Molecular brain imaging during gender‑affirming hormone therapy: a pilot synthesis of PET and MR spectroscopy
What has been measured, in whom, and what a focused paper could add.
The short answer
A focused synthesis of PET and MR spectroscopy during gender-affirming hormone therapy (GAHT) looks more promising than a PET-only review. The treatment PET evidence comes from only two Vienna cohort families and two neurochemical targets, so it cannot support a target-level meta-analysis. MRS adds different measurements and one clearly separate longitudinal cohort, from Ghent.
The contribution such a paper could make lies less in collecting studies than in counting them properly. That means separating reports from cohorts and visits from people, and separating follow-up observations from people with a valid measurement both before and after treatment. Then each molecular signal can be read for what it does and does not measure. That is the thread this report follows, and the reason for its title.
The commonest error in this literature is a unit error: adding visits as if they were people, or papers as if they were cohorts. So every load-bearing number here is underlined in the colour of what it counts. Hover over or tap any of them.
Small chips mark what kind of claim a sentence makes: an result, the interpretation, reasoning across sources, a , or an question that needs more source work. The dots show how deeply the source behind the claim was read. Citations such as Kranz 2015 open a summary card. Defined terms such as binding potential carry a soft highlight. Brief mode, in the top bar, keeps only headings, key paragraphs, figures and tables.
Five separations organise everything that follows. Each one prevents a specific, common overstatement.
Each section opens with the same four lines: what was measured, in whom, what it can mean, and what would settle it. The sections then earn those lines.
Treatment PET during gender-affirming hormone therapy has measured two proteins, the serotonin transporter and monoamine oxidase A, each in a single Vienna study family of a few dozen people. Transporter binding rose under testosterone. The monoamine oxidase results run from regional decreases to no significant change, but they are two analyses of one cohort, not a replication.
Both targets belong to the serotonin system, and both measurements count tracer held in tissue. They differ in what that count is compared with, and that sets how far each number can be read.
A positron emission tomography (PET) study starts with a radiotracer: a drug-like molecule that carries a short-lived positron emitter and is injected in trace amounts. Here the emitter is carbon-11, with a half-life of 20.4 minutes, so each dose is made on site shortly before the scan. [11C]DASB binds the serotonin transporter (SERT), the membrane protein that clears released serotonin and that SSRI antidepressants block. [11C]harmine binds monoamine oxidase A (MAO-A), the mitochondrial enzyme that breaks serotonin and other monoamines down. In both Vienna programmes each scan ran for 90 minutes on the same full-ring scanner.
Neither number is a reading of serotonin in the synapse. The authors of the SERT study suggest that testosterone raised serotonin, which then raised transporter expression . That is a mechanistic proposal, and the scan does not test it.
Five PET reports include transgender participants. Four measure a molecular target, three analyse change during treatment, and behind those three stand only two recruitment families.
| Report | Target · tracer | Number reported | Design | Transgender participants | Family |
|---|---|---|---|---|---|
| Berglund 2008 | No molecular target: blood flow · H2[15O] | Activation (perfusion) | Before treatment; odour task | 12 TW | Stockholm |
| Kranz 2014 | SERT · [11C]DASB | BPND asymmetry | Before treatment | 14 TW, reused in 2015 | Vienna SERT |
| Kranz 2015 | SERT · [11C]DASB | BPND | Baseline, 4 weeks, 4 months | 33 → 25 → 24 | Vienna SERT |
| Kranz 2021 | MAO-A · [11C]harmine | VT, 12 atlas regions | Baseline, about 4 months | 18 enrolled; 16 → 10 usable scans | Vienna MAO-A |
| Handschuh 2024 | MAO-A · [11C]harmine | VT, MRI-defined regions | Baseline, about 4.5 months | 18 at baseline; 11 paired | Vienna MAO-A |
| Registry record only | Serotonin synthesis · [11C]AMT | Planned | Planned beside harmine | No verified scans or results | Outside the tally |
| Searched, not found | Steroid receptors, SV2A, dopamine, TSPO, FDG | None | None | None verified in this pilot | Not proof of absence |
Broader reviews place these reports beside structural and functional MRI (Kranz 2020; Nielsen 2025). This section narrows the lens to the molecular targets and what their numbers can carry. The full count of people across PET and spectroscopy, including the open questions about overlap, is in section 03.
In the only longitudinal SERT study, binding rose over four months of testosterone in transgender men (TM). It fell in a few cortical regions after four months of estradiol plus antiandrogen in transgender women (TW).
The story begins with Kranz 2014, a baseline study of left-right asymmetry in SERT binding in 14 TW and cisgender controls before any treatment. Its methods section later states plainly that this group is a subsample of the 2015 study, so the two papers form one family and their participants cannot be added.
Kranz 2015 scanned 14 TM and 19 TW before treatment, with 35 cisgender controls. Subsets returned after four weeks (9 TM + 16 TW) and four months (11 TM + 13 TW). TM received 1,000 mg testosterone undecanoate every 12 weeks. TW received cyproterone acetate or triptorelin with oral or transdermal estradiol, and six also took finasteride.
In TM, binding rose in the amygdala, caudate, putamen and median raphe nucleus, and only the amygdala and caudate survived correction for the number of tests. Larger rises in testosterone went with larger rises in binding at four weeks, but not at four months. In TW, binding fell at four months in the insula and in the anterior and mid-cingulate cortex, and only the anterior cingulate survived correction. Regions lost less binding where estradiol rose more, which the authors read as a protective effect.
The MAO-A evidence comes from one registered Vienna trial (NCT02715232). Its first report found regional decreases in TM. A later report on the same PET subgroup asked about different regions and found no significant change.
Kranz 2021 calls itself a preliminary report. The trial enrolled 11 TM, 7 TW and 17 controls before the scanner was irreparably damaged in December 2019, leaving 52 usable scans. Its results table counts 10 TM and 6 TW at baseline and 6 TM and 4 TW at about four months. Distribution volume was computed with an arterial input function in 12 regions chosen in advance from the SERT study. In TM, six regions fell by 8 to 10.5%. In TW, no region changed significantly.
Two further details shape the reading. The controls' distribution volume was on average higher at their second scan, so the comparison group itself moved. And the rise in testosterone did not correlate with the change in MAO-A in any region after correction.
Handschuh 2024 returns to the same trial: the same registration, ethics number, PET subgroup sizes and scanner failure. This time the PET regions were those where grey-matter density and diffusivity had both changed during treatment, among them the fusiform and lingual gyri, insula, cingulum and cerebellum. No main or interaction effect on MAO-A distribution volume was significant. Seventeen people completed both PET scans, 11 of them transgender (6 TM and 5 TW).
One more PET study measured blood flow rather than a molecule. One planned tracer has no verified output, and several plausible targets did not turn up at all.
Berglund 2008 scanned 12 untreated TW and 24 cisgender controls with H2[15O] water while they smelled odorous steroids. Its twelve scans per person are task conditions within one session, not treatment visits, and 22 of its 24 controls came from earlier odour studies. It is a separate Stockholm programme and counts only in the all-PET total.
The MAO-A trial registration and its grant also planned [11C]AMT, a tracer for serotonin synthesis, at baseline and four months. Registry records, the grant's final report and its publication list give no transgender AMT scan count and no result. The scanner failure explains why the harmine dataset stops, but it does not establish what happened to the AMT arm. It stays an open question, outside the measured-target tally.
Finally, this pilot's searches found no transgender brain PET with direct steroid-receptor tracers, the synaptic-density marker SV2A, dopamine ligands, TSPO for neuroinflammation, or FDG. The searches were bounded, and a database hit count is not an eligibility count, so this is a search finding, not proof that no such study exists.
Even where binding changed, five features of these designs stop the change from being assigned cleanly to testosterone or to estradiol.
Spectroscopy has followed transgender men through testosterone in two places. Vienna's data come from the same trial as the newer PET; Ghent is a separately recruited cohort. Both report ratios of tissue metabolite pools, not neurotransmitter release.
Proton magnetic resonance spectroscopy (MRS) reads the chemistry of one block of tissue at a time. Hydrogen nuclei in each molecule resonate at their own characteristic frequency, so the result is a row of peaks rather than a picture.
The block is the voxel, usually a few cubic centimetres. Water dominates the raw signal and is suppressed, leaving metabolites present at millimolar levels. Position along the axis is chemical shift, given in parts per million (ppm). It is a property of the molecule rather than of the scanner, so N-acetylaspartate (NAA) appears at 2.01 ppm at any field strength. Peak area scales with how much of a molecule the voxel holds, summed over every cell type and compartment inside it.
Every number in this section describes a pool of molecules in a volume of tissue. Four properties of those pools limit what a change in them can mean.
Ghent used an unedited sequence that captures the large peaks. Vienna and the Bangkok study used editing, which isolates GABA at the cost of almost everything else.
The scales differ too. Vienna acquired a grid of spectra across a large slab, nominal voxel about 4 cm³, and averaged them within regions segmented from each participant's own scan. Ghent placed two single boxes of 6 and 12 cm³. Their hippocampal values sample tissue differently.
Vienna's spectroscopy found one treatment effect, in hippocampal GABA+/tCr. It comes from the same registered trial as the newer harmine PET.
Spurny-Dworak 2022 scanned 15 TM and 15 cisgender women (CW) twice, at least 12 weeks apart. The actual interval was 146 ± 30 days for TM and 166 ± 40 days for CW. The TM were on seven different regimens: intramuscular or transdermal testosterone, in some cases with desogestrel or triptorelin.
In the hippocampus, GABA+/tCr fell in TM relative to CW across visits: a group-by-time interaction with p = .048 after Bonferroni correction across five regions, two ratios and two groups. There was no Glx/tCr effect in any region, no GABA+ effect in insula, putamen, pallidum or thalamus, and no correlation between hormone levels and either ratio.
The report places itself within NCT02715232, the registration behind Kranz 2021 and Handschuh 2024. It therefore widens what that family measured, from MAO-A to GABA+ and Glx, rather than adding a verified independent cohort. Whether these 15 TM are the people in the PET reports cannot be read from the papers.
The authors read the result through hippocampal androgen metabolism and neurotrophic effects on interneurons, mechanisms drawn from animal work. The measurement cannot separate GABA from co-edited macromolecules, which they assume stable.
Ghent is the only cohort here with molecular imaging before and during treatment that was recruited outside Vienna. Its treatment signal is a parietal decrease in mI+Gly/Cr; the amygdala voxel showed no simple change within TM.
Collet 2023 followed TM starting testosterone undecanoate, 1000 mg every 12 weeks, alongside cisgender men (CM) and CW with no intervention. Of 31 TM at baseline, 28 were retained at follow-up after 7.7 ± 3.5 months; the planned interval was six months, extended by COVID-19 delays. At follow-up, spectra of acceptable quality came from 25 TM in the parietal voxel and 20–22 in the amygdala voxel, depending on metabolite.
Within TM, parietal mI+Gly/Cr decreased (p = .015). Visit-by-group interactions appeared for parietal mI+Gly, mI+Gly/Cr and NAA/mI+Gly relative to CM, and for amygdala Glx/Cr relative to CW. In the amygdala/anterior hippocampus there was no within-TM change, although after treatment TM had lower Glx and Glx/Cr than CW.
Collet 2021 is this cohort's baseline, and Kiyar 2022 combines fMRI with the same MRS data; neither adds people. Collet 2024 is a corrigendum that restores the ethics, funding and acknowledgement statements and changes no results. For counting: one family, one report on change during treatment.
One result needs reconciling before Ghent can enter any pooled estimate. The reported within-TM change in parietal mI+Gly/Cr is −0.045 (95% CI −0.076 to −0.015). The estimated means in the next sentence of the paper, 0.948 before and 0.944 during treatment, differ by only 0.004, which matches the paper's own later description of a 0.4% decrease. The supplementary medians, 0.954 and 0.908, imply a shift close to the coefficient.
Ask the authors which quantity the coefficient describes and on what scale, and how many TM had valid parietal spectra at both visits. The supplement's follow-up table is headed 31 TM, but its valid and missing counts sum to 28.
The only spectroscopy report on TW found here is a single session. It shows that edited GABA measurement in TW is feasible, not how treatment changes it.
Jarukasemkit 2025 scanned 20 TW recruited for depressive symptoms, using MEGA-PRESS in the left hippocampus and dorsolateral prefrontal cortex. Lower hippocampal GABA/Cr went with higher depression scores until the participant with the highest GABA value was removed. About half were on GAHT (the methods say half, the results 45%); on- and off-treatment comparisons were not significant.
| Report | Cohort | Acquisition | Regions | Signals | Valid transgender spectra |
|---|---|---|---|---|---|
| Spurny-Dworak 2022 | Vienna, NCT02715232 | 3D MEGA-LASER MRSI, 3 T, TE 68 ms | Hippocampus, insula, putamen, pallidum, thalamus | GABA+/tCr, Glx/tCr | TM 12–15 per region, ratio and visit |
| Collet 2021 | Ghent, baseline | PRESS, 3 T, TE 75 ms | Left amygdala/anterior hippocampus; right lateral parietal | NAA, Cho, Cr, Glx, mI+Gly; ratios | TM 23 and 24, one visit |
| Kiyar 2022 | Ghent, same MRS data | As Collet 2021, with fMRI | Amygdala/anterior hippocampus | Choline as moderator | TM 25, one visit |
| Collet 2023 | Ghent, follow-up | PRESS, 3 T, TE 75 ms | As Collet 2021 | As Collet 2021 | TM 25 parietal, 20–22 amygdala at follow-up |
| Jarukasemkit 2025 | Bangkok | MEGA-PRESS, 3 T | Left hippocampus, left DLPFC | GABA/Cr | TW 20, one visit |
MRS widens a molecular synthesis in two ways: one more longitudinal family, and chemical signals that PET does not measure. It does not supply a larger paired sample.
Where spectroscopy might meet PET, three directions for a protocol:
Section 03 places both spectroscopy families on the full cohort map, next to the PET families, and counts each participant once.
GABA gives three groups of peaks, near 3.01, 2.28 and 1.89 ppm, and all three sit under larger signals: creatine, Glx and NAA. The 3.01 and 1.89 ppm groups belong to neighbouring hydrogens in the same molecule, so they are coupled. A narrow pulse at 1.9 ppm changes how the 3.01 ppm signal evolves, but only in molecules with that coupling.
Recording spectra with and without that pulse and subtracting them removes the uncoupled creatine peak and keeps GABA. Macromolecules with a resonance near 1.7 ppm are partly hit by the same pulse and contribute to the 3.0 ppm difference peak, hence the name GABA+. Glutamate and glutamine are co-edited and appear near 3.75 ppm. Separating GABA from macromolecules needs additional acquisitions, and Spurny-Dworak 2022 point to 7 T for separating glutamate from glutamine.
Ten PET and spectroscopy reports trace back to five cohort families, and only three of those families measured the same people before and during treatment. Inside them, the number a change estimate rests on, people with a valid measurement at both times, is smaller than any headline sample size and is sometimes not reported at all.
A sample size means little until its unit is named. This literature mixes five units, and most overstatements come from adding one as if it were another.
The units nest, and each step narrows. A study enrols people, scans them at baseline, rescans a subset, loses some scans or spectra to quality control, and ends with a paired sample for each endpoint.
Placed by what they measured and whom they recruited, the ten PET and spectroscopy reports fall into five families. The newer Vienna trial alone supplies three of them, plus two of the MRI reports used for context.
Read down a column and the repetition shows. NCT02715232 is behind Kranz 2021, Handschuh 2024 and Spurny-Dworak 2022, and also behind Reed 2023; Konadu 2023 pools it with an earlier Vienna MRI trial, NCT01292785. The Ghent column holds three reports and one recruitment: Collet 2021 describes the baseline, Kiyar 2022 states that its spectroscopy data were published there before, and Collet 2023 follows the same project through testosterone treatment.
Read across a row and the thinness shows. Each PET target has exactly one longitudinal family: the serotonin transporter (SERT) in the earlier Vienna PET trial and monoamine oxidase A (MAO-A) in the newer one. GABA-edited spectroscopy has one longitudinal family (Vienna) and one single-session cohort (Bangkok), and the PRESS metabolite ratios have one family (Ghent). The Stockholm perfusion study and the Bangkok study measured each person in a single session, so neither can estimate change during treatment.
The same arithmetic, animated on the three PET treatment reports: papers collapse into families, scans into people, and people narrow to pairs.
Those three moves are the method of this section, and every number below carries its unit so that a reader can make them too.
For each longitudinal molecular report, the ledger keeps enrolled, baseline-valid, follow-up-valid and explicitly paired counts in separate columns. Where a paper does not report a number, the cell stays open.
The column that matters most for a change estimate is the last, and few rows fill it from an explicit statement. Handschuh 2024 states that 17 participants completed both PET scans and breaks them down by group. In Kranz 2015 everyone rescanned had a baseline scan by design. Elsewhere the paired count has to be bounded by arithmetic, which narrows it but does not settle it.
| Report · endpoint | Family | Started | Valid at baseline | Valid at follow-up | Explicitly paired |
|---|---|---|---|---|---|
| Kranz 2015 SERT · 4 weeks | Vienna SERT | 3314 TM + 19 TW | 33 | 259 TM + 16 TW | 25by design |
| Kranz 2015 SERT · 4 months | Vienna SERT | same 33 | same 33 | 2411 TM + 13 TW | 24by design |
| Kranz 2015 · people with any follow-up scan | The two follow-up visits overlap; they are not added. | not reported27–32 by arithmetic | |||
| Kranz 2021 MAO-A · ~4 months | Vienna multimodal | 1811 TM + 7 TW | 1610 TM + 6 TW | 106 TM + 4 TW | not reported8–10 by arithmetic |
| Handschuh 2024 MAO-A · MRI-defined regions | Vienna multimodal | 18same PET subgroup | 1811 TM + 7 TW | 116 TM + 5 TW | 11stated; not added to 2021 |
| Spurny-Dworak 2022 hippocampal GABA+/tCr | Vienna multimodal | 15TM | 13 | 13 | not reported11–13 by arithmetic |
| Collet 2023 parietal metabolite ratios | Ghent | 31TM; 28 retained | not reported in this report | 25of 28 retained | not reportedat most 25 |
| Collet 2023 amygdala / anterior hippocampus | Ghent | same 31 | not reported in this report | 20–22by metabolite | not reportedat most 20–22 |
One report can also carry several paired samples: valid values in Spurny-Dworak 2022 range from 12 to 15 per visit across regions and ratios. The reports' mixed models can use unpaired scans, so each analysis is sound on its own terms; a synthesis of within-person change still needs the pairs.
Each trap is a sum that looks natural and is wrong. All six come from the published reports themselves. None needs an error by the authors, only a reader who adds across units.
None of these gaps needs new data. Each could be closed with information the research groups already hold, and each ask is specific enough to answer in a table.
Four MRI analyses help interpret the molecular signals. None adds a molecular study, and three of them reuse the Vienna multimodal trial already counted.
Handschuh 2024 is the closest bridge: grey-matter and microstructure change measured in the same people whose MAO-A was scanned. Reed 2023 analyses insular connectivity in 14 TM and 11 TW scanned twice in the newer Vienna trial. Konadu 2023 pools that trial with an earlier Vienna MRI trial to study hypothalamic volume in 38 TM and 15 TW scanned twice. Kranz 2018 reports hypothalamic diffusion in 25 TM over the same three-visit schedule as the DASB study; its registration and overlap were read only at abstract level.
Their samples are larger than the molecular ones, which makes them tempting to add. They should not be added. A Konadu participant from NCT02715232 may also be a harmine, spectroscopy or connectivity participant, and the report does not list which trial each person came from. The right use of these studies is mechanistic: what changed in tissue alongside a molecular signal, in people already counted.
The closest reviews already read several of the same primary reports, each under its own question. A focused synthesis of PET and MR spectroscopy during hormone therapy, with selected MRI as context, could complement them by counting reports, cohorts and valid pairs once, and by reading each measure for what it measures. How much it adds depends on the current study table of the scoping review with Michael Winterdahl, which we have not yet seen.
The closest reviews are not rivals to a new paper. They ask different questions of overlapping material, and each does what it set out to do.
Positron emission tomography (PET) and magnetic resonance spectroscopy (MRS) during gender-affirming hormone therapy (GAHT) have been touched by three reviews that matter here. Two were read in full; the third, the most recent and the closest in scope, is available to us only as a congress abstract.
Asks: what longitudinal MRI and PET studies show about the brain during GAHT.
Molecular content: 1 PET treatment report, the 2015 serotonin transporter (SERT) study with [11C]DASB, described across baseline, one month and four months.
Written before the monoamine oxidase A (MAO-A) and MRS treatment reports appeared.
Asks: how brains differ by gender identity or sexual orientation before treatment; post-treatment data are excluded by design.
Molecular content: 3 PET reports (Berglund 2008, Kranz 2014 and the baseline of Kranz 2015) and 1 single-photon emission computed tomography (SPECT) report, too few to meta-analyse.
Search to January 2018 for gender identity; a Medline update to March 2021.
Asks: how the brain changes during GAHT across imaging methods, and how future studies could be designed better.
Content: 52 included studies from 212 records found in PubMed and Embase.
The accessible abstract has no study table, search cutoff, PET-target count or cohort map. The current manuscript was not available to us.
Read side by side, they show that overlap belongs to a question, not to a paper. The 2015 DASB study appears in Kranz 2020 as a treatment study and in Frigerio 2021 through its baseline comparisons alone. Whether a new paper overlaps an existing review therefore has to be judged report by report and question by question.
Of 13 primary reports, Kranz 2020 reviews two and Frigerio 2021 includes four, two of them only through baseline data. Everything published after 2020 is new to both, and the scoping review's coverage stays unknown until its study table can be compared.
Two things stand out. The MAO-A reports, both longitudinal MRS cohorts and the newer Vienna MRI analyses all postdate the two full-text reviews, so a molecular update has real material to add. And the one review that could already cover that material is the one whose table we have not seen. Any statement about what a new paper adds should wait for that comparison, and no paper here should claim to be the first review of anything.
A focused PET + MRS synthesis is the working preference. A PET-only paper is possible but thin, and an all-modality catalogue would mostly re-cover ground that existing reviews, including the scoping review (Nielsen 2025), already map.
The preferred design is a critical review with a reproducible search, or a scoping review if the team prefers a formal map. Its core is change during treatment: three PET reports from two Vienna families and two longitudinal MRS reports, one of them from the separate Ghent cohort. Baseline group contrasts sit in their own background section: Berglund 2008, Kranz 2014, Collet 2021, Kiyar 2022 and the cross-sectional Bangkok study in transgender women (TW), Jarukasemkit 2025. That way a difference before treatment is never read as an effect of treatment.
There would be no meta-analysis per target. Two families and two PET targets leave nothing independent to pool, and the two harmine analyses ask different regional questions of one study family, so their different results are not a replication test. The contribution would be the accounting itself: baseline versus treatment comparisons, cohort reuse, valid paired endpoints, and what binding potential, distribution volume and metabolite ratios each measure.
Four MRI analyses earn a place as context. Three reuse the newer Vienna trial already counted; the fourth, Kranz 2018, comes from the older Vienna programme, and its overlap with the DASB cohort is unresolved. None is counted as more molecular evidence or more people.
Used this way, MRI helps interpret a molecular result (did tissue change where binding did not?) without being counted as more molecular evidence or more independent people.
None of these needs new data. Most need a careful re-reading or a short, collegial question to the groups that ran the studies.
| # | Open item | Why it matters | Possible route | Owner |
|---|---|---|---|---|
| 1 | Compare against the scoping review's current study table | Settles overlap and any statement of what a new paper adds | Ask for the current manuscript version or its table | to agree |
| 2 | Reconcile harmine sample allocations and exact valid pairs | The 2021 and 2024 papers divide 52 usable scans differently; 10 four-month observations are not 10 verified pairs | Re-extract both tables; possibly ask the Vienna authors | to agree |
| 3 | Settle the planned [11C]AMT arm | Registered and funded, with no verified transgender output | Ask the Vienna group whether AMT scans were acquired | to agree |
| 4 | Pair counts per MRS endpoint | 15 Vienna and 28 Ghent TM were rescanned; valid pairs are fewer and differ by region | Supplements first, then an author query | to agree |
| 5 | Full screen of the retrievals, if a systematic review is wanted | 573 records are retrievals, not eligible studies | Two-reviewer screen; add Embase, Scopus, Web of Science and registries | to agree |
| 6 | Separate adult GAHT, puberty suppression and regimens | Exposures differ; estradiol plus antiandrogen is not estradiol alone | Write eligibility rules before extraction | to agree |
| 7 | Targets not found here: steroid receptors, SV2A, dopamine, neuroinflammation | A qualified search finding, not proof of absence | A targeted search before any gap is stated | to agree |
The third item deserves care in both directions. The trial registration NCT02715232 and the Austrian Science Fund programme FWF KLI504 planned serotonin-synthesis PET with [11C]AMT (α-methyl-L-tryptophan) alongside harmine, before and after about four months of treatment, and the EU register entry (EudraCT number 2015-000502-19) marks the trial completed. No transgender AMT scan count or result was found. The harmine papers report that the PET scanner was irreparably damaged in December 2019. That explains the truncated harmine data; it does not show that every planned AMT scan was cancelled.
The seventh item is phrased narrowly on purpose. This pilot did not find GAHT studies of steroid receptors, synaptic vesicle glycoprotein 2A (SV2A), dopamine or neuroinflammation targets, but its search was bounded. That is a reason to search for them deliberately, not a gap to announce.
Each line is a sentence that would be easy to write and wrong to publish. Together they are the spine of any paper built from this pilot.
Back matter
Each card states the design, who was measured, what the source showed, its main limitation, and how deeply it was read. Filter by kind, or search.
The same definitions that appear on hover, gathered in one place.
This is an exploratory pilot. It is not a completed systematic review, and its counts are source-checked candidates rather than a final eligibility denominator.
Reproducible PubMed and Europe PMC queries were run for PET, MR spectroscopy, longitudinal MRI and existing reviews, with a modality-blind rescue search and citation chasing. Together they returned 573 de-duplicated records. These are retrievals and have not been fully screened. Every primary report cited here was read in full, including methods, tables and supplements where available, unless its card says otherwise. Sample counts were extracted separately for enrolment, valid baseline, valid follow-up and explicit pairs. Cohort families were assigned from trial registrations, ethics numbers, funding and authors' own statements of reuse; participant-level identity across programmes has not been established. Registry and grant records were consulted for planned tracer arms. Unresolved values are left unresolved.
The published version of the existing scoping review (Nielsen 2025) is a congress abstract; its study table was not available, so overlap is assessed against the abstract and two earlier full-text reviews (Kranz 2020, Frigerio 2021).