Hematocrit: the test most likely to change your prescription
Testosterone reliably raises the red cell count. Who rises, how far, and what that means for clotting risk depend more on the man and the preparation than on the target.
Hematocrit is the test most likely to change a testosterone prescription. It is the proportion of blood volume made up of red cells, reported as a percentage; haemoglobin measures the oxygen-carrying protein inside those cells and usually moves in parallel. A rise on treatment is expected pharmacology rather than a complication, and the practical question is which men rise far enough for it to matter.
The size of the expected rise
Across 51 studies pooled in a 2010 meta-analysis, testosterone raised hematocrit by a weighted mean of 3.18 percentage points (95% CI 1.35 to 5.01) and haemoglobin by 0.80 g/dL (0.45 to 1.14). Compared with placebo, the pooled relative risk of developing erythrocytosis was 8.14 (95% CI 1.87 to 35.40) across four trials at low risk of bias — a wide interval, with the direction not in doubt. It is the commonest reason testosterone therapy is modified or stopped.
Those averages conceal a wide spread. In young men treated to mid-normal concentrations, haemoglobin barely moved; it climbed meaningfully only at clearly supraphysiological levels. The men in whom the rise becomes clinically awkward are generally older, heavier, and on a preparation that delivers testosterone in peaks.
The mechanism
In a randomized trial of 166 older men on testosterone gel, haemoglobin and hematocrit rose by 7% and 10%, erythropoietin rose 58% at one month, and hepcidin fell by about 49%. Hepcidin restricts iron release, so suppressing it makes more iron available for red cell production. Both the erythropoietin rise and the hepcidin fall had returned toward baseline by six months despite continued treatment, which the authors interpreted as the body resetting its erythropoietin-to-haemoglobin set point rather than running the old system harder.
The causal step was demonstrated in mice: testosterone inhibits hepcidin transcription through the androgen receptor acting on BMP/Smad signalling, and mice engineered to overexpress hepcidin in the liver did not show the haemoglobin rise at all. Erythropoietin alone does not account for the whole effect, however. In a graded-dose study of 121 men, haemoglobin and hematocrit rose with dose as expected, while the changes in erythropoietin did not differ across dose groups and did not correlate with the change in testosterone.
What the thresholds say
The published thresholds are less uniform than their reporting suggests.
| Source | Before starting | During treatment |
|---|---|---|
| Endocrine Society (2018) | The graded recommendation lists "elevated hematocrit" as a contraindication and attaches no number to it. Its 2010 predecessor used >50% | Erythrocytosis defined as >54%; withhold until it normalizes, then resume at a lower dose |
| AUA (published 2018, reviewed and validity confirmed 2024) | The graded statement says only to measure haemoglobin and hematocrit and counsel on the risk — no number. The supporting text describes withholding at >50% | Reduce the dose or temporarily stop at ≥54% |
| VA Pharmacy Benefits Management Services, Recommendations for Use (January 2026) | Not specified | Reduce the dose if hematocrit exceeds 51%; stop if it exceeds 54% |
| FDA label, testosterone gel (2025) | Check hematocrit before starting — no number | "If hematocrit becomes elevated, stop therapy until hematocrit decreases to an acceptable concentration" — no number |
The 54% stopping threshold is close to universal and can be relied on. The starting threshold is not. Where societies do commit to a pre-treatment number they land in a narrow band: the European Academy of Andrology gives 48 to 50%, the American Association of Clinical Endocrinology and the International Consultation on Sexual Medicine give 50%, as did the Endocrine Society's 2010 guideline, and the International Society for the Study of the Aging Male gives 52%. A figure of 48% is widely attributed to the current Endocrine Society guideline, and a 2024 systematic review in the Asian Journal of Andrology tabulates it that way, but it does not appear in the graded recommendation, which is the part of the document carrying the Society's formal strength rating. Somewhere between 48 and 52% a pause is warranted, and no single authoritative line exists.
Monitoring follows the guideline minimum: haemoglobin and hematocrit at 3 to 6 months after starting, then annually. The Endocrine Society, the VA document and the FDA label all say the same.
What raises the risk
The preparation. Short-acting injectables raise hematocrit most and patches least. In a network meta-analysis of 29 placebo-controlled trials in 3,393 men, the mean rise against placebo was 4.0 percentage points for injectable enanthate or cypionate (2.9 to 5.1), 4.3 for oral undecanoate (0.7 to 8.0), 3.0 for gel (1.8 to 4.3), 1.6 for injectable undecanoate (0.3 to 3.0) and 1.4 for the patch (0.2 to 2.6). Only injectable cypionate or enanthate against patch differed significantly head to head; the rest of the ordering is real but not statistically separated.
Incidence figures circulating elsewhere deserve care. The widely quoted 67% on injectables, 13% on gel and 35% on pellets come from a single retrospective review of 178 men, which states itself that its injectable group was more than a decade younger and started with higher haemoglobin and hematocrit. No meta-analysis of erythrocytosis incidence by formulation exists. The cleanest randomized figure comes from the T4DM trial, in older and heavier men on injectable testosterone undecanoate over two years: 106 of 491 men (22%) crossed the 54% safety trigger, against 6 of 484 (1%) on placebo. That figure describes that population and that preparation, not treatment in general.
Sleep apnoea. Independently associated, with an adjusted odds ratio of 2.09 (95% CI 1.17 to 3.76) for a hematocrit at or above 52% in one cohort of 474 men. See when testosterone isn't the answer.
SGLT2 inhibitors. Taken alongside testosterone, the adjusted odds of a new hematocrit above 54% within a year were 3.80 (2.27 to 6.11) compared with the SGLT2 inhibitor alone and 2.49 (1.51 to 3.59) compared with testosterone alone.
Time. With transdermal treatment the rise continues for roughly the first five to six months and then plateaus. Time to cross 50% differed by route in one series: about 10.5 months on injectables, 14.0 on gel, 16.4 on pellets. In a transmasculine cohort the mean time to peak was 31 months, so a six-month plateau is not a universal rule.
Erythrocytosis and clotting
The claim that a high hematocrit on testosterone causes blood clots has a plausible mechanism and real physiology behind it. The outcome data are thinner than the confidence with which it is usually asserted. A 2018 review concluded that no randomized or prospective study had observed a direct relationship between testosterone-induced erythrocytosis and thromboembolic events. A 2025 systematic review of 45 studies found exactly one retrospective study reporting increased events in treated men who developed erythrocytosis compared with those who did not, and concluded that the existing literature does not suggest a clear association. In a cohort of 519 transgender patients on testosterone, 20% developed a hematocrit above 50% and five had a thromboembolic event — none of whom had erythrocytosis at the time of the event.
TRAVERSE found more pulmonary embolism, more atrial fibrillation and more acute kidney injury in the testosterone arm, which belongs in the decision. But TRAVERSE actively titrated doses to keep hematocrit below 54%, so by design it cannot report what erythrocytosis itself does, and it did not report thrombotic events stratified by hematocrit. The 54% threshold is therefore a precaution built on physiology rather than a line at which measured risk is known to change.
Phlebotomy
Phlebotomy is frequently offered, and the evidence behind it is weaker than its popularity suggests. Two independent reviews, in 2024 and 2025, concluded that evidence for the efficacy or safety of therapeutic phlebotomy in this setting is lacking, and both raised the possibility of net harm: phlebotomy depletes iron and lowers tissue oxygen, and in a different erythrocytosis context arterial thrombosis was more common among those who underwent it. No study has tested phlebotomy here against a clinical outcome rather than against the number on the report. Lowering the dose, lengthening the interval or changing preparation addresses the cause; phlebotomy addresses the measurement.
Because age, weight, baseline count and preparation drive this more than the target does, the same number is comfortable in one man and not in another, and sorting out which is which is part of the individual assessment at the practice.
Questions patients ask
My hematocrit went up on testosterone, so something has gone wrong.
A rise is the expected pharmacology, not a complication.
What the evidence showsPooled across 51 studies, testosterone raised hematocrit by a weighted mean of 3.18 points and haemoglobin by 0.80 g/dL. The mechanism — suppression of hepcidin, increasing iron availability — has been demonstrated causally in animals and tracked in men. In young men at mid-normal concentrations haemoglobin barely moved.
What remains uncertainWhich men will rise the most; the predictors are only partly understood.
Bottom lineExpect it, monitor it, and act on it at the agreed threshold.
Strong
There is one hematocrit level above which I definitely cannot start testosterone.
There is a band rather than a line, and the societies do not agree on where it sits.
What the evidence showsThe Endocrine Society's 2018 graded recommendation lists "elevated hematocrit" as a contraindication without a number, and its 2010 predecessor used 50%. The AUA's graded statement contains no number either; its supporting text describes withholding above 50%. The FDA gel label says to check hematocrit before starting and gives no figure. Societies that do commit cluster between 48 and 52%.
What remains uncertainNo pre-treatment threshold anywhere has been derived from outcome data. Each is a judgment about where the expected rise becomes unacceptable.
Bottom lineSomewhere in the high forties a pause is warranted. A single quoted figure is convention rather than settled evidence.
Moderate
High hematocrit on testosterone will give me a blood clot.
The mechanism is solid; the outcome data are not.
What the evidence showsA 2025 systematic review of 45 studies found a single retrospective study linking treatment-induced erythrocytosis to increased events, and concluded the literature does not show a clear association. In 519 transgender patients, five thromboembolic events occurred and none of those patients had erythrocytosis at the time. TRAVERSE did report more pulmonary embolism overall, but titrated hematocrit below 54% throughout.
What remains uncertainEssentially the central question. No trial has been designed to answer it.
Bottom lineTreat 54% as a sensible precaution rather than a measured cliff edge.
Mixed — Strong on mechanism, Limited on outcomes
Donating blood or getting phlebotomy fixes it.
It fixes the number. Whether it fixes anything else is unknown.
What the evidence showsTwo reviews in 2024 and 2025 concluded that evidence for the efficacy or safety of therapeutic phlebotomy in testosterone-induced erythrocytosis is lacking, and both raised the possibility of harm through iron depletion. No study has measured a clinical outcome.
What remains uncertainWhether phlebotomy helps, does nothing, or harms.
Bottom lineAddress the dose or the preparation first, which treats the cause.
Limited
Switching from injections to gel would help.
Plausibly, and the trial data support the direction.
What the evidence showsIn a network meta-analysis of 29 trials, the mean rise was 4.0 points on injectable enanthate or cypionate against 3.0 on gel, 1.6 on injectable undecanoate and 1.4 on the patch. Only injectable against patch reached statistical significance head to head.
What remains uncertainHow much of an individual man's rise is the preparation and how much is him.
Bottom lineA reasonable move, alongside dose reduction. Both are worth discussing.
Moderate
If I have sleep apnoea I just need to treat it and my hematocrit will settle.
Treating sleep apnoea is worth doing on its own merits, but it does not undo the testosterone effect.
What the evidence showsSleep apnoea independently raised the odds of polycythaemia on testosterone (adjusted OR 2.09, 95% CI 1.17 to 3.76). Separately, CPAP does not raise testosterone: a meta-analysis of seven studies in 232 men found no significant effect.
What remains uncertainWhether treating apnoea reduces testosterone-associated erythrocytosis. It has not been studied.
Bottom lineTwo real problems that interact. Treat both, and expect neither to solve the other.
Moderate
Ready for testosterone care built on a diagnosis?
The first step is a full endocrine evaluation, not a prescription. We see patients across San Diego County and welcome referrals from other physicians.
Our endocrinologists evaluate low testosterone in La Jolla, La Mesa and Poway: low testosterone care at Diabetes & Endocrine Specialists.


