Is testosterone therapy safe long term? The evidence runs about three years

Whether testosterone therapy is safe long term is not known, because no trial has run long enough to answer it. The largest attempt, TRAVERSE, gave a daily testosterone gel or a dummy gel to 5,246 men aged 45 to 80 who had symptoms of deficiency plus two fasting blood readings below 300 nanograms per decilitre, the unit a laboratory prints testosterone in [2].

Inside its own window, the trial was reassuring. A heart attack, stroke or heart-related death reached 182 men on testosterone (7.0%) against 190 on the dummy gel (7.3%) [2]. But the men used the gel for about 22 months on average and were followed for about 33 months [2], so what the trial cleared is roughly two years of treatment.

Three other problems were more common in the testosterone group over those same months: atrial fibrillation, an irregular heart rhythm; sudden kidney injury; and a blood clot in the lungs [2].

The one study that looked past the trial window found the opposite direction. Researchers used National Health Service records from Glasgow to compare 440 men aged 51 and over who had been prescribed testosterone across an interval of at least two years with 136,051 men who had not, then followed both groups from 2017 to 2022 [14]. The treated men were about 55% more likely to have what the researchers called a major adverse cardiovascular event - a heart attack, unstable angina, stroke, heart failure or death from a heart cause - after accounting for age, ethnicity, deprivation and other illnesses [14]. That is a relative figure, and the underlying rate is not reported beside it.

A 2024 review of testosterone and heart health concluded that the newer data confirm the medium-term cardiovascular safety of testosterone therapy in middle-aged and older men with low testosterone [6]. The word there was medium-term, not long-term. Before TRAVERSE no trial had been designed or built large enough to detect heart events at all [7], and the two regulators read the older evidence in opposite ways: the American FDA issued a warning about possible cardiovascular risk, and Europe's medicines regulator reviewed the same material and declined to endorse it [15].

Testosterone reaches the cardiovascular system from several directions at once, which is why the biological arguments run both ways. Laboratory work shows it acting on skeletal muscle, heart muscle cells, blood vessels, fat tissue, insulin action and the production of red blood cells [6].

One of those effects is measurable and points towards risk. Testosterone raises haematocrit, the share of blood made up of red blood cells, which makes blood thicker - though most studies have not found a matching rise in clots forming in the veins [6].

The argument on the other side rests on the lining of the blood vessels, and it is the weaker of the two. The endothelium is the single-cell layer lining every blood vessel, and it releases nitric oxide, a gas that relaxes the vessel and works against the processes that build clots and fatty plaque [11]. Androgens, the hormone family testosterone belongs to, appear to help that lining function and to help repair it after injury [8], and earlier reviews argued testosterone might protect arteries this way [10]. Whether the body's own testosterone helps or harms the lining is still unsettled, and the data on it have been described as scanty and contradictory [11].

This is the trouble with reasoning from mechanism here. There is a plausible biological story for benefit and a plausible one for harm, and they cannot both be right about the same men. Counting actual heart attacks settles the argument in a way the biology cannot, and the trial counted them for about two years.

The long-term signal is a single observational study, and its design carries a specific weakness. Only 440 men in it had taken testosterone across an interval of two years or more, and nobody randomised them: a doctor decided who received a prescription, so the treated men may have been less well in ways the adjustment for age, ethnicity, deprivation and other illnesses did not capture [14]. A pattern of that shape cannot show that testosterone caused the extra events.

The trial evidence has the reverse problem - clean design, narrow entry. TRAVERSE enrolled only men who had both a blood-confirmed deficiency and existing heart disease or a high risk of it [2].

The evidence here does not describe:

• men whose testosterone sits in the normal range and is drifting down with age, or falling with obesity; prescriptions for those two reasons rose sharply in the decade before 2019 even though neither is an approved reason to prescribe it [7]
• women, whose evidence base is separate and thin: testosterone applied through the skin has not shown cardiovascular harm in women, but the link between testosterone and heart disease in older women has not been established [4]
• transgender men taking testosterone, a group whose cardiovascular research has been underpowered and short of proper comparison groups [9]
• men under 45, who were outside the trial's age range [2]
• anyone treated for a decade or more, which nothing here covers

One more gap is worth naming. No trial published before TRAVERSE was designed or large enough to detect heart events [7], so much of the older literature that both sides of this argument quote was never built to settle it.

A 52-year-old man with fatigue and low libido has a blood test, and it comes back below the threshold twice. His doctor raises testosterone gel. For the first two or three years, the trial evidence covers a man like him reasonably well [2].

The question it cannot answer is the one he is more likely to face, because men who start treatment for a confirmed deficiency often stay on it: what the picture looks like at year eight. That is the gap the Glasgow researchers were pointing at when they called for longer-term data [14]. It is also why clinical guidance for male hypogonadism, the medical term for a testosterone deficiency confirmed by blood test, treats monitoring for both benefits and side effects as part of the treatment rather than an optional extra [5].