Testosterone Therapy

The cardiovascular fear that defined twenty years of testosterone prescribing has collapsed under one large trial — but the evidence that remains is narrower than the clinics selling "optimization" suggest: testosterone reliably treats genuine, symptomatic deficiency, yet it is not a longevity drug or a tune-up for healthy men.

Testosterone replacement therapy (TRT) has lived under a heavier regulatory cloud than almost any drug in routine internal medicine — a 2015 FDA boxed cardiovascular warning, retrospective signals of acute harm, and decades of observational data linking low endogenous testosterone to early death. The 2023–2026 evidence has reorganized the picture. The TRAVERSE trial settled the cardiovascular question prospectively. A 737-man treatment registry reported an extraordinary — but, on inspection, non-credible — mortality reduction in the men with the most to lose. A genetic study using lifelong-exposure data flagged a subtler hypertensive risk that clinical trials are too short to see. None of these alone tells the whole story; together they sketch a calibrated answer to who actually benefits.

What testosterone does, and what "low" means

In adult men testosterone and its metabolites (dihydrotestosterone, estradiol) drive muscle protein synthesis, suppress visceral adiposity, maintain bone mineral density, stimulate erythropoiesis, and support libido, executive cognition, and mood.

Endogenous production falls with age. In the Massachusetts Male Aging Study — a population-based cohort of 1,709 men aged 40–70 at baseline, re-measured 7–10 years later — total testosterone declined 1.6% per year within individuals, with bioavailable testosterone falling 2–3% per year.^[1] The Baltimore Longitudinal Study of Aging, following 890 men, quantified how often that decline crosses into deficiency: hypogonadal total-testosterone values were present in about 20% of men over 60, 30% over 70, and 50% over 80.^[2]

Two thresholds are in active use:

AUA: total testosterone <300 ng/dL (10.4 nmol/L) on two separate morning draws plus symptoms.^[3]
Endocrine Society: <264 ng/dL (9.2 nmol/L) — slightly stricter, more conservative, and anchored to CDC-harmonized assay reference ranges.^[4]

Symptoms of late-onset hypogonadism — fatigue, low libido, poor recovery, mood flatness, lost lean mass — are nonspecific and overlap with poor sleep, alcohol, untreated obstructive sleep apnea (OSA), depression, overtraining, and several common medications. Both societies require lifestyle correction first before pharmacological replacement.

Lifestyle first — endogenous rescue (Strong)

Before any prescription, the modern guideline path is correction of the inputs that suppress endogenous production:

Sleep. In a controlled laboratory study, restricting 10 healthy young men to 5 hours in bed for one week lowered daytime testosterone by 10–15%, with the largest drop in the afternoon and evening.^[5] Untreated OSA is the single largest reversible suppressor — and it causes the same fatigue/low-libido phenotype TRT is often prescribed for. Screen for OSA before pursuing TRT (insomnia-treatment, sleep-breathing).
Exercise. 150–300 minutes/week of moderate aerobic activity combined with resistance training reduces visceral fat and improves sexual function; the Endocrine Society and AUA both put weight loss and lifestyle correction ahead of pharmacological replacement in obese and metabolically unwell men.^[6] See resistance-training.
Stress and alcohol. Chronic hypercortisolemia suppresses the hypothalamic-pituitary-gonadal (HPG) axis, and heavy alcohol use compounds the suppression. In the Massachusetts Male Aging Study, men in apparent good health — no chronic illness, no prescription medication, no obesity, no excessive drinking — carried androgen levels 10–15% higher than the cohort as a whole at the same age.^[7]
Endocrine-disrupting chemicals. Phthalates, bisphenols, and parabens depress sex hormone-binding globulin (SHBG) and disrupt androgen signalling; the IRECO trial showed urinary biomarkers drop within five days of personal-care substitution. See environmental-toxins.
Varicocele repair (where present) durably raises mean total testosterone — a structural rather than pharmacological fix.

A meaningful fraction of men presenting as "low T" recover into the normal range after these inputs are corrected. About 25% of men start TRT without comprehensive baseline lab evaluation — a reckless starting point given how many of these substitutes for diagnosis.^[8]

Cardiovascular safety — what TRAVERSE settled (Strong)

The defining trial of the field is TRAVERSE — an FDA-mandated, event-driven, double-blind RCT of 5,204 men aged 45–80 with documented hypogonadism (two morning T <300 ng/dL plus symptoms) plus either established cardiovascular disease or elevated CV risk, randomized to daily transdermal testosterone gel vs placebo for a mean 21.7 months of treatment with 33 months follow-up.^[9]

The primary composite of cardiovascular (CV) death, non-fatal myocardial infarction (MI), and non-fatal stroke occurred in 7.0% of the testosterone arm vs 7.3% of placebo — formally non-inferior. There was no excess of prostate cancer or worsening lower urinary tract symptoms (LUTS).^[10] In 2025 the FDA removed the boxed cardiovascular warning from all testosterone product labels on the strength of TRAVERSE — and in the same action added a class-wide blood-pressure warning after required post-marketing ambulatory blood-pressure studies confirmed a rise across every product.^[11]

The trial's pre-specified diabetes substudy is often misread. In the men with prediabetes, testosterone did not significantly slow progression to diabetes — the rates tracked closely and non-significantly across four years (for example 10.1% vs 14.6% at 24 months; overall P = 0.49), with no improvement in glucose or HbA1c and no diabetes remission in those who already had it.^[12] The authors concluded plainly that testosterone alone should not be used to prevent or treat diabetes. The dosing was also conservative — about 25% of treated men never reached the normal range, with mean testosterone rising only to about 9.3–12.9 nmol/L (roughly 270–370 ng/dL) — so the trial cannot fully rule out a glycemic effect at higher doses, but on its hard endpoint the metabolic case for testosterone is not supported.^[13]

Secondary signals to monitor

TRAVERSE was not unambiguously clean. The testosterone arm showed higher rates of:

Atrial fibrillation — a real arrhythmia signal, important in older men screening for AF.
Acute kidney injury and pulmonary embolism — modest absolute increases, more relevant in patients with baseline risk factors.
Elevated blood pressure — confirmed on ambulatory monitoring; the FDA added a label warning for some testosterone products.

These belong on the consent conversation, not in the headline.

The Mendelian randomization paradox (Moderate)

The opposite-direction finding came from a 2026 Mendelian randomization (MR = using inherited genetic variants as lifelong-exposure proxies to test causality) study. Genetic instruments derived from more than 425,000 UK Biobank adults of European ancestry were cross-referenced with coronary artery disease (CAD) outcomes from over 1.16 million people in CARDIoGRAMplusC4D.^[14] Men with a genetic predisposition to higher lifelong testosterone had a 17% higher CAD risk — an odds ratio of 1.17 (95% CI 1.07–1.27; P = 3.3 × 10⁻⁴), meaning roughly 1.17 times the odds — equivalent in absolute terms to a lifetime CAD risk rising from ~7.3% to ~8.5%. (The confidence interval is the range where the true effect most plausibly lies; because this one sits entirely above 1.0, the association is unlikely to be neutral or protective — whereas a range crossing 1.0 would be compatible with no effect.) The signal was null in women (OR 1.01, 95% CI 0.94–1.10).^[15] Adjusting for systolic and diastolic blood pressure substantially weakened the association — most of the genetic signal appears to be lifelong hypertensive shear stress, not lipid or inflammatory mechanisms. The broader lesson is that genetic instruments for lifelong testosterone do not reproduce a cardiovascular benefit — the causal signal on hard endpoints runs the wrong way for the "more is better" case that optimization is chasing.

The result looks contradictory next to TRAVERSE but isn't, because the two studies answer different questions. Physiologic replacement in genuinely deficient, symptomatic men — the TRAVERSE population — is not the same intervention as lifelong higher or supraphysiologic exposure, which is what the genetic instrument models. MR captures cumulative exposure from puberty onward; clinical trials capture months-to-years of restoration in older men whose deficiency itself drives visceral adiposity, insulin resistance, and inflammation. Restoring testosterone in a metabolically compromised 60-year-old rescues the dominant short-term killer; lifelong elevated testosterone in an otherwise healthy 30-year-old slowly raises blood pressure for decades.^[16] So the MR signal does not overturn the post-TRAVERSE safety story for true hypogonadism; it argues specifically against testosterone "optimization" in eugonadal men.

The take-home: target the middle of the normal range, not the top. Optimization toward 25-year-old physiology in a 55-year-old buys subjective gains at the cost of decades of higher hypertensive load. This is the Wilder pattern again — benefit concentrates in hypogonadal, symptomatic men, while in healthy men the expected value tilts negative.

The mortality registry — a number that doesn't survive scrutiny (Weak)

A 2024 longitudinal registry of 737 men with adult-onset testosterone deficiency, followed for a median of 114 months, reported a hazard ratio of 0.23 (95% CI 0.14–0.40) for all-cause mortality — on its face a 77% lower death rate — with long-acting testosterone undecanoate versus untreated controls — adjusted for age, waist circumference, HbA1c, lipids, blood pressure, smoking, and diabetes.^[17]

Do not take that number at face value. A hazard ratio of 0.23 is a 77% reduction in all-cause mortality — a larger effect than statins, smoking cessation, or exercise, from a hormone injection. That is, on its face, implausible, and the study design explains why: this is a non-randomised, unblinded, single-centre registry in which the men who received treatment were the men who chose it, stayed on it, and kept turning up for eleven years of follow-up. That is the textbook setup for healthy-adherer bias (people who adhere to any therapy, including placebo, die less) and immortal-time bias (you cannot be counted as a treated survivor until you have survived long enough to be treated). Multivariable adjustment does not remove either. The same site makes this argument at length about a far more modest effect: in glucosamine, a hazard ratio of 0.85 is shown to be reproducible from selection bias alone, with a true effect of zero. An HR of 0.23 from a weaker design deserves more scepticism, not less.

The decisive comparison is with the randomised evidence. TRAVERSE — 5,204 men, placebo-controlled — found no mortality benefit from testosterone. When a registry and an RCT disagree by that margin, the RCT wins.

What survives is the shape, not the magnitude. In the registry's stratified analysis the apparent survival advantage was concentrated in men with the highest baseline cardiometabolic risk, and near-absent in relatively healthy hypogonadal men — the pattern the authors labelled Wilder's "Law of Initial Value," and the same pattern seen with Ozempic-class drugs: biggest effect in the sickest patients, smallest in healthy adults seeking optimization. That directional finding is consistent with the rest of the evidence base. The 77% figure is not. TRT is a treatment for hypogonadism, not a life-extension lever.

Cognition and mood (Moderate)

A 2025 systematic review and meta-analysis of 14 trials of androgen replacement in hypogonadal men reported domain-specific improvements. The effect sizes below are standardized mean differences (SMD) — a difference of about 0.2 counts as small, 0.5 as moderate, 0.8 as large — with the bracketed 95% confidence interval:^[18]

Domain	Standardized mean difference	p
Executive function	0.49 (0.37–0.60)	<0.001
Verbal/working memory	0.46 (0.34–0.58)	<0.001
Visuospatial	0.23 (0.15–0.31)	<0.001
Attention	0.22 (0.08–0.35)	0.001

Executive function and memory show the most robust signal — consistent with hippocampal androgen receptor density. Qualitative reviews also report symptom improvement in treatment-resistant depression and mild Alzheimer's.^[19] Despite this, the AUA still advises clinicians to label cognitive benefits as "inconclusive" (Grade B) — a notable institutional lag behind the meta-analytic data.^[20] For a healthy adult, TRT is not first-line for cognitive complaints; the dementia-prevention levers (hearing, BP, exercise, MIND diet, sleep) outperform on absolute terms.

Epigenetic effects (Weak / preliminary)

The TRIIM trial — 9 men aged 51–65 — used a multi-agent regimen of recombinant human growth hormone, dehydroepiandrosterone (DHEA), and metformin and reported a 1.5–2.5 year mean regression in epigenetic age across four major clocks over 12 months, plus restoration of thymic mass.^[21] A larger TRIIM-X extension is ongoing. Testosterone wasn't the principal driver, but the trial established that targeted hormone manipulation can move epigenetic clocks measurably.

In population data, higher male testosterone-to-estradiol ratios link to younger PhenoAge and substantially lower DNA methylation-based PAI-1, a biomarker of thrombosis and senescence — a one-SD increase in total T was associated with a 481 pg/mL drop in DNAm PAI-1.^[22] See hallmarks-of-aging for the underlying altered intercellular communication hallmark.

A controlled longitudinal study of 13 trans men and 13 trans women showed sexually dimorphic effects in the first year of gender-affirming hormone therapy. Testosterone in trans men roughly held the pace of aging steady — as measured by DunedinPACE, a DNA-methylation estimate of how fast a person is biologically aging — while slightly shortening methylation-estimated telomere length (p=0.037); feminizing regimens in trans women did the opposite, speeding the pace of aging (p=0.002) with a small telomere gain.^[23] Testosterone is a tissue- and sex-specific regulator, not a generic anti-aging molecule.

Female testosterone therapy (Moderate, narrow indication)

Testosterone is a primary female sex steroid; women produce more testosterone than estrogen by mass. The only globally endorsed indication for female testosterone therapy is Hypoactive Sexual Desire Disorder (HSDD) in postmenopausal women.^[24]

Claims that the case extends beyond libido rest on much weaker evidence than the HSDD indication. The most-cited recent example is a retrospective observational study of 332 women aged 27–78 receiving individualized testosterone through a telehealth platform, in which 84.3% reported improved energy and 89.7% reported improved quality of life.^[25] Read those numbers with the design in mind. There was no placebo arm, no blinding, and no baseline symptom score — the outcomes come from a single survey administered after treatment, to women who were still on the drug and had paid for it. Energy, mood, and quality of life are precisely the domains where placebo response in hormone trials runs 30–50%, and two of the authors hold a scientific-advisory role with the telehealth provider whose patient data were analysed. The paired biomarker changes (total testosterone +151.8%, triglycerides −12.6%) are real; the symptom percentages should be read as hypothesis-generating, not as an effect size. A 10-year retrospective on subcutaneous testosterone pellets in menopausal women showed sustained reductions across all 11 Menopause Rating Scale categories — also uncontrolled.^[26]

The short-term cardiovascular safety profile is reassuring: a 2026 systematic review of 13 randomised trials in 2,628 cisgender women on transdermal testosterone for 8–52 weeks reported no cardiovascular deaths. The same review rated the certainty of long-term evidence low — these are short trials, not lifetime safety data.^[27] Dose-dependent androgenic effects (mild acne, hirsutism) are managed by titration or spironolactone.

The biggest current obstacle is supply: there is no FDA-approved female-specific testosterone formulation. Clinicians fractionate male transdermal gels off-label or use compounded products, which the global consensus statement explicitly advises against because of inconsistent dosing.^[28] A licensed female transdermal patch is the most pressing regulatory frontier.

Diagnostic and monitoring framework (Strong on protocol)

If lifestyle correction has been adequate and symptoms persist with documented biochemical deficiency, the standard protocol is:

Baseline (in addition to the general midlife lab panel):

Two morning total testosterone draws (peak production is shortly after waking).
Hematocrit (must be <50%); luteinizing hormone (LH), follicle-stimulating hormone (FSH), prolactin, estradiol; prostate-specific antigen (PSA) in men over 40.
Comprehensive metabolic, lipids (apoB), HbA1c.

Formulation choice (pharmacokinetics differ markedly):

Formulation	Frequency	Profile	Trade-offs
Cypionate / enanthate IM	7–14 d	Sharp peaks 36–48 h post-dose, troughs by day 14	Cheap; precise titration; highest erythrocytosis rate and mood/libido swings
Undecanoate IM (long-acting)	Initial, 4 wk, then every 10 wk	Stable steady state by day 7	Most convenient; FDA Risk Evaluation and Mitigation Strategy (REMS) — 30-min post-injection observation for rare pulmonary embolism risk
Transdermal gel (1%, 1.62%, 2%)	Daily	Approximates circadian profile	Avoids injection peaks; transfer risk to women and children on skin contact
Subcutaneous pellets	Every 3–4 mo	Steady release after mild initial spike	Excellent compliance; minor surgical procedure; hard to dose-adjust if AE occurs

Sources: ^[29], ^[30].

Targets and follow-up:

Aim for the middle tertile of the reference range (~450–600 ng/dL, roughly 15.6–20.8 nmol/L), not the top — consistent with the MR finding that lifelong high-normal carries hypertensive load.
Recheck at 2–4 weeks for gels; at mid-cycle after 3–4 cycles for short-acting IM.
Stable phase: every 6–12 months on testosterone, hematocrit, PSA, lipid panel, BP.
Hematocrit >54% is the bright line: dose-reduce, lengthen interval, switch to gel, or therapeutic phlebotomy. Severe erythrocytosis raises blood viscosity and thrombotic risk.^[31]
Three-to-six-month efficacy review. If serum normalized but no symptom improvement, discuss cessation.

What's overrated

TRT as a generic longevity drug. The mortality benefit is concentrated in metabolically compromised men with documented deficiency. In healthy eugonadal men the risk-benefit is dominated by HPG-axis suppression (testicular atrophy, infertility) and erythrocytosis.
"Optimizing" healthy men into the high-normal range. The MR signal puts a real cost on lifelong elevated testosterone via blood pressure. Aim middle-tertile, not 25-year-old peak.
TRT as a libido or energy fix without OSA screening. OSA causes both, suppresses endogenous T directly, and is one of the most cost-effective things to fix in midlife — see sleep-breathing.
Compounded female testosterone products. Global consensus advises against until licensed female formulations exist.
The "beyond libido" case for female testosterone. The eye-catching energy and quality-of-life percentages come from uncontrolled, unblinded, industry-linked retrospective cohorts in a symptom domain with a large placebo response. HSDD in postmenopausal women remains the only indication supported by randomised evidence.
Starting TRT without a baseline workup. The estimated 25% of patients who do this are skipping the easiest reversible causes.

Cautions

Fertility loss. Exogenous testosterone suppresses the HPG axis and halts spermatogenesis. Men who want to preserve fertility need different protocols (clomiphene, hCG, enclomiphene) — TRT is not the answer.
Polycythemia / erythrocytosis. The single most consistent adverse effect; monitor hematocrit every 6–12 months.
Atrial fibrillation, BP elevation, AKI, PE. All real signals from TRAVERSE; relevant for men with baseline risk.
Prostate. TRAVERSE found no increase in prostate cancer or worsening LUTS, debunking the older fear, but PSA monitoring remains standard of care over age 40.
Drug interactions. SSRIs, opioids, glucocorticoids, and several common medications suppress endogenous T independently — review the medication list before treating.