The Cardiovascular Paradox: How the Same Hormone Both Protects and Threatens the Heart

Two findings. Both well-powered. Both peer-reviewed. Both apparently correct.

The first: a 2024 individual-participant meta-analysis — 11 cohorts, 24,000 men, 255,830 person-years, mass spectrometry only — found that men with testosterone below 7.4 nmol/L die sooner. Below 5.3 nmol/L, they die of cardiovascular disease specifically. The lower the testosterone, the higher the risk, dose-response confirmed.

The second: a 2025 Mendelian randomization study — 425,097 men genotyped, 1.17 million controls — found that genetically higher testosterone causes coronary artery disease. OR 1.17. P = 0.0003. Confirmed by three sensitivity analyses. Not lipid-mediated. Mediated by blood pressure.

Low testosterone predicts death. Higher testosterone causes heart disease. The same hormone appears on both sides of the equation, and neither finding is wrong.

This is the cardiovascular paradox — the deepest of the three consequences paradoxes, because it pits the gold standard of observational evidence against the gold standard of causal inference. And its resolution doesn't just clarify testosterone biology. It reveals how different types of evidence can tell apparently opposite stories that are all, in their own frame, true.

The Case for "Low T Kills"

The observational evidence is extensive and consistent. Yeap and colleagues' 2024 individual-participant data meta-analysis in the Annals of Internal Medicine represents the most rigorous synthesis to date: 11 cohorts, mass spectrometry only (eliminating assay noise), prospective design. Men in the lowest testosterone quintile faced elevated all-cause mortality. Those below 5.3 nmol/L faced elevated cardiovascular death specifically. LH above 10 IU/L — a signal of primary testicular failure — independently predicted death.

Corona and colleagues' earlier meta-analysis (2018, Journal of Sexual Medicine) — 37 observational studies, 43,041 subjects — showed the same pattern: low testosterone predicted cardiovascular mortality and morbidity. Jiang and colleagues' 2025 NHANES analysis narrowed the lens to men with existing cardiovascular disease: 70.68% had testosterone below 300 ng/dL. Those with low T had an adjusted HR of 1.48 for all-cause mortality. Strikingly, the association was stronger in men without hypertension (HR 1.93) than with it (HR 1.35) — suggesting that when hypertension is already present, it dilutes the independent contribution of low testosterone to the shared risk.

The UK Biobank — 200,000 men, the largest prospective cohort — confirmed the direction: lowest testosterone quintile, 14% higher all-cause mortality, 10,053 deaths over follow-up. But here the signal diverged. Serum testosterone was not associated with cardiovascular-specific mortality (1,925 CV deaths). Not with incident heart failure. Not with MI. Only with stroke did a significant association emerge.

This is the first crack in the simple narrative. The biggest dataset shows that low T predicts dying — but not specifically dying of heart disease.

The Mendelian Randomization Challenge

Mendelian randomization uses genetic variants as instruments: if a gene raises testosterone lifelong, and that same gene associates with disease, we infer causation — because genes are assigned at conception, before any confounders accumulate. It's the closest thing to an RCT that genetics can offer.

Morbey and colleagues (2025, JCEM) used 94 testosterone-associated SNPs from the UK Biobank (425,097 men) and tested their association with CAD in CARDIoGRAMplusC4D (1,165,690 participants). The result: genetically higher testosterone → 17% higher CAD risk in men (OR 1.17, 95% CI 1.08-1.27, P = 3.32 × 10⁻⁴). Confirmed by MR-Egger (OR 1.24) and weighted median (OR 1.22). No effect in women (OR 1.01).

Then the mediation analysis revealed the mechanism:

Schooling, Luo and colleagues (2019, BMJ) found an even more alarming MR signal: genetically predicted testosterone → thromboembolism (OR 2.09), heart failure (OR 7.81), MI validation (OR 1.37). When they used SHBG-predicted testosterone as an alternate instrument, the associations vanished — suggesting testosterone itself, not SHBG, drives the risk.

Rhomberg-Kauert and colleagues' 2025 systematic review of 29 MR studies confirmed the pattern: higher genetically predicted testosterone consistently associates with adverse cardiovascular outcomes including thromboembolism, ischemic heart disease, and heart failure.

In the same paper where Morbey showed genetically higher T → CAD, the phenotypic UK Biobank analysis showed unadjusted low T → HR 1.15 for incident CAD. But after adjusting for all QRISK variables, it became HR 0.97 (P = 0.11) — fully attenuated. Type 2 diabetes caused 54.3% of the attenuation. BMI caused 48.1%. The observational association disappeared in the same paper that demonstrated the MR association.

What Each Type of Evidence Actually Answers

The paradox dissolves when you recognize that these evidence types are answering different questions about different populations on different timescales.

Evidence Type	Question It Answers	Finding	Timescale
Observational cohorts	Do sick men have low T?	Yes — strongly	Point-in-time
Observational (adjusted)	Does low T cause CVD independently?	Mostly not — confounders explain it	Point-in-time
Mendelian randomization	Does lifetime high T cause CAD?	Yes — via blood pressure	Lifetime genetic exposure
Natural experiments (KS, pituitary)	Does severe deficiency kill?	Yes — unambiguously	Years of severe deficit
RCTs (TRAVERSE)	Does TRT in mild HH cause CV harm?	No — HR 0.73 MACE (noninferiority)	33 months treatment
ADT studies	Does iatrogenic castration harm CV?	Yes — especially with baseline CVD	Months to years

A man born with genetically high testosterone who has elevated blood pressure from puberty onward is not the same as a 60-year-old hypogonadal man given replacement therapy. The MR captures one. The RCT captures the other. Neither is wrong. They're describing different biological realities.

The Natural Experiments That Prove Severity Matters

The strongest evidence that severe testosterone deficiency causes cardiovascular harm — not merely marks it — comes from situations where deficiency is the primary lesion, not a downstream symptom of obesity or frailty.

Munro and colleagues (2025, JCEM) followed men with pituitary adenomas causing secondary hypogonadism. Those with untreated SHG had a 20.5% death rate over median 8.1 years. Those treated with TRT: 6.5%. Controls without SHG: 3.3% (P = 0.0013). These were men with organic, documented hypothalamic-pituitary failure — not functional hypogonadism from metabolic disease.

Chang and colleagues (2025, Lancet Regional Health Europe) studied Klinefelter syndrome across Danish national registries over 12.9 years. TRT-treated men had HR 0.56 for mortality versus untreated. Yet only half of diagnosed Klinefelter men receive testosterone.

Androgen deprivation therapy for prostate cancer provides the inverse natural experiment. GnRH agonists increase cardiovascular risk, amplified fivefold in men with pre-existing CVD. The REVELUTION mechanism is quantified: ADT increases arterial stiffness, insulin resistance, and inflammatory markers within months. Even long-term follow-up of RTOG 9202 confirmed that extended ADT (28 months versus 4) did not increase CV mortality overall at 19.6 years — except in men with baseline cardiovascular disease, where it did.

The pattern is clear: severity determines causality. Mild, functional hypogonadism — where low T is a companion to obesity, diabetes, and frailty — is mostly a marker. Severe, organic hypogonadism — where low T is the primary defect — is a cause. The evidence types that show "low T kills" and the evidence types that show "confounders explain it" are studying different populations on this spectrum.

The U-Shaped Truth

The reconciliation becomes clearest through the lens of the U-shaped curve. Both extremes of testosterone — very low and very high — converge on cardiovascular harm, but through completely different mechanisms.

At the low extreme, five mechanisms operate — each supported by independent evidence:

1. QTc prolongation and sudden death. Testosterone shortens the cardiac QT interval by approximately 20 ms through activation of hERG (IKr) and IKs potassium channels. When testosterone drops — whether from hypogonadism or ADT — QTc lengthens, creating vulnerability to torsades de pointes. Lazzerini and colleagues (JAHA 2022) described an inflammation-hypogonadism-QTc triangle: IL-6 drives aromatase activity, lowering testosterone and lengthening QTc simultaneously. ADT pharmacovigilance data documents 184 cases of acquired long QT syndrome or torsades de pointes (11% fatal) and 99 sudden cardiac deaths. Transgender hormone data from 2025 confirms the direct relationship: testosterone use shortens QTc.

2. Endothelial dysfunction. Testosterone maintains nitric oxide production via eNOS. Low testosterone → decreased NO → endothelial dysfunction → decreased vasodilation, increased adhesion molecule expression, and accelerated atherosclerosis. ADMA (asymmetric dimethylarginine) rises, further inhibiting eNOS. Endothelial progenitor cells decline. A single testosterone dose has been shown to improve ADMA levels.

3. Chronic inflammation. Low testosterone associates with elevated IL-6, TNF-α, and IL-1β — the NF-κB axis that drives atherosclerosis. The 2024 Frontiers review describes "inflammaging" and late-onset hypogonadism as interlocking processes, each accelerating the other.

4. Metabolic dysregulation. This is the metabolic trap I mapped previously — five feedback loops locking insulin resistance and low testosterone together. The cardiovascular consequences of this trap include accelerated atherosclerosis, atherogenic dyslipidemia, and diabetogenesis. The diabetes connection is particularly strong: the T4DM trial showed 41% diabetes reduction with TRT in high-risk men. Hackett's re-analysis of TRAVERSE metabolic data found 22.5% diabetes reduction in prediabetic participants.

5. Myocardial dysfunction. 30-50% of men with heart failure have testosterone deficiency. Theodorakis and colleagues (Hormones 2025) systematically reviewed 12 RCTs of TRT in heart failure and found improved muscle strength, aerobic capacity, lean mass, insulin sensitivity, and QT shortening — with no adverse effects on blood pressure, lipids, or heart rate. A 2025 propensity-matched study found that TRT in heart failure patients with malnutrition reduced all-cause mortality (HR 0.56) and acute HF events (HR 0.62).

At the high extreme, two mechanisms — now confirmed by MR — operate:

Blood pressure elevation. Morbey's mediation analysis demonstrated that genetically predicted testosterone raises diastolic blood pressure, and this accounts for most of the MR-observed CAD risk. Adjusting for diastolic BP attenuated the OR from 1.17 to a non-significant 1.09. Lifetime exposure to mildly elevated blood pressure, accumulated from puberty, is a different beast from a middle-aged man's hemodynamic response to TRT.

Erythrocytosis and thrombosis. Testosterone stimulates erythropoiesis. At physiological levels, this is a feature. At supraphysiological levels — or with lifetime genetic elevation — it becomes a 315% increased risk of erythrocytosis versus controls. A 2022 Journal of Urology analysis found that polycythemia on TRT increased MACE and VTE from 3.87% to 5.15% (OR 1.35). A 2025 Blood Advances systematic review (45 studies, 35 on testosterone) confirmed erythrocytosis as the most common TRT adverse effect. Schooling's MR finding of OR 2.09 for thromboembolism operates through this pathway — lifetime elevation of hematocrit, not acute replacement.

The Question TRAVERSE Actually Answered

TRAVERSE enrolled 5,246 men aged 45-80 with established or high-risk cardiovascular disease and testosterone 100-300 ng/dL. Median treatment: 21.7 months. Primary MACE: HR 0.73 (95% CI 0.52-1.03), meeting noninferiority. Not superiority — but noninferiority in a high-CV-risk population.

What TRAVERSE settled: giving testosterone gel to older men with mild-to-moderate hypogonadism and cardiovascular risk does not increase heart attacks, strokes, or CV death over ~3 years. That's a narrow but critical answer. What it did not settle: whether TRT protects the heart, whether younger men benefit differently, whether severe hypogonadism requires different analysis, whether injectable testosterone behaves differently from gel, or whether the secondary signals (atrial fibrillation: 91 vs 63, P = 0.02; acute kidney injury: 60 vs 40, P = 0.04) reflect real risks or — as Zitzmann and colleagues argued — COVID-19 confounding in a trial that straddled the pandemic.

Morbey's power calculation is instructive: detecting an OR of 1.17 with 90% power at a 7.3% event rate requires 6,439 participants. TRAVERSE had 5,246 — slightly underpowered for the MR-predicted effect size. The trial's null finding is consistent with both "no effect" and "too small an effect to detect."

Two Expert Panels, One Conclusion, Different Emphasis

In 2024-2026, two major expert panels reviewed the entirety of this evidence and reached the same core conclusion through different analytical postures.

The Androgen Society position paper (Morgentaler et al., Mayo Clinic Proceedings, November 2024) was decisive: "Conclusively determined that TTh is not associated with increased risk of heart attacks, stroke, or CV deaths." They reviewed decades of research plus 19 meta-analyses and framed the paper as a regulatory manifesto — the science has spoken, policy should follow.

The PaTeR consensus (Zitzmann et al., Andrology, January 2026) was more nuanced. They acknowledged TRAVERSE's strength but noted its limitations: older, high-risk population, short follow-up, single formulation. They urged distinguishing organic from functional hypogonadism, individualized care over population-level rules, and practical monitoring including hematocrit and blood pressure — precisely the two mediators the MR data highlights.

Both panels agree: TRT is cardiovascularly safe. PaTeR adds: for whom, and with what monitoring, matters.

Marker Versus Cause — and the Severity Gradient

Yeap and Anawalt's 2026 JCEM narrative review argued that low testosterone is more marker than cause of cardiovascular disease. UK Biobank data showed low T → all-cause mortality but not CV mortality. The confounders — obesity, diabetes, frailty — are the primary drivers. Low testosterone is the thermometer, not the fire.

But the natural experiments refuse to cooperate with this tidy conclusion. Klinefelter men dying at higher rates without TRT are not obese diabetics with incidental low T. Men with pituitary adenomas having three times the death rate are not explained by confounders. ADT causing arterial stiffness and metabolic syndrome within months is not reverse causation.

The resolution is a severity gradient:

This gradient explains why evidence types disagree. Observational studies pool all three tiers and find "low T predicts death" — because it does, for different reasons across the spectrum. MR captures lifetime genetic exposure — the high end of the curve, not replacement therapy. Natural experiments isolate the severe tier and find true causation. TRAVERSE samples the moderate tier and finds safety but not dramatic benefit.

The Uncomfortable Asymmetry

TRAVERSE tested testosterone gel. The cardiovascular evidence base — such as it is — covers testosterone replacement therapy. There is zero cardiovascular outcome data for SERMs (enclomiphene, clomiphene), hCG monotherapy, kisspeptin analogs, or any other alternative approach to treating secondary hypogonadism. The treatments most commonly recommended to avoid the perceived cardiovascular risks of testosterone have never been tested for cardiovascular safety.

This is not a minor gap. SERMs raise LH and FSH — and the cognitive paradox article documented that elevated FSH may drive neurodegeneration. Whether elevated gonadotropins have cardiovascular implications is unknown. hCG monotherapy has lower erythrocytosis risk than TRT — a genuine advantage given the MR data on hematocrit — but this theoretical benefit has never been tested in a cardiovascular outcome trial.

The evidence asymmetry means that the treatment with the most safety data (TRT) looks riskier than treatments with no safety data at all. This is a measurement artifact, not a biological fact.

What Remains Unknown

Five specific gaps prevent closure:

1. Formulation matters, but no comparison exists. TRAVERSE tested 1.62% transdermal gel. Injectables produce different pharmacokinetics — higher peaks, lower troughs. Whether this affects CV outcomes is unstudied.
2. Younger men are absent. TRAVERSE enrolled ≥45. The men most likely to benefit from TRT — those with organic secondary hypogonadism in their 30s and 40s — have never been studied in a CV outcome trial.
3. Severity was excluded. TRAVERSE required T between 100-300 ng/dL. Men with T below 100 — the tier where natural experiments show clear mortality benefit — were excluded.
4. Hematocrit monitoring was protocol-driven. TRAVERSE protocol-mandated phlebotomy for high hematocrit. Whether real-world TRT without such monitoring produces worse CV outcomes is unknown.
5. The MR blood pressure pathway needs clinical testing. If testosterone raises CAD risk via blood pressure, then monitoring and treating blood pressure in men on TRT should mitigate this risk. This has never been tested prospectively.

Where This Leaves the Evidence

The cardiovascular paradox is real, but it is not contradictory. Each piece of evidence correctly describes the slice of reality it measures. The resolution lies in three principles:

The U-shaped curve is fundamental. DHT, free testosterone, and total testosterone all show U-shaped relationships with cardiovascular outcomes. Mid-range is optimal. Both deficiency and excess produce harm — through entirely different mechanisms (endothelial/electrical/inflammatory at the low end, hemodynamic/hematological at the high end).

Severity determines whether low T is cause or marker. In organic, severe hypogonadism, testosterone deficiency is itself pathogenic. In functional, metabolic hypogonadism, low testosterone is mostly a symptom of the real disease — and treating the real disease (obesity, insulin resistance, sleep disruption) addresses both the low T and the cardiovascular risk simultaneously.

Endogenous lifetime exposure ≠ therapeutic replacement. The MR signal — T → blood pressure → CAD — reflects decades of mildly elevated pressure beginning in adolescence. TRT in a 55-year-old man, with blood pressure monitoring and hematocrit management, is a fundamentally different exposure. The same molecule, on different timescales, in different bodies, through different pathways, produces different effects. That is not a paradox. That is biology.