep a correction because it is faithful to the evidence, whichever way it cuts.
Free to read and reuse
Everything is published under CC BY 4.0, free to read, share and adapt, including commercially.
Controls on our own bias
We are not neutral, and we say so. Eden is a transgender woman who holds a declared view in favour of access to gender-affirming care. That view is filed as a competing interest, as is the non-profit organisation she founded, which teaches its own standards of care. The controls below exist because declaring a bias does not neutralise it. The method has to.
No third-party money
No grants, advertising, sponsorship, affiliate links or donations, including from advocacy organisations on our own side. Independence that runs only one way is not independence.
A dated register of interests
Every interest is listed with the date it began. Entries are never edited or deleted. Earlier grades we’ve given to the same evidence count as interests too.
Rules fixed before reading
The method for appraising the guidelines is registered publicly, and the Asia study is pre-registered; the Cass Review appraisal is being registered. Questions and rules are frozen before any judgement is made or any data are extracted.
The same rules in both directions
Benefits and harms are judged by identical rules. Claims that cut against our view are graded at full strength, and we have published findings unhelpful to positions we hold.
Independent checking
Each guideline appraisal is checked by an independent human verifier, who is not shown our findings, before any release. Native-language extractors read the Asia sources, with a second reader checking blind. The Cass protocol has had four rounds of commissioned adversarial review.
Randomness we can’t choose
Scoring order and samples are drawn from public randomness sources after the relevant work is fixed, so nobody can pick the convenient sample.
Report first, publish after
A defect in someone else’s work goes to the people who can fix it before we say anything publicly. Our own errors sit in the same register as everyone else’s.
AI disclosed, never a source
AI is never treated as a source, and never screens studies, extracts outcome data or assigns a grade in our reviews. How we use it is published.
There are limits, and we state them. Our evidence reviews are not systematic reviews: they are not pre-registered, dual-screened or peer reviewed, and the deposit standard sets out what they do instead. Most of our work has a single author. These safeguards reduce the risks of a one-person publication. They do not remove them.
→ How we evaluate evidence · Deposit standard · Independence · Declared interests · Corrections
Campaigns
Our campaigns are run differently from our research, and are measured by what changes rather than by what we publish.
The single patient record
Safeguards for the Health Bill’s single patient record provisions, now before the House of Lords.
→ The campaign · Lords briefing · DPRRC submission