In 1735, a boy in the Chosun Dynasty was castrated before puberty and entered the Korean royal court as a eunuch. He lived to be 109. He was not unusual. Of the 81 eunuchs whose lifespans could be verified from the Yang-Se-Gye-Bo genealogical records, three became centenarians — a rate 130 times higher than in any modern developed nation. On average, the eunuchs outlived intact men of identical social status by 14.4 to 19.1 years.
Remove testosterone and men live dramatically longer. The data are unambiguous.
Now consider the opposite: men with low testosterone die younger. The Yeap 2024 individual-patient-data meta-analysis in the Annals of Internal Medicine — 11 prospective cohorts, 24,109 men — found that total testosterone below 7.4 nmol/L predicted all-cause mortality. Min's eunuch data says remove testosterone and gain two decades. Yeap's data says lose testosterone and die sooner.
The same molecule. Opposite survival outcomes. Both backed by robust evidence.
This isn't a contradiction waiting to be resolved by better data. It's a paradox that is the data — and resolving it requires abandoning the assumption that aging is one thing.
Two Clocks, Two Verdicts
The paradox sharpened in 2025 when epigenetic clocks started delivering contradictory verdicts on the same hormone.
Sugrue et al. (PNAS 2025) built an androgen clock — a set of CpG sites whose methylation state tracks cumulative androgen exposure. In sheep, castration halted the clock. In mice, androgen receptor knockout confirmed it: the receptor, not just the hormone, drives age-associated demethylation at these sites. The earlier Sugrue eLife 2021 study identified the mechanism: 48 of the top 50 androgen-sensitive differentially methylated positions (asDMPs) show progressive hypomethylation in intact males. The top site — MKLN1 — overlaps with AR binding, DNase I hypersensitivity, H3K27ac, and vertebrate conservation. It's active in tissues with high androgen receptor expression (epidermis, kidney, skeletal muscle) and silent in tissues with low AR (cerebellum, cortex, liver, blood).
Verdict from the androgen clock: testosterone accelerates biological aging.
That same year, Kusters et al. (GeroScience 2024) analyzed three cohorts — Framingham Offspring, BLSA, and InCHIANTI, totaling 1,612 men. Higher testosterone-to-estradiol ratio was associated with decreased PhenoAge acceleration, decreased GrimAge acceleration, and decreased DNAm PAI-1. A "highly robust finding" detected by both first- and second-generation clocks.
Verdict from the mortality clocks: testosterone slows biological aging.
The Clock Disagreement Problem
Borrus, Sehgal & Higgins-Chen (Yale, bioRxiv 2024) measured variability across six epigenetic clocks and found a fundamental distinction:
Type 1 Methylation
Programmed aging and damage signals. CpG sites that change because the organism is deteriorating. Androgen-sensitive asDMPs are Type 1 — AR-driven demethylation is the aging process itself.
Type 2 Methylation
Repair and compensatory responses. CpG sites that change in response to damage — inflammaging, metabolic stress, immune decline. PhenoAge and GrimAge composite these signals.
Mixing Type 1 and Type 2 CpGs in a single clock produces contradictory intervention results. The testosterone contradiction is partly about which dimension of aging each clock measures.
Six Dimensions of Male Aging
The clocks disagree because aging isn't one process. Testosterone doesn't uniformly accelerate or protect — it pushes different dimensions of aging in different directions simultaneously. Here are the six that matter:
The six dimensions are not equally weighted. Which dimensions dominate depends on clinical context — and this is where the paradox resolves.
The Mechanistic Bridge: mTOR Connects Testosterone to the Epigenetic Clock
Dimensions 1 and 2 aren't independent. mTOR is the mechanistic bridge connecting testosterone to accelerated CpG aging.
Testosterone activates mTOR signaling. mTOR drives cellular hyperfunction — the Blagosklonny model where continued growth-program activity after development causes age-related pathology. This hyperfunction includes AR-driven demethylation at asDMPs, which is what the androgen clock measures. The chain is: T → mTOR → hyperfunction → AR-driven demethylation at asDMPs → accelerated epigenetic aging.
Leontieva and Blagosklonny (2012) showed males have constitutively higher mTOR activity (pS6, pAKT) across multiple organs, with males 28% heavier by 6 months. Rapamycin extends female lifespan more than male in both mice and Drosophila. A Nature Aging 2022 study revealed why: Drosophila enterocyte sexual identity regulates autophagy — males have high basal autophagy that rapamycin cannot further increase. The drug has less room to work in male cells because testosterone-driven mTOR has already pushed them further along the hyperfunction trajectory.
The PEARL trial (2024) provided the first human data: rapamycin improved bone mineral content in men and lean mass in women — sex-divergent responses consistent with different baseline mTOR states.
The Bridge
mTOR is not a parallel aging pathway. It is the mechanism by which testosterone converts reproductive vigor into epigenetic aging. The androgen clock doesn't measure a separate process from mTOR hyperfunction — it measures its downstream epigenetic consequence. This is why Dimensions 1 and 2 always move together.
DeepStrataAge: The Clocks Aren't Linear
If the clock disagreement were the whole story, resolution would be straightforward — just measure the right CpGs. But Lin et al. (npj Aging 2026) complicated things further with DeepStrataAge, a deep neural network trained on 29,167 samples that reveals aging is not a linear decline but a phase-structured, wave-like process:
Male Pattern
Abrupt, Compressed
Men show a sudden midlife transition — a sharp molecular shift consistent with andropause as an abrupt epigenetic event rather than a gradual decline.
Female Pattern
Earlier, Gradual
Women show an earlier but more gradual transition — phased remodeling rather than abrupt collapse. Menopause unfolds; andropause ruptures.
SHAP interpretability reveals CpG-specific timing differences between sexes. The key implication: testosterone-maintained methylation stability during young adulthood collapses at andropause. The androgen clock doesn't tick at a constant rate — it accelerates precisely when testosterone begins to fall, because the AR-maintained methylation pattern destabilizes without its hormonal input.
This means the longevity paradox has a temporal dimension. In young men, testosterone is simultaneously accelerating Dimension 1 (CpG aging) while powerfully protecting Dimensions 3 and 4 (metabolic and inflammatory aging). The net balance favors survival. At andropause, the Dimension 1 acceleration slows — but Dimensions 3 and 4 lose their protector. Which collapse matters more depends on what's already broken.
Why Evolution Built This Trap
The paradox isn't a design flaw. It's the intended design.
Colebank et al. (Science Advances 2024) provided the genomic evidence: mutations affecting reproduction are 4.9 times more likely to influence lifespan, and 7.5 times more likely to show antagonistic pleiotropy — genes that help you reproduce but cost you longevity. Testosterone is the master mediator of this tradeoff. It drives muscle mass, aggression, secondary sexual characteristics, and immune suppression (the immunocompetence handicap) — all features that enhance reproductive success in early life while eroding somatic maintenance in later life.
The classical disposable soma theory (Kirkwood, 1977) predicted this: organisms allocate finite resources between reproduction and somatic maintenance. Males, under stronger sexual selection, invest disproportionately in growth and competition. But the 2024 update from Mitchell et al. (PNAS 2024) refined the picture: reproduction itself doesn't cause lasting somatic damage in mice. The cost is not in the act of reproduction but in the hormonal infrastructure that supports it — the constitutively elevated mTOR, the AR-driven demethylation program, the inflammatory suppression that leaves gaps in immune surveillance.
Testosterone is not draining a finite repair budget. It is running a developmental program past its expiration date.
The X-Chromosome Escape Advantage
Females have a biological counterbalance males lack entirely. Dubal et al. (Science Advances, March 2025) showed that aging activates escape of genes from the "silent" inactive X chromosome (Xi) in the female mouse hippocampus. The Xi escapee Plp1 — a myelin component — increased with age in XX but not XY mice. AAV-mediated Plp1 overexpression in aging mice of both sexes improved cognition.
Females have a backup copy of every X-linked gene. As aging degrades the active X, the inactive X selectively activates neuroprotective genes. Males — with one X and a Y that's progressively lost with age — have no backup. Earlier work from Dubal's group at UCSF demonstrated that XX mice live longer than XY mice regardless of gonadal sex — most of the longevity effect comes from the chromosomes, not the gonads.
This adds a seventh dimension to the aging matrix — one that testosterone doesn't drive but that males cannot escape. The Y chromosome carries no compensatory aging-resistance genes. XX females have an architectural advantage that XY males, regardless of their testosterone levels, do not.
Context-Dependent Resolution
The paradox resolves when you stop asking "does testosterone accelerate or protect against aging?" and start asking "in this specific person, which dimensions of aging dominate?"
+14–19 years
Childhood Castration
Removes Dimensions 1 + 2 entirely (no AR-driven aging, no mTOR hyperfunction). Loses Dimensions 3 + 4 (metabolic and inflammatory protection), but in an era before metabolic disease dominated mortality, the net effect was dramatic lifespan extension. The Korean eunuchs lived in a context where infection and violence were the leading killers — and castration removed the immune suppression and risk-taking that testosterone drives.
+43% mortality
ADT in Elderly Men
Androgen deprivation therapy in prostate cancer removes Dimensions 3 + 4 in men who are already metabolically and immunologically compromised. The metabolic trap locks in. Inflammatory markers surge. Cardiovascular events spike. At this age and health status, the protective dimensions matter more than the accelerating ones.
−22.5% diabetes
TRT in Hypogonadal Middle-Aged Men
Restoring testosterone in men whose levels have fallen below the threshold (~7.4 nmol/L) reinstates Dimensions 3 + 4 protection. TRAVERSE showed cardiovascular safety (HR 0.73) and metabolic benefit. The androgen clock will tick faster — Dimension 1 reactivates — but the mortality clocks improve because metabolic and inflammatory aging are the bigger killers in this population.
U-shaped curve
The Balance Point
The cardiovascular paradox is a U-shaped mortality curve. Both extremes kill — too low loses metabolic protection, too high drives mTOR hyperfunction and erythrocytosis. The optimal zone is where Dimensions 3 + 4 are engaged without maximally driving 1 + 2. This is the evolutionary compromise, and it differs by individual.
The Evidence Void
The resolution framework above is a synthesis of converging evidence. But the direct test — does TRT change epigenetic clock age in hypogonadal men? — has never been conducted.
What Doesn't Exist
What This Means
The longevity paradox is the deepest question in testosterone biology, and it connects every thread in this series:
The metabolic trap is Dimension 3 collapsing. The immune paradox is Dimension 5's inherent duality. The cardiovascular paradox is the U-shaped curve where Dimensions 1-2 and 3-4 cross. The treatment framework is an attempt to navigate these dimensions without a clock to guide us.
Korean eunuchs gained 14-19 years because they eliminated Dimensions 1 and 2 in an era where those dimensions were the primary killers. Modern hypogonadal men die younger because Dimensions 3 and 4 are the primary killers in a world of metabolic disease and chronic inflammation. TRT in the right clinical context restores the protective dimensions — but it reactivates the accelerating ones too.
The missing piece is the human androgen clock. When it arrives, we'll be able to measure the cost of TRT in Dimension 1 and weigh it against the benefit in Dimensions 3-4. Until then, we're navigating by proxy — using mortality clocks to infer what we cannot yet directly measure, and trusting that the net survival benefit of treating hypogonadism outweighs the epigenetic cost we cannot yet quantify.
That trust is probably well-placed. But probably is not the same as proven.