A man on testosterone replacement therapy has his hematocrit measured at 52%.
If he sees his urologist, therapy is stopped. The AUA threshold is 50% — he crossed it two points ago. Dose reduction, formulation switch, or discontinuation.
If he sees his endocrinologist, he's fine. The Endocrine Society threshold is 54%. Two points of margin. Continue therapy, recheck in three months.
If he sees a hematologist, neither threshold matters. Gangat, Szuber, and Tefferi's 2026 update in the American Journal of Hematology manages drug-induced erythrocytosis by symptoms, not arbitrary cutoffs. Phlebotomy for symptom control. No cytoreduction.
If he reads the literature, he finds a fourth answer. A 2024 paper in Endocrine Connections argues that phlebotomy — the intervention the hematologist just recommended — may paradoxically increase thrombotic risk through HIF-α activation and iron depletion.
Same patient. Same blood draw. Same number. Four doors, four protocols, zero randomized controlled trials supporting any of them.
The Number Everyone Watches
Erythrocytosis is the most common side effect of testosterone therapy. Depending on formulation and definition, rates range from 7% to 66.7%. Intramuscular testosterone — the most prescribed formulation — produces the highest rates. A 2025 retrospective study found 23% of patients exceeded 50% hematocrit and 5% exceeded 54%. The side effect is common, predictable, and the subject of more disagreement than almost any other monitored parameter in endocrinology.
The clinical concern is thrombosis. Higher hematocrit increases blood viscosity, which theoretically increases clotting risk. The fear comes from polycythemia vera, a myeloproliferative neoplasm where JAK2-driven erythrocytosis genuinely causes thrombotic events. But testosterone-induced erythrocytosis is a different mechanism entirely — no clonal expansion, no JAK2 mutation, no myeloproliferation. The field borrowed a threshold from one disease and applied it to another without testing whether the risk transfers.
The borrowed threshold: The 54% hematocrit cutoff used by the Endocrine Society and the European expert panels originates from polycythemia vera management, where the CYTO-PV trial showed maintaining Hct below 45% reduced cardiovascular events compared to 45-50%. The field extrapolated from a JAK2-mutant clonal disease to a pharmacologically induced secondary erythrocytosis. No study has validated any Hct threshold specifically for testosterone-induced erythrocytosis.
Seven Layers of Void
The polycythemia management question in testosterone therapy isn't a single evidence gap. It's seven interlocking gaps, each making the others harder to fill.
Layer 1: The threshold has no evidence base
Six guideline bodies have published hematocrit thresholds for testosterone therapy. None cite a study validating their specific number for this population.
| Body | Hct Threshold | Management | Year |
|---|---|---|---|
| AUA | 50% | Dose reduction / formulation switch | 2018 |
| Endocrine Society | 54% | Withhold TRT, phlebotomy listed | 2018 |
| ICSM 2024 | 54% | Multiple options, 48-50% for high-risk | 2024 |
| European Expert Panel | 54% | Phlebotomy recommended | 2026 |
| Canadian | 55% | Not detailed | — |
| Gangat (Hematology) | Symptom-based | Phlebotomy for symptoms | 2026 |
A patient at 52% is two points above one threshold and two points below another. The AUA says stop. The Endocrine Society says continue. The ICSM says it depends on comorbidities. The hematologists say ignore the number and ask about symptoms. The Canadian guidelines add another number entirely. This is not a debate about where to draw the line — it's a demonstration that nobody knows where the line belongs.
Layer 2: The standard intervention may cause harm
When hematocrit rises, three of these six bodies recommend or list phlebotomy as a management option. But a growing body of evidence questions whether phlebotomy is safe in this population.
The concern centers on HIF-α. Phlebotomy acutely lowers hematocrit, but it also lowers tissue oxygen tension, activating hypoxia-inducible factor pathways. Siebert and colleagues (2024) argue that this activation may paradoxically increase thrombotic risk — the exact outcome phlebotomy is meant to prevent. The Chuvash polycythemia model provides a natural experiment: patients with constitutive HIF-α activation have elevated thrombotic risk despite chronic phlebotomy. Iron depletion from repeated phlebotomy compounds the problem by further activating HIF-α through the iron-sensing pathway.
Meanwhile, the hematology literature takes the opposite position. Gangat (2026) recommends phlebotomy for symptom control without discussing the HIF-α concern. Neither camp cites the other. Two specialties, two literatures, zero cross-specialty trials.
Layer 3: The mortality paradox
If managing hematocrit down is protective, higher hematocrit on testosterone therapy should correlate with worse outcomes. Strange and colleagues (2021) found the opposite. In a long-term registry of 353 men on testosterone undecanoate (up to 18 years of follow-up), men who died had significantly lower hematocrit. Hematocrit in the 49-52% range was associated with lower mortality than 46-49%.
Funding disclosure: The Strange 2021 study used the Haider/Saad registry. Bayer AG partially funded statistical analyses. Saad is a Bayer consultant and shareholder, Haider a company speaker. The finding is suggestive and requires independent replication, but the inverse correlation between hematocrit and mortality on TRT is biologically plausible: better oxygen-carrying capacity, better tissue perfusion, better outcomes. What it means is that the field is actively managing down a parameter that, in the only long-term data available, correlates with survival.
Layer 4: Baseline matters more than the number
Kohn and colleagues (2024), analyzing a database of 74 million patients, found that the change in hematocrit from baseline — not the absolute value — was associated with increased risk of major adverse cardiovascular events. Two men at identical hematocrit of 52% have different risk profiles depending on whether they started at 42% (a 10-point rise) or 49% (a 3-point rise).
No guideline incorporates delta-Hct. Every threshold is absolute. The most relevant dimension of risk — how far the patient has traveled from their personal baseline — is invisible to the current management framework.
Layer 5: The combination that nobody tested
In Morgentaler and Clift's 2024 analysis of 9,537 patients across 42 clinics, 74.23% were on hCG co-therapy with testosterone. This is the real-world standard of care for fertility-preserving testosterone treatment — nearly three-quarters of patients on combination protocols.
Here is what we know about hCG and hematocrit: men switched from TRT to hCG monotherapy show a significant decrease in hematocrit (45.27% → 44.16%). hCG stimulates endogenous testosterone production, which is pulsatile — the Leydig cells produce testosterone in a diurnal rhythm rather than the sustained supraphysiologic levels that follow an intramuscular injection.
The question writes itself: if 74.23% of real-world TRT patients are also taking hCG, and hCG monotherapy lowers hematocrit, does the combination incidentally protect against the side effect that the guidelines spend the most effort monitoring? Nobody has tested this. The most common clinical protocol — the one 74.23% of patients are actually on — has zero hematocrit data. Every threshold, every guideline, every management recommendation was developed for testosterone monotherapy in a world where three-quarters of patients are on something else.
Layer 6: The drug interaction nobody warned about
Kabha and colleagues (2025), analyzing 5,235 patients from the Clalit Healthcare database, found that concurrent SGLT2 inhibitor use with testosterone therapy produced an odds ratio of 4.85 (95% CI 3.06-7.69) for hematocrit exceeding 54%. SGLT2 inhibitors — prescribed for diabetes and heart failure, often to the same metabolically unhealthy men who receive testosterone — independently raise hematocrit through increased erythropoietin production. Combined with testosterone's hepcidin-suppressive effect, the two mechanisms compound.
This interaction appears in no testosterone management guideline. No guideline asks about concurrent SGLT2i use before initiating testosterone. No threshold adjusts for the combination. The 54% threshold was set without knowing that a commonly co-prescribed drug class nearly quintuples the risk of exceeding it.
Layer 7: The field reviewed itself and found no system
Liu and colleagues (2025), publishing in Blood Advances, conducted a systematic review of drug-induced erythrocytosis — the definitive synthesis of the field's own literature. Their conclusion: "No clear consensus among clinicians about diagnosis or management." The systematic review of the system found no system.
The Pattern Underneath
Beneath the management void, one principle keeps resolving individual puzzles: the pattern of delivery matters more than the molecule.
This is the same principle that governs every other HPG axis paradox. GnRH given continuously downregulates its receptor; given in pulses, it activates it. Kisspeptin analogs stimulate LH acutely but suppress it chronically. The molecule is identical. The signal pattern is everything.
For erythrocytosis, three lines of evidence converge on the same resolution:
Peak at 40 minutes, baseline by 6 hours. Rapid absorption and elimination produces a pulsatile testosterone pattern. Erythrocytosis rate: approximately 3% (University of Miami, 60 men).
Stimulates endogenous Leydig cell testosterone production in a diurnal pattern. Hematocrit decreases on switch from TRT (45.27% → 44.16%).
Supraphysiologic peak sustained for days after injection. Chronic hepcidin suppression, EPO set-point recalibration, continuous erythropoiesis. Erythrocytosis rate: approximately 40%.
Same molecule. Same therapeutic intent. Erythrocytosis rates spanning 3% to 40%, determined entirely by how the drug enters the body. Pulsatile delivery allows hepcidin recovery between peaks. Sustained delivery suppresses hepcidin continuously, recalibrates the erythropoietin/hemoglobin set point upward, and drives uninterrupted red cell production.
The pulsatility resolution explains why the threshold debate is unanswerable on its own terms: the risk of a given hematocrit depends on how the patient arrived there. A 52% on nasal testosterone (rare, and built slowly through mild hepcidin modulation) is not the same biological state as a 52% on biweekly IM injections (common, driven by chronic hepcidin suppression and EPO recalibration). The number is identical. The underlying physiology is not.
The Pipeline That Confirms the Mechanism and Ignores the Population
The pharmaceutical industry has developed a solution to hepcidin-mediated erythrocytosis. Rusfertide, a hepcidin mimetic, controlled erythrocytosis in polycythemia vera patients in a Phase 3 trial, maintaining hematocrit below 45% and reducing phlebotomy need for over 2.5 years. The FDA accepted the NDA with priority review in March 2026, with a PDUFA date in Q3 2026. DISC-3405, an anti-TMPRSS6 antibody that stimulates endogenous hepcidin production, is in Phase 2 for the same indication.
Both drugs directly address the mechanism that testosterone disrupts: hepcidin suppression leading to unchecked iron availability for erythropoiesis. Both are being developed exclusively for polycythemia vera. Neither has been studied in testosterone-induced erythrocytosis, despite the mechanism being identical at the iron-regulatory level.
The patient community has noticed. Forum threads already ask whether rusfertide could help men on TRT with high hematocrit. The question is mechanistically sound. The answer is: nobody knows, because nobody has asked.
What Fifty-Two Percent Means
The man at 52% is the measurement problem in a single blood draw. His number is simultaneously dangerous and safe, actionable and ignorable, a reason to stop treatment and a reason to continue it — depending on which guideline applies, which specialist reads the result, whether anyone asks about his baseline, his formulation, his co-medications, or his symptoms.
The management void has a specific architecture:
The structure of the void
The threshold was borrowed from a different disease without validation.
The intervention (phlebotomy) may cause the harm it's meant to prevent.
The outcome data suggest higher hematocrit correlates with survival, not death.
The relevant dimension (delta from baseline) appears in no guideline.
The standard protocol (TRT + hCG) has zero hematocrit data despite 74.23% real-world use.
The common co-medication (SGLT2i) nearly quintuples the risk with no guideline acknowledgment.
The pharmaceutical fix (hepcidin mimetics) exists but is being developed for the wrong population.
Every layer makes every other layer harder to resolve. You can't set a threshold without knowing whether the intervention at that threshold is safe. You can't evaluate the intervention without outcome data. You can't generate outcome data when three-quarters of your patients are on a combination you haven't studied. You can't study the combination when you haven't characterized the drug interactions that modify the risk.
The Blood Advances systematic review surveyed this landscape and returned the only honest verdict: no consensus. Not because the field hasn't tried, but because the problem is structural. The evidence needed to set a threshold requires trials in the population actually being treated, with the formulations actually being used, at the combinations actually being prescribed. Those trials don't exist. They have never existed. And the management protocols that fill the void — the thresholds borrowed from polycythemia vera, the phlebotomy extrapolated from myeloproliferative neoplasms, the dose adjustments based on expert opinion — are not evidence. They are the absence of evidence wearing evidence's clothes.
The pulsatility resolution suggests the real question isn't what number is dangerous but what delivery pattern produces risk. Sustained supraphysiologic exposure drives erythrocytosis. Pulsatile delivery does not. The formulation — not the threshold — may be the variable that matters. But that question, too, awaits the trial that no one is running.
The man at 52% deserves an answer. He doesn't have one.