Testosterone and glioblastoma: what the Cleveland Clinic 38% signal actually means.

What Lee et al. actually reported

On May 6, 2026, a Cleveland Clinic team led by Juyeun Lee and senior author Justin Lathia published "Androgen loss accelerates brain tumour growth via HPA axis activation" in Nature [Lee 2026]. The paper has two arms: a retrospective clinical signal and a mechanistic mouse study. Both arms are interesting on their own. Together they are more interesting still.

The clinical arm pulled records on more than 1,300 men diagnosed with glioblastoma (GBM) — a Grade IV astrocytoma and the most aggressive primary brain tumour in adults — and stratified by whether the patient was receiving exogenous testosterone supplementation for unrelated indications (typically age-related hypogonadism) at the time of diagnosis. Men receiving supplemental testosterone showed a 38% lower mortality risk compared with men not on TRT [NIH 2026]. The Cleveland Clinic press materials and downstream coverage all agree on the 38% figure; the precise hazard ratio and 95% confidence interval live in the published paper. Researchers explicitly framed the clinical association as observational and not establishing causality [Inside Precision Medicine 2026].

For context: glioblastoma median survival, on the standard Stupp protocol of maximal safe resection plus concurrent radiotherapy and temozolomide, is roughly 14–16 months from diagnosis [Stupp 2005]. Two-year survival hovers near 27%. A 38% mortality reduction, if it survives the move from association to replication, is not a small signal in a disease that has barely moved in twenty years.

That "if" is doing a lot of work. The point of this article is to explain what the signal is, why it is interesting, and what it does and does not justify — without pretending the biology is uninteresting and without pretending the clinical data is more than it is.

The HPA-axis mechanism, in plain English

The mechanistic arm of the paper is the part that flips the cultural framing. The dominant story for decades — particularly in oncology culture and in patient-education materials around TRT — has been that androgens fuel growth in male tumours. That framing comes from prostate-cancer biology, where androgen-deprivation therapy is a cornerstone of treatment. The Cleveland Clinic group set out to test whether the same logic transfers to a tumour in a fundamentally different anatomical compartment: the brain.

It does not. In their mouse models of glioblastoma, surgical castration — which acutely removes androgen — did not slow tumour growth. It accelerated it. The signal it pulls runs through the hypothalamic-pituitary-adrenal axis (HPA axis), the neuroendocrine circuit (hypothalamus → pituitary → adrenal cortex) that governs cortisol output and the stress response.

Androgen loss in their male mice triggered HPA-axis hyperactivation, which drove a sustained surge in circulating stress hormones — corticosterone in mice, the rodent equivalent of human cortisol [Lee 2026]. The downstream effect of that surge was an immunosuppressive brain environment: tighter blood-brain barrier segregation, fewer infiltrating immune cells, less anti-tumour T-cell activity at the tumour edge. The tumour grew faster because the body's immune surveillance, particularly inside the brain compartment, was effectively sedated by chronic glucocorticoid exposure.

Lathia's framing of why this matters anatomically captures it well: "The brain has evolved to keep stuff out and that includes immune cells from elsewhere in the body" [Medical Xpress 2026]. The brain is an immune-privileged compartment. The cortisol-driven tightening of that compartment is the punch line. The same sex-specific finding ran through the female-mouse arm: androgen loss in female mice did not produce equivalent HPA hyperactivation, and did not accelerate the same tumours [News Medical 2026]. This is endocrine-immune crosstalk, sex-specific, in a tumour compartment that doesn't behave like prostate.

Androgen loss accelerated tumour growth — not by removing a growth signal but by unmasking a chronic cortisol one. The story is the immune-suppressive cost of low testosterone in the brain compartment.

The prior literature this builds on

The 2026 paper did not arrive in a vacuum. Fariña-Jerónimo and colleagues published a small prospective study in 2024 in European Journal of Medical Research that — at first glance — seems to point the opposite way [Fariña-Jerónimo 2024]. In 40 patients with primary glioblastoma (16 women, 24 men), they measured circulating testosterone at diagnosis and tracked progression-free survival. Patients with androgen deficiency showed better progression-free survival than those with normal testosterone levels (252 vs. 135 days; p = 0.041). Multivariate adjustment for age, performance status, extent of resection, and MGMT methylation status returned a hazard ratio of 6.346 (95% CI 1.812–22.223; p = 0.004) for normal androgen status as a risk factor for progression.

These two readouts are not as contradictory as the headlines make them look. Fariña-Jerónimo measured endogenous testosterone in a small mixed-sex cohort at diagnosis. Lee et al. measured outcomes in a much larger male-only cohort stratified by whether the patient was receiving exogenous supplementation. Endogenous androgen status in a sick patient and exogenous androgen supplementation in a patient previously hypogonadal-for-other-reasons are not the same biological condition. A patient with low endogenous testosterone at diagnosis may have low testosterone because they are sick — reverse causation. A patient on stable TRT for two decades whose hormone levels are clinically normal when GBM arrives is a different cohort entirely.

The mouse arm of Lee et al. resolves the apparent conflict the way good mechanistic work tends to: by showing that acute androgen loss — the experimental condition closest to "what happens when you take testosterone away from a previously normal male" — is what triggers the HPA-axis surge. The story is not that testosterone is good for tumours. The story is that chronic loss of testosterone in previously testosterone-replete males sets up a stress-axis cascade that suppresses brain-compartment immunity.

That is a narrower, more careful claim than the headline-friendly "testosterone protects against brain cancer." It is also the claim the biology supports.

Why this isn't a license to start TRT for cancer prevention

We need to be direct here. A 38% mortality-reduction signal in a 1,300-man retrospective cohort, paired with a clean mouse mechanism, is the kind of finding that gets misread two ways. The first misread: "testosterone prevents brain cancer, everyone should be on TRT." The second misread: "this is just an observational study, ignore it." Both are wrong.

On the first misread: glioblastoma is rare. Annual US incidence sits near 3 per 100,000 adults. Even if the 38% signal is causally real, starting exogenous testosterone in a man who does not have biochemical hypogonadism, in order to lower the lifetime risk of an uncommon tumour, is a calculation that fails on absolute-risk math alone. TRT is not consequence-free. It changes erythropoiesis, lipid handling, endogenous gonadotropin signalling, and fertility. It commits the user to clinical monitoring indefinitely. The right question for any TRT-curious man remains the same one it was last month: am I biochemically hypogonadal, do I have symptoms, and is the clinical risk-benefit conversation honest?

On the second misread: dismissing a 38% mortality signal with a plausible biological mechanism because the human data is observational is also wrong. Observational data backed by mechanistic data is the signal that gets formally tested. The path now is a registered prospective trial — testosterone supplementation in hypogonadal men diagnosed with GBM, or even glucocorticoid-modulating co-therapies in the standard-of-care arm — and the Cleveland Clinic press materials hint that follow-on work is in design.

The bigger therapeutic insight from this paper isn't testosterone. It's the HPA axis. If chronic cortisol is the actual lever that suppresses anti-tumour immunity inside the brain, then any intervention that interrupts that cortisol surge — selective glucocorticoid receptor modulators, behavioural stress reduction in the post-diagnosis period, even careful management of the dexamethasone protocols routinely used to control GBM-associated cerebral oedema — becomes a target worth examining. Dexamethasone is not testosterone, but it is a potent glucocorticoid, and Lee et al. will force a re-examination of the trade-off between controlling oedema and chronically suppressing the immune response oncologists increasingly want to harness.

Where TRT-and-cancer-risk actually stands in 2026

The most important recent context for the Cleveland Clinic finding is the TRAVERSE trial — the largest randomised, placebo-controlled trial of testosterone safety in middle-aged-to-older hypogonadal men run to date. TRAVERSE enrolled 5,246 men aged 45 to 80 with biochemical hypogonadism and either pre-existing cardiovascular disease or elevated cardiovascular risk [Lincoff 2023 TRAVERSE]. The primary cardiovascular readout, published in 2023, found testosterone non-inferior to placebo for major adverse cardiac events.

Prostate safety was a pre-specified secondary endpoint. The published prostate-events analysis reported 5 cases of high-grade prostate cancer in the testosterone arm vs. 3 in the placebo arm — incidences described as low and similar across the two groups [Bhasin 2023 TRAVERSE prostate]. Rates of acute urinary retention and invasive surgery for benign prostatic hyperplasia were also similar. The honest 2026 read on TRT-and-prostate-cancer risk: in men who have been appropriately screened to exclude pre-existing high-risk prostate findings, testosterone replacement at standard therapeutic dosing does not appear to materially increase clinically-meaningful prostate cancer risk over multi-year follow-up.

The Lee paper adds a different axis to that conversation. The prostate-cancer question is the one that historically loomed largest for TRT users. Brain tumours were not previously part of the risk-benefit conversation at all, because there was no human data either way. Now there is a 1,300-man signal pointing toward protection, not harm. That doesn't change the prostate screening protocol. It does change the framing that men on TRT have been given for years — that they are taking on cancer risk in exchange for symptomatic and metabolic benefit. The picture is more complicated than that, in both directions, and the honest conversation is becoming harder for any clinician to caricature.

Where the screening conversation belongs

The Cleveland Clinic data does not collapse the case for cancer screening in men on testosterone — it does the opposite. If testosterone is now plausibly active in the cancer-immune conversation in more than one tissue compartment, the case for coherent baseline and annual screening gets stronger, not weaker.

For men on or considering TRT, the screening framework remains:

• Baseline PSA and digital rectal exam before initiating, with thresholds for urology referral pre-defined.
• Annual PSA monitoring, with attention to velocity (rate of change) rather than only absolute level.
• Haematocrit every 3–6 months in the first year, then annually — testosterone reliably raises haematocrit, and erythrocytosis is a manageable but real risk.
• Lipid panel and fasting glucose annually, since TRT alters metabolic parameters in ways that are clinically meaningful over time.
• Symptomatic surveillance for new neurological symptoms, headache patterns, or any change that warrants imaging in a man over 50. This was true before the Lee paper and remains true after.

What the Lee paper does not do is justify lowering the threshold for brain imaging in asymptomatic TRT users. The absolute incidence of glioblastoma in any given year is too low — even a 38% relative-risk modulation does not move screening math for an asymptomatic patient. Any neurological symptom worth investigating gets investigated; the prevention story stays a clinical-trial question for now.

A tiered framework for TRT users with cancer-family-history

We do not write protocols. We write frameworks that you take to a clinician. With that established: