Free testosterone: three different tests with the same name

Three different tests are reported under this name, on numeric scales that differ by as much as tenfold. Before you interpret a free testosterone result, you have to know which one you got.

What it measures

The fraction of circulating testosterone not bound to a carrier protein. Measured directly by equilibrium dialysis in 146 men, the median was 1.5% of total testosterone — the commonly quoted “about 2%” is a modelled figure and runs slightly high.

The reason to measure it is straightforward: total testosterone is a mixture of a bound majority and an available minority, and when SHBG is abnormal, the total misrepresents the minority. That is when this test earns its place, and it is the only situation in which any guideline recommends it.

The three tests

Equilibrium dialysis is the reference method. Serum is dialysed against a buffer until the unbound hormone equilibrates across a membrane, and the dialysate is measured — in the best current protocols, by mass spectrometry.

Calculated free testosterone is arithmetic, not measurement. Total testosterone, SHBG and albumin are entered into an equation derived from the law of mass action. The Vermeulen equation, published in 1999, is the one in general use.

Direct analog immunoassay is a single-tube commercial assay that reports a number labelled “free testosterone, direct.” It is the one you are most likely to have received if nobody specified, and it is the one every guideline body that has addressed the question says not to use.

These are not three routes to one answer. In one clinical series, the direct analog result was about one seventh of the equilibrium dialysis value on the same samples. A single national reference laboratory publishes an adult male range of 52 to 280 pg/mL for its dialysis test and 9.3 to 26.5 pg/mL for men aged 20 to 29 on its direct analog test — the same analyte, the same laboratory, scales differing by 5.6-fold at the lower limit and 10.6-fold at the upper.

The practical consequence: a threshold of 64 pg/mL applied to a direct analog result would classify essentially every adult man as hypogonadal, because the entire direct-analog reference interval sits below it. This is not a hypothetical error. It is the single commonest misreading of this test.

Evidence: Established.

Why direct analog immunoassay should not be used

The guideline language here is unusually blunt, and unusually unanimous.

The Endocrine Society (2018): clinicians “should not use direct analog-based free testosterone immunoassays, as they are inaccurate.”

The EAU (2026): available immunoassays “are not able to provide an accurate estimation of fT; therefore, direct fT evaluation with these methods is not recommended and should be avoided.”

The VA (January 2026): measuring free testosterone by immunoassay, “also called the ‘direct analog displacement’ method,” is not preferable, “as these assays have poor accuracy, sensitivity, and between-assay comparability.”

The mechanism of the failure was described in the same 1999 paper that gave us the calculation: the direct analog values “were only a fraction of either” the dialysis or calculated value, “the fraction varying as a function of SHBG levels.” That last clause is the important one. The error is not a fixed offset you could correct with a multiplier — it is patient-specific, and it varies with exactly the variable the test was ordered to account for.

The same paper rejected the free androgen index in the same sentence: “neither aFT nor FAI is a reliable index of bioavailable T.” A study of 5,350 men later showed how badly the index can fail — for incident type 2 diabetes it produced a significant association in the opposite direction (hazard ratio 1.74, 95% CI 1.17 to 2.59) to two properly calculated free testosterone estimates (0.74, not significant, and 0.59, 0.39 to 0.91). It does not add noise; it inverts conclusions.

Evidence: Established (that these methods are unreliable).

What the calculation actually does

The Vermeulen equation is a good enough tool used with more confidence than it deserves, and most of what is said about its weaknesses is backwards.

In the cleanest head-to-head — equilibrium dialysis with mass-spectrometric quantification of the dialysate, run on samples from the European Male Ageing Study — the Vermeulen calculation overestimated measured free testosterone by a median of 19% (2.5th to 97.5th percentile: 10% under to 47% over). A newer “ensemble allosteric” model, published with the claim that its values did not differ from dialysis, was measured at roughly double the dialysis value in the same comparison (median ratio 2.05). An empirical formula was unbiased at the median but scattered ±40%.

Two things follow that contradict what is usually said.

First, the direction. Vermeulen overestimates; three independent datasets agree, including one of 2,159 male samples. If you have been told the calculation underestimates free testosterone, that is wrong.

Second, the failure mode. The received wisdom is that the Vermeulen equation breaks down at the extremes of SHBG. In that comparison it was the only one of the three whose error was not significantly related to SHBG, testosterone or albumin — while the other two were strongly SHBG-dependent. Its flaw is a near-constant proportional bias, which is a much easier flaw to live with. The formula it is usually contrasted unfavourably against is the one with the SHBG problem.

We are stating an open disagreement rather than resolving it: the authors of the allosteric model reported agreement with dialysis, and the leading independent dialysis laboratory published a twofold discrepancy. Both are in print. Neither has been settled.

One further correction worth making, because it is repeated everywhere: the claim that an abnormal albumin invalidates the calculation. In 2,159 male samples, actually measuring albumin rather than assuming it did not improve the formulas’ accuracy — “the accuracy of the formulae is not influenced by correction for serum albumin, obesity, ethnicity or gonadal status.” The dominant error in men is the binding model, not the albumin input. The caveat is reasonable at genuinely abnormal albumin — nephrotic syndrome, decompensated cirrhosis, severe malnutrition — but we have found no study that puts a number on it, so we will not.

Evidence: Promising — the calculation is usable and its bias is characterised, but no formula has been shown to reproduce measured free testosterone.

Where 64 pg/mL comes from

This number, and its unit-variants 6.4 ng/dL and 220 pmol/L, is quoted as though it were a diagnostic threshold. All three are the same single number, from one paper.

It comes from the European Male Ageing Study’s 2010 report in the New England Journal of Medicine, a random population sample of 3,369 men aged 40 to 79. The sentence it comes from reads: late-onset hypogonadism can be defined by the presence of at least three sexual symptoms together with a total testosterone below 11 nmol/L and a free testosterone below 220 pmol/L. And in the methods: free testosterone was calculated with Vermeulen’s formula.

So the number is a research case definition, not a clinical cutoff; it requires the symptoms and the total testosterone alongside it; and it is a calculated value, which means it should not be applied to a dialysis result — since the calculation runs about 19% high, the dialysis-equivalent of that cut point is nearer 185 pmol/L than 220.

The EAU is the one guideline body that has adopted it, stating in its 2026 edition that “limited evidence indicates that a threshold of 220 pmol/l (6.4 ng/dl) for fT can be considered as a reliable threshold” — with the qualifier about limited evidence in the EAU’s own words, and in the specific setting of altered SHBG. The other three decline to give a number, and say why.

Why nobody publishes a normal range

The Endocrine Society: “A harmonized reference range for FT has not been established, so reference ranges may vary considerably depending on the specific equilibrium dialysis method or the algorithm used to calculate FT.”

The VA: “there is no harmonized reference range for free testosterone, and the lower limit of the provided reference range should be used.”

That is not institutional caution. It is a description of the data. Two reference-range studies, both in healthy non-obese young men, both using equilibrium dialysis with mass spectrometry — the reference method, correctly applied, twice — disagree as follows:

Study	Population	2.5th–97.5th percentile
Boston, 2023	145 healthy non-obese men, 19–39 subgroup	415–1,274 pmol/L (120–368 pg/mL)
Ghent, 2026	1,202 White community-dwelling men; 18–39, BMI <30 subgroup	184–749 pmol/L (53–216 pg/mL)

The lower limits differ by a factor of 2.26. The explanation is that equilibrium dialysis is a method, not a standardised measurement: duration, temperature, whether the serum is diluted, and the ionic composition of the buffer all vary between laboratories, and all of them move the answer. The published standardised protocol specifies 16 hours at 37°C with undiluted serum and a plasma-mimicking buffer — every one of those is a choice another laboratory may make differently.

Interpret a free testosterone against the performing laboratory’s own interval, and only against that.

When to order it

Only when SHBG is likely to be abnormal, or when the total testosterone sits near the lower limit. Both the Endocrine Society and the EAU frame it that way; the AUA has no graded statement on free testosterone at all.

The conditions that call for it, from the Endocrine Society’s own table: lowering SHBG — obesity, diabetes, glucocorticoids, some progestins and androgenic steroids, nephrotic syndrome, hypothyroidism, acromegaly, SHBG gene polymorphisms; raising SHBG — aging, HIV, cirrhosis and hepatitis, hyperthyroidism, some anticonvulsants, estrogens, SHBG gene polymorphisms.

One item on that list deserves a caveat the guidelines do not give it. The best-powered studies of the common SHBG variants find that they move SHBG and total testosterone substantially while leaving measured free testosterone intact — in one study of 999 men with dialysis-measured free testosterone in a subset, the variants shifted SHBG by up to a quarter and total testosterone by up to a fifth, with “no clear effects on measured free T.” The genotype changes what the total testosterone reads. It does not appear to change what the man has.

Two examples make the case better than the list does.

In 51 men with epilepsy on enzyme-inducing anticonvulsants, SHBG was higher and total testosterone was higher than in controls — while free testosterone was lower. Total and free moved in opposite directions in the same men.

In 90 obese boys compared with 90 age-matched controls, total testosterone and SHBG were both significantly lower, while median free testosterone was comparable to controls, with normal pubertal progression. The authors’ conclusion was that in obese boys free testosterone “is a better marker of androgen activity than TT.”

One counterexample, because it cuts against the tidy rule: in 80 obese men with biopsy-proven steatohepatitis, total testosterone was not significantly different from controls while calculated free testosterone was significantly lower. “Liver disease raises SHBG” is not one phenotype.

What high and low values mean

There is less to say here than for total testosterone, because the differential is almost entirely the total-testosterone differential seen through the SHBG lens.

A low free with a normal total points to high SHBG — aging, liver disease, enzyme-inducing anticonvulsants — or to a total that is normal only by population standards. Thyrotoxicosis belongs in a different box: it raises SHBG and total testosterone sharply and leaves free testosterone where it was, so it produces a misleadingly high total rather than a low free.

A normal free with a low total points to low SHBG, most often obesity and insulin resistance. This is the pattern behind the common reassurance that obese men are not really hypogonadal, their SHBG is merely low — and that reassurance is only half right. Obesity also lowers directly measured free testosterone: by about 14% at a BMI of 30 and 22% at 35, against a reference BMI of 22. In the European Male Ageing Study, men with a BMI of 30 or above had free testosterone lower by 53.7 pmol/L with LH unchanged or reduced, a pattern the investigators read as genuine hypothalamic-pituitary dysfunction rather than a measurement artefact. A preserved free testosterone does not by itself rule out real central hypogonadism.

The exception worth flagging: free testosterone falls with age roughly 2.5 times faster than total testosterone, because SHBG is rising at the same time. In the Massachusetts Male Aging Study, total testosterone declined about 0.8% per year cross-sectionally while free and albumin-bound testosterone declined about 2%, with SHBG rising 1.6%. Directly measured, free testosterone falls about 4.5 pmol/L per year.

How it relates to treatment

Two things are worth knowing, and the second is not widely appreciated.

Free testosterone does not confer stability. The study that measured biological variation most carefully concluded that “the biological variation parameters of calculated bioavailable testosterone and calculated free testosterone confer no advantage over total testosterone.” In a separate week-to-week series, 10 of 16 men crossed between the eugonadal and hypogonadal ranges on repeat bioavailable testosterone sampling. Repeat measurement is as necessary here as it is for the total.

Calculated free testosterone is least trustworthy in men on testosterone. Using dialysis-measured free fractions from the Testosterone Trials, treatment raised the percent free testosterone even though SHBG did not change — and small changes in estradiol moved the free fraction more than changes in testosterone or DHT did. A calculation that takes only total testosterone, SHBG and albumin as inputs cannot see that. The men most often monitored with calculated free testosterone are the men in whom the calculation is weakest.

A note on bioavailable testosterone

Bioavailable testosterone is free plus albumin-bound — the fraction not held by SHBG. It can be measured by ammonium sulfate precipitation or calculated, and the calculated version tracks the measured one closely.

No guideline prefers it, and the comparative data do not support preferring it. In 1,072 men, total testosterone outperformed every calculated index for identifying biochemical hypogonadism overall — although in the borderline band of 7.5 to 12 nmol/L the ordering inverted and the free testosterone estimates did better, which is precisely the quantitative argument for ordering free testosterone when the total is equivocal. In 51 older men, total and bioavailable testosterone correlated poorly (r = 0.17 to 0.19, not significant) and the authors judged that bioavailable testosterone “reflected gonadal function poorly.”

Evidence: Limited.

Related

The Lab Library · SHBG · Total testosterone · How testosterone should actually be tested

Questions patients ask