Content uploaded by William Rosner
Author content
All content in this area was uploaded by William Rosner
Content may be subject to copyright.
POSITION STATEMENT: Utility, Limitations, and
Pitfalls in Measuring Testosterone: An Endocrine Society
Position Statement
William Rosner, Richard J. Auchus, Ricardo Azziz, Patrick M. Sluss, and Hershel Raff
St. Luke’s/Roosevelt Hospital Center and Columbia University College of Physicians and Surgeons (W.R.), New York, New
York 10019; Division of Endocrinology and Metabolism (R.J.A.), University of Texas Southwestern Medical Center, Dallas,
Texas 75390; Center for Androgen Related Disorders and Department of Obstetrics-Gynecology (R.A.), Cedars-Sinai Medical
Center and Department of Obstetrics-Gynecology and Department of Medicine, The David Geffen School of Medicine at
University of California, Los Angeles, Los Angeles, California 90048; Reproductive Endocrine Unit (Department of Medicine)
and Department of Pathology (P.M.S.), Massachusetts General Hospital and Harvard Medical School, Boston,
Massachusetts 02114; and Endocrine Research Laboratory (H.R.), Aurora St. Luke’s Medical Center, Division of
Endocrinology, Metabolism, and Clinical Nutrition, Medical College of Wisconsin, Milwaukee, Wisconsin 53215
Objective: The objective of the study was to evaluate the current
state of clinical assays for total and free testosterone.
Participants: The five participants were appointed by the Council of
The Endocrine Society and charged with attaining the objective using
published data and expert opinion.
Evidence: Data were gleaned from published sources via online
databases (principally PubMed, Ovid MEDLINE, Google Scholar),
the College of American Pathologists, and the clinical and laboratory
experiences of the participants.
Consensus Process: The statement was an effort of the commit-
tee and was reviewed in detail by each member. The Council of
The Endocrine Society reviewed a late draft and made specific
recommendations.
Conclusions: Laboratory proficiency testing should be based on the
ability to measure accurately and precisely samples containing
known concentrations of testosterone, not only on agreement with
others using the same method. When such standardization is in place,
normative values for total and free testosterone should be established
for both genders and children, taking into account the many variables
that influence serum testosterone concentration. (J Clin Endocri-
nol Metab 92: 405– 413, 2007)
1. Introduction
The measurement of testosterone (T) in plasma or serum,
as done in most laboratories, suffers from a number of se-
rious problems. In women and children, the lack of accuracy
and sensitivity has resulted in severely limited utility. For
men, most T assays have adequate sensitivity and reasonable
clinical utility but are relatively inaccurate. The importance
of this issue is highlighted by recent publications in The
Journal of Clinical Endocrinology & Metabolism including two
Original Articles and an Editorial (1), a Position Statement on
polycystic ovary syndrome (2), and Clinical Guidelines on
Androgen Therapy in Women (3).
The Council of The Endocrine Society, after noting that
substantial difficulties exist in the measurement of T in bi-
ological fluids, appointed a task force, consisting of the au-
thors of this position paper, to review the problem and make
recommendations based on that review. The task force
reviewed the literature, gathered data by interview and
discussion to assess current practice, evaluated profi-
ciency survey data from the College of American Pathol-
ogists (CAP), and came to consensus by both discussion
and revision of drafts of the manuscript. The manuscript
was reviewed by the Council; their comments were eval-
uated and included, if appropriate, before finalizing the
document.
The problems of sensitivity and specificity of T assays have
been addressed by extracting steroids from plasma or serum
and separating them chromatographically before subjecting
them to immunoassay. These methods are labor intensive and
expensive. Hence, high-throughput, relatively inexpensive
methods are in wide use that employ whole serum or plasma
(“direct” assays) but, for the most part, are too insensitive and
inaccurate to measure the total T (TT) concentration in the
plasma of women and children. We have the technology to
improve the accuracy and precision of T assays and must
choose these properties over simplicity and economy.
2. Background
Assays for T in plasma, and their evaluation, pose a num-
ber of challenges:
• TT concentrations in plasma vary over 3 orders of mag-
nitude depending on age, gender, and the presence of
disease.
First Published Online November 7, 2006
Abbreviations: bio-T, Bioavailable T; CAP, College of American Pa-
thologists; ED, equilibrium dialysis; FAI, free androgen index; FT, free
T; ID/GC-MS, isotope-dilution gas chromatography-MS; Kd, dissocia-
tion constant; LC/MS-MS, liquid chromatography/MS-MS; MS, mass
spectrometry; MS-MS, tandem MS; PCOS, polycystic ovary syndrome;
T, testosterone; TT, total T.
JCEM is published monthly by The Endocrine Society (http://www.
endo-society.org), the foremost professional society serving the en-
docrine community.
0021-972X/07/$15.00/0 The Journal of Clinical Endocrinology & Metabolism 92(2):405–413
Printed in U.S.A. Copyright © 2007 by The Endocrine Society
doi: 10.1210/jc.2006-1864
405
• The concentration of TT varies with time of day.
• Other steroids of similar structure and abundance in the
circulation lead to assay interference.
• Only 1–3% of T is not bound to plasma proteins, raising
questions about whether TT or free T (FT) is the most
clinically useful measure.
• Age- and gender-corrected normal ranges, using a stan-
dardized assay, are generally lacking.
• There is no universally recognized T-calibrating standard.
3. Methods for Measuring TT
The commonly used methods for measuring TT and their
strengths and shortcomings are summarized in Table 1. Fol-
lowing, we briefly detail these methods.
a. Immunoassay methods for measuring TT
RIAs and chemiluminescence immunoassays are the most
widely used methods for measuring plasma TT. These assays
are performed directly on serum or plasma or after extraction
and/or chromatography. The more labor-intensive assays
that incorporate extraction and chromatography offer sev-
eral advantages, including removal of interfering proteins,
separation of cross-reacting steroids, and use of large serum
aliquots to increase sensitivity. Such assays are more accurate
and sensitive than direct assays but still require proper
validation.
b. Mass spectrometry (MS) methods for measuring TT
MS both identifies and quantifies the analyte and, for TT,
routinely incorporates extraction and chromatography be-
fore assay (4–9); the specificity of this method has been
enhanced by tandem MS (MS-MS), which still must be val-
idated for accuracy, sensitivity, and precision.
c. Comparison of TT assay methods
The CAP administers a quality-control program that dis-
tributes blinded samples to participating laboratories and
gauges accuracy relative to others using the same method-
ology. The samples they distribute are not in plasma but in
material that allows the samples to be stable, although not
frozen, and hence more easily distributable to large numbers
of laboratories. Assay results may very well be influenced by
this artificial matrix. Table 2 shows the CAP results for three
samples of TT expected for a normal woman [no. 1; 33 ⫾ 11
ng/dl (1.1 ⫾ 0.4 nm)], a hypogonadal man or an androg-
enized woman [no. 2; 97 ⫾ 31 ng/dl (3.4 ⫾ 1.1 nm)], and a
normal man [no. 3; 465 ⫾ 81 ng/dl (16.1 ⫾ 2.8 nm)]. For no.
1, the coefficients of variation within the same methodology
performed at different laboratories ranged from 13–32%. Re-
sults for the same sample using the same method varied 2-
to 6-fold, demonstrating the unacceptable reliability of these
methods for measuring TT in normal women. Examination
of “All Instruments” for no. 1 indicates that clinical useful-
ness is severely compromised. As the concentration of T
increases (no. 2 and 3), the coefficients of variation within a
methodology decrease. However, there is still considerable
variability and lack of standardization between methods as
demonstrated by, for example, the range of values for no. 2
[45–365 ng/dl (1.6–12.7 nm)]. Approximately one third of all
the laboratories used the same instrument and about two
TABLE 1. Comparison of methods available for measuring TT in the circulation
Method Strengths Shortcomings
Direct assay by RIA, ELISA, or CLIA ● Technically simple, rapid, and ● T concentration often overestimated
relatively inexpensive ● Susceptible to matrix effects
● High throughput and fast turnaround
time
● Not standardized; results and reference
intervals are method dependent
● Can be automated ● Limited accuracy at T ⬍ 300 ng/dl
● Reference intervals in different
populations not well documented
● For RIA: generates radioactive waste
RIA after extraction and chromatography ● Extensively used, with well-
documented reference intervals in
different populations
● Labor intensive, cumbersome, time
consuming, and costly
● Requires a high degree of technical
● Relatively large serum volumes can be
used for the assay, increasing
sensitivity
expertise
● Use of organic solvents requires special
facilities and waste disposal
● Potential to assay multiple steroids
separated by the chromatography in
the same sample aliquotfacilities and
waste disposal
● Susceptible to matrix effects
● Imprecise: measurements must be
corrected for recovery
● Generates radioactive waste
● T released from steroid-binding
proteins during extraction
MS, after extraction and liquid (LC) or
gas chromatography (GC)
● Multiple steroids can be measured in
the same sample aliquot
● Relatively expensive
● Currently standardization still lacking
● Highly accurate when properly
validated
● Limited throughput and relative long
turnaround times
● Throughput comparable with RIA after
extraction and chromatography
● Derivatization steps can introduce
additional error
● Use of organic solvents requires special
facilities and waste disposal
CLIA, Chemiluminescent immunoassay. Adapted in part from Ref. 46.
406 J Clin Endocrinol Metab, February 2007, 92(2):405– 413 Rosner et al. • Testosterone Assays
thirds used one of three instruments (Table 2). The wide-
spread use of a limited number of methodologies gives
promise that standardization is achievable. There are com-
parable problems with variability and a lack of standards for
accuracy for FT.
More reliable MS methods for TT have begun to appear
(10–12), although only five of more than 1100 participating
laboratories used MS in late 2005. However, the values avail-
able from mass spectrometric measurements are consistent
with reports of larger bias at low concentrations of TT (7, 9).
We propose that the best prospect for a gold standard lies in
extraction and chromatography followed by MS or MS-MS
in which the chemical structure of the molecule measured is
identified.
Taieb et al. (7) compared 10 direct commercial immuno-
assays with isotope-dilution gas chromatography-MS (ID/
TABLE 2. Selected CAP proficiency survey results for TT (ng/dl)
Instrument No. of labs Mean/SD (ng/dl) CV (%) Median (ng/dl) Low/high value (ng/dl)
Test sample 1
All instruments 1108 32.7/11.4 34.9 31 7–100
Abbott Architect 5 26 17–32
Bayer ACS: 180 19 30.9/9.5 30.7 34 13–43
Bayer ADVIA Centaur 349 30.3/8.6 28.4 30 9–53
Beckman Acess/2 140 32.4/4.3 13.4 32 22–44
Beckman UniCel Dxl 73 31.3/5.3 17.0 31 12–43
DPC Coat-a-Count 40 27.1/4.2 15.4 26 20–36
DPC Immulite 1000 126 47.8/9.3 19.4 49 26–72
DPC Immulite 2000 9 52 31–56
DPC Immulite 2500 59 51.2/9.2 18.0 51 31–77
Roche Elecsys 1010/2010 70 25.1/7.0 27.7 24 7– 43
Roche Elecsys/E170 72 31.2/4.7 15.1 31 20–43
Tosoh AIA-Pack 12 43.8/14.0 32.0 43 17–71
Vitros Eci 83 18.4/2.7 14.5 18 13–26
MS 5 31.8/- 33 27–37
Diagnostic Sys Liquid 5 25 24–33
Diagnostic Sys Solid 5 20 8–23
Test sample 2
All instruments 1133 97.1/31.3 32.2 87 45–365
Abbott Architect 8 75 66–108
Bayer ACS: 180 23 97.6/14.1 14.5 95 64–122
Bayer ADVIA Centaur 358 96.9/10.8 11.1 97 65–130
Beckman Acess/2 150 76.8/5.8 7.6 77 62–94
Beckman UniCel Dxl 57 71.4/6.2 8.7 72 56– 87
DPC Coat-a-Count 42 79.9/8.0 10.4 76 65–92
DPC Immulite 1000 60 147.1/19.6 13.3 146 103–197
DPC Immulite 2000 133 154.3/17.0 11.0 153 120–200
DPC Immulite 2500 5 157 151–195
Roche Elecsys 1010/2010 82 69.7/10.0 14.3 68 55–105
Roche Elecsys/E170 66 81.1/7.3 9.0 81 60–102
Tosoh AIA-Pack 12 87.8/12.3 14.0 89 71–108
Vitros Eci 85 78.3/6.2 7.9 78 64–91
MS 5 68.6/6.1 8.9 69 60–77
Test sample 3
All instruments 1135 464.9/80.6 17.3 449 276–744
Abbott Architect 8 383 353–395
Bayer ACS: 180 23 439.8/42.7 9.7 440 344–509
Bayer ADVIA Centaur 359 424.1/42.6 10.0 421 328–546
Beckman Acess/2 152 402.3/21.6 5.4 402 338–473
Beckman UniCel Dxl 57 377.1/23.2 6.2 379 332–428
DPC Coat-a-Count 42 413.3/35.7 8.6 410 324–516
DPC Immulite 1000 60 550.3/60.5 11.0 546 436–673
DPC Immulite 2000 133 566.0/59.7 10.5 563 423–744
DPC Immulite 2500 5 635 546– 667
Roche Elecsys 1010/2010 81 511.8/28.7 5.6 509 451–589
Roche Elecsys/E170 69 550.6/25.7 4.7 546 501–626
Tosoh AIA-Pack 11 636.9/44.8 7.0 652 555–706
Vitros Eci 84 519.4/26.0 5.0 519 453–581
MS 5 354.4/45.4 12.8 365 281–395
The CAP distributed three different unknown samples to the indicated number of laboratories. The measured TT results were returned to
CAP and summarized by them (reproduced here in modified form with permission from CAP). Test samples 1, 2, and 3 contain, respectively,
concentrations of T similar to those in the plasma of normal women, hypogonadal men, or androgenized women and normal men. Only methods
in use in a sufficient number of reporting laboratories (No. of labs) are shown; some do not have sufficient numbers to calculate reliably mean
and
SD. The coefficient of variation (CV) gives an estimate of the variability both within and between methodologists (All instruments). The
CV (Mean/SD) is unitless but is commonly multiplied by 100 and reported as a percent. Also notice that, in some cases, the ranges (Low/high
value) are lower or higher than those shown for individual methods. That is because there are several participants in the CAP survey that do
not use one of the major methods or instruments listed. To convert to nanomolar, multiply by 0.03467.
Rosner et al. • Testosterone Assays J Clin Endocrinol Metab, February 2007, 92(2):405– 413 407
GC-MS) and also compared samples from 55 women by
ID/GC-MS with an extraction and chromatography RIA (13)
(Fig. 1). Below approximately 8.0 nm (230 ng/dl), the meth-
ods disagreed by up to 5-fold, with immunoassays generally
overestimating the T concentration. Some of the methods
were better than others, but even the best method showed up
to a 2-fold higher T concentration by immunoassay in
women. In men, seven of the 10 tested assays had highly
statistically significantly different medians compared with
ID/GC-MS. The salutary outcome is that three of the com-
mercial assays had medians not significantly different from
those measured by the MS-based method in men. Even this
salve contains an irritant because “. . . this statistical analysis
compared the differences between medians and did not ad-
dress the scatter of the results between each immunoassay
and ID/GC-MS.” Thus, concern remains that even the good
assays in men are only good on average.
Wang et al. (9) obtained results much like those in Fig. 1
by comparing six different direct immunoassays to liquid
chromatography/MS-MS (LC/MS-MS). Two of the assays
tested were the same as shown in Fig. 1 and four were unique
to this study. For T less than 150 ng/dl (⬍5.2 nmol/liter), the
values were neither analytically nor clinically useful. For
higher T concentrations, the values have some utility, but the
discrepancies among methods are unacceptable. There are
three additional studies that support these findings (14 –16)
and apparently none that contradict them.
4. Methods for Measuring FT
T circulates bound to at least two plasma proteins, SHBG
and albumin (17). That which is unbound, FT, is often con-
sidered the component that has access to the cell and results
in androgenic effects. The situation is more complicated than
that (18, 19), but as a practical matter, FT often correlates
better with the androgenic state of the patient than does TT
(20). In addition, there exists the concept of bioavailable T
(bio-T), defined as the concentration of T that is free, plus that
which is weakly bound, e.g. albumin bound. This is a widely
used measurement that we will discuss in concert with the
discussion of FT. The commonly used methods for measur-
ing FT, and their strengths and weaknesses, are summarized
in Table 3.
a. Measuring FT
An appropriate assay for TT is necessary but not suf-
ficient for the measurement of FT. Because FT is such a tiny
portion of TT, an assay that is accurate and precise when
measuring very small amounts of T is required. The indirect
measurement of FT is accomplished by adding
3
H-T to the
sample to be assayed and, after equilibrium has been attained,
separating bound from free
3
H-T, and then measuring free
3
H-T. The fraction of free
3
H-T is multiplied by the amount of
TT, obtained in a separate assay from the same plasma. The
FIG. 1. T concentrations in women (n ⫽ 55; E) and men (n ⫽ 50; F)
obtained by 10 immunoassays and ID/GC-MS. The abcissa is the T
concentration (nanomoles per liter) measured by ID/GC/MS. (To con-
vert to nanograms per deciliter, multiply by 28.8.) The y-axis is the
ratio of T concentration determined by immunoassay divided by T
concentration determined by ID/GC-MS. The vertical dotted line sep-
arates the data for men and women. [Reprinted with permission
from Taieb J, Mathian B, Millot F, Patricot MC, Mathieu E, Queyrel
N, Lacroix I, Somma-Delpero C, Boudou P 2003 Testosterone mea-
sured by 10 immunoassays and by isotope-dilution gas chromatog-
raphy-mass spectrometry in sera from 116 men, women, and children.
Clin Chem 49:1381–1395 (7).]
408 J Clin Endocrinol Metab, February 2007, 92(2):405– 413 Rosner et al. • Testosterone Assays
major potential problem with such assays is the possible pres-
ence of radiochemical impurities. For example, a 2% contam-
ination with a radioactive substance that does not bind to pro-
teins will cause an apparent doubling of the FT in which 2% of
T actually was free. If both direct and indirect methods are done
well, they should yield the same answer.
b. Data on FT measurement
The values of FT by equilibrium dialysis (ED) are influ-
enced by a number of variables, the most important of which
is the assay for T. However, the details of how ED is done (21)
as well as the population being studied all have an impact on
TABLE 3. Comparison of methods available for measuring FT, unbound T, or bio-T in the circulation
Method Strengths Shortcomings
Direct RIA ● Simple, rapid, and relatively
inexpensive
● Poor accuracy, sensitivity, and
between-laboratory comparability:
● Requires minimum technical expertise X Major biasing effects due to
● Can be automated dilution of serum samples
X Significant binding of the analog
to serum proteins
X Lack of parallelism between
measurements of serially diluted
serum samples and FT
Physical separation of protein-bound
from FT
a
● Relatively accurate (the equilibrium
dialysis assay is considered the gold
standard method for quantifying FT)
● Relatively expensive
● Technically cumbersome and difficult
X Equilibrium dialysis is influenced
䡠●Relatively sensitive and reproducible by dilution of the serum sample
X Centrifugal ultrafiltration is
subject to adsorption of
testosterone to the membrane and
difficulty with temperature control
X Both the dialysis and
ultrafiltration methods can be
affected by
3
H-labeled impurities
bound differently from T by SHBG
and/or albumin
䡠●As for all indirect measures, highly
dependent on the accuracy of the TT
assay
䡠●At this time, none of the methods are
sensitive enough to accurately measure
free T directly in women and children
Ammonium sulfate precipitation to
measure bio-T
● Technically simple ● Can be inaccurate due to:
X Use of impure
3
H-T
X Incomplete precipitation of
globulins
X Lack of uniformity of methodology
between labs
Calculation: The FAI (T/SHBG) ● Simple
● Good correlation with physical
separation measures in women
● Poor correlation with physical
separation measures in men
● Highly dependent on the accuracy and
sensitivity of the total T and SHBG
assays
Calculation: using algorithms based on
law of mass action
b
● Simple
● Excellent correlation with physical
separation measures
● Highly dependent on the accuracy and
sensitivity of the total T and SHBG
assays
● Assumptions and reference intervals
not standardized
Calculated: using empirical equations ● Excellent correlation with physical
separation measures
● Relatively sensitive
● Equations are derived from computer
modeling based on known
concentrations of T, SHBG, and FT
obtained in individual laboratories
● Hundreds to thousands of samples are
needed to generate the equations
䡠●Lack of transportability of the
equations among laboratories
Adapted in part from Refs. 36, 46, and 47. FAI, Free androgen index.
a
Separation by means of a membrane (i.e. ED) or a filter (i.e. centrifugal ultrafiltration). Either use
3
H-T and multiply by total T concentration
(obtained in separate assay) to measure percent FT or measure FT directly.
b
Requires total T and SHBG (and possibly albumin) concentrations and the Kd between T and SHBG and albumin.
Rosner et al. • Testosterone Assays J Clin Endocrinol Metab, February 2007, 92(2):405– 413 409
the results. Before citing some examples of how FT is mea-
sured, our agreement with the following should be noted:
“. . . it is clear that even the best available and scrupulously
performed measurement procedures have technical and fun-
damental limitations and that, consequently, the scientific
community will have to accept that there will remain a de-
gree of arbitrariness about the best way to measure free
hormone concentrations” (8). We believe that the degree of
arbitrariness can be small and that the best approaches ap-
proximate the FT concentration.
c. Direct and indirect measurement of FT
Both Adachi et al. (22) and Van Uytfanghe et al. (8) mea-
sured FT by assaying for T in an ultrafiltrate of plasma or
serum. The former study gave inadequate attention to the
measurement of TT, thus invalidating the use of their results
for this stringent review. The latter used a method based on
GC-MS for the measurement of TT and paid close attention
to the validation of their method. Even then, these investi-
gators concluded that further sensitivity must be attained for
the method to be useful in women.
To our knowledge, there is only one major study in which
T was measured directly when FT was separated by ED (23).
Although the work was meticulously performed, T was mea-
sured by a direct RIA using
125
I-T as tracer, and the method
was not validated against a criterion (gold) standard method.
All of the other communications in which FT was measured
by ultrafiltration or ED used the indirect method. Thus, when
FT obtained by ED is compared with other methods, we are
considering studies that use the indirect method. Otherwise,
studies that measure FT do so by calculating it, using a
surrogate for it [either the free androgen index (FAI) or
bio-T], or a direct assay for FT. There is considerable con-
fusion in the use of terms. For example, the term FT was used
when what was reported was a surrogate (24, 25) or the
calculated FAI was specified in the title and then reported as
FT in nmol/liter (26).
d. Measurement of bioavailable T (bio-T)
Bio-T is albumin bound plus FT, the fraction thought to be
available to tissues. It is measured by adding
3
H-T to serum
and precipitating SHBG bound
3
H-T with ammonium sul
-
fate. The fraction of
3
H-T not precipitated is used to calculate
bio-T by multiplying by the TT obtained in a separate assay.
Variations in precipitation and assay methodology makes
comparison of results between different communications dif-
ficult (27). Furthermore, the concept itself may be misleading
and confusing. Despite this, bio-T has been reported to cor-
relate with FT by ED from fairly (28) to extremely well (29)
and to be a useful index of some biological changes (28).
e. The Free Androgen Index (FAI)
The FAI is the unitless quotient T/SHBG and depends on
appropriate measurements for T and SHBG. Therefore, there
is a reasonable correlation between FAI and FT, particularly
in women. However, FT depends on not only the ratio but
also the absolute concentration of both T and SHBG (30).
Thus, the correlation will depend on these variables and will
be biased, depending on the concentration of T as, at lower
levels, measurements are less precise. This is illustrated by
the range of correlations obtained by different investigators
(29, 31, 32). Having measured both T and SHBG, FT should
be calculated, which is easily done using a fixed formula
in a spreadsheet or using a FT calculator (e.g.
http://www.issam.ch/freetesto.htm).
f. One-step direct assay for FT
This assay, which uses an
125
I-T analog, has been consis
-
tently found to be inaccurate (29, 32, 33), and its use is highly
questionable.
g. Calculation of FT concentration
One can also use the law of mass action to calculate con-
centrations of T that are free or bound to SHBG and albumin
(29, 34). The calculation depends on the measurement of TT,
total SHBG and total albumin, and the use of the equilibrium
dissociation constants (Kd) for the binding of SHBG and T
and albumin and T. Within rather broad limits, the concen-
tration of the low affinity binding protein, albumin, varies
too little to significantly affect FT levels (29). Although the Kd
for SHBG-T is about 1 nm, this number needs to be verified
and universally agreed on. In addition, a universal standard
for SHBG has not been agreed on. With those caveats, cal-
culation of FT is the most useful estimate of FT in plasma (29,
32) except in pregnancy (29).
Calculation of FT compares extraordinarily well with FT
measured by ED (29, 32, 35). Miller et al. compared calcula-
tion with ED in more than 400 women with a variety of
disorders and looked at five clinical subgroups (32). The
correlation coefficients between ED and calculation were
greater than 0.96; the intercepts were not different from zero,
but the slopes indicated a 20% bias between the measured
and calculated values due either to a systematic error in the
method used for ED or in the calculation. The error in the
calculation, in turn, could arise from using the wrong Kd,
errors in the measurement of SHBG, or both. Studies com-
paring calculation with ultrafiltration mostly (36), but not
always (8), show excellent correlation.
5. Uses of T Assays
For scientific, analytical purposes, T assays need to be as
accurate and reproducible as possible and as sensitive as
necessary for the job at hand. We must be able to compare
studies from multiple laboratories using different methods.
Clinically used assays must be held to the same high stan-
dard. The diagnosis and management of the hyperandro-
genic woman, or the hypoandrogenic woman or man, will
depend to a large degree on highly accurate androgen mea-
surements. In addition, because there is a diurnal variation
in plasma T that may be superimposed on variations over
smaller time intervals, samples should be multiple (three
should suffice) and be obtained between 0800 and 1000 h.
a. Evaluating adult males
The most common use of clinical T assays is to diagnose
hypogonadism in men for which almost any assay will do.
410 J Clin Endocrinol Metab, February 2007, 92(2):405– 413 Rosner et al. • Testosterone Assays
Whether the subtle decrease in plasma T in the aging male
is normal or represents hypogonadism will not be answered
with the use of almost any assay. Furthermore, the evaluation
of the risks and benefits of T replacement requires sensitive
and accurate assays. When the TT lies near the lower limit of
the normal range, a calculated FT may prove useful.
b. Evaluating adult females
The measurement of T in women is used for evaluating
states of androgen excess to both exclude androgen-produc-
ing tumors and help in the diagnosis of other hyperandro-
genic states. Most commercial assays are adequate for iden-
tifying, but not accurately quantifying, elevated TT in
women. However, these assays frequently fail to detect the
moderately androgenized patient, e.g. most patients with
polycystic ovary syndrome (PCOS) (20). FT, or one of its
surrogates, correlates better with the clinical presentation of
these patients than does TT. FT measurements may be the
most sensitive marker of hyperandrogenemia; they are above
the upper normal limit in 60 –70% of women with clinical
signs and symptoms of hyperandrogenism (20).
The influence of T on female sexual desire and sense of
well-being has received considerable attention; two studies
have shown no correlation between circulating T concentra-
tions and female sexual function (37, 38), although there is
evidence that T replacement in ovariectomized premeno-
pausal women (39) and those with hypopituitarism (40) con-
fers some benefit. The normative values for T and FT across
a woman’s life span are not adequate. Thus, the measure-
ment of serum T for the evaluation of poor libido in women
is unlikely to be informative, and we recommend that such
measurements not be used for this purpose until improved
methods are available.
c. Evaluating children
In boys, TT measurements are used during adolescence in
the evaluation of early or delayed puberty or at birth during
the evaluation of undervirilized males. In girls, TT assays are
used to assess and treat disorders of sexual development and
in the evaluation of contrasexual pubertal development. As
in women, TT determination in children should be carried
out only with assays of sufficient sensitivity and in conjunc-
tion with appropriate normative data. FT in children is of
limited value.
6. Normal Ranges for T
a. In adult males
Although the measurement of TT in normal men does not
represent a problem in sensitivity, a precise definition of the
lower limit of normal TT for adult males remains elusive.
Because the clinical presentation of hypogonadism is highly
variable, particularly in the setting of comorbid conditions,
hypogonadism is often a laboratory rather than clinical di-
agnosis. TT greater than 320 ng/dl (11.1 nm) is considered
normal (41). TT less than 200 ng/dl (6.9 nm) is diagnostic of
hypogonadism, but TT 200 –320 ng/dl (6.9 –11.1 nm) is equiv-
ocal. The agreement among platform assays is marginal in
the difficult range of 200 –320 ng/dl (6.9–11.1 nm) in which
a difference of 10% might alter clinical decisions. Although
standardization (or replacement) of platform assays with
MS-based methods holds the promise of obviating the assay-
based confusion at the lower end of the normal range, the
variability in FT consequent to alterations in SHBG must be
considered. Because T secretion is pulsatile and varies diur-
nally, more than a single measurement is sometimes required
to make a therapeutic decision.
For values in the equivocal zone, the determination of FI
or bio-T is recommended to distinguish eugonadism from
hypogonadism. An FT of 6.5 ng/dl (0.23 nm) and a bio-T of
150 ng/dl (5.2 nm) are considered the lower limits of normal
(41). Measurement of FT or bio-T does not avoid the problem
of TT assay standardization, because both use TT as part of
the measurement.
b. In adult women
The need for defining an accurate lower range for T in
women has recently become significant. Platform and con-
ventional RIAs are unreliable in this range, whereas immu-
noassay after extraction and chromatography or LC/MS-MS
appears capable of yielding meaningful data.
In constructing normal ranges, care must be taken to ex-
clude subjects with PCOS, or other forms of androgen excess.
T distributions are bimodal in families of PCOS subjects (42)
and normal ranges show a tail at the high end of the distri-
bution. T in women varies not only with the menstrual cycle
but also with age, race, and body mass index (38).
The FAI is often used as a surrogate for FT, and the FAI
correlates well with FT in women (32) but not men. Because
T production is regulated by gonadotropin feedback in men,
changes in SHBG, which alter FT concentrations, will be
compensated by autoregulation of T production but not so in
women. In addition, much circulating T in women is derived
from the peripheral conversion of adrenal dehydroepiandro-
sterone and dehydroepiandrosterone sulfate (43) that also is
not subject to feedback control. Because SHBG is present in
such large excess in women (10 –100:1), FT concentrations are
driven primarily by SHBG abundance. In addition, T excess
in women lowers SHBG concentrations, which raises the FT
concentration and contributes to the strong correlation of
1/SHBG with FT.
c. In children
The testes secrete large amounts of T during the first year
of life, but gonadal steroidogenesis is very low in both boys
and girls thereafter until the start of puberty. Consequently,
T concentrations are extremely low throughout childhood;
the measurement of T in children poses the same problems
as those in women. Consequently, assay of T in children
should use immunoassay after extraction and chromatogra-
phy or LC/MS-MS (5). One recent report indicates that der-
ivatization before LC/MS-MS improves assay characteristics
(4). Total and FT reference intervals must specify gender, age,
and Tanner stage, as has recently been done (5) because all
these variables influence T concentrations. Normative data
for infants are difficult to obtain, so historical data are used
(44, 45).
Rosner et al. • Testosterone Assays J Clin Endocrinol Metab, February 2007, 92(2):405– 413 411
7. Summary of Key Findings and Recommendations
This review demonstrates that the manner in which most
assays for TT and FT are currently performed is decidedly
unsatisfactory. The technology exists to perform accurate,
precise, and reproducible assays for T, and we should move
forward to ensure that these assays become the standard by
which all assays are validated. We have summarized our
findings in Tables 1 and 3.
• Our most salient recommendation is: laboratory profi-
ciency testing should be based on the ability to accurately
and precisely measure a sample containing a known con-
centration of T and not only on agreement with peers
using the same method. When such standardization is in
place, normative values for TT and FT should be estab-
lished taking into account all the appropriate variables, e.g.
gender, age, race, stage of puberty, time of day, etc.We
believe that this goal can be accomplished. It has been
done for cholesterol.
In the interim we offer the following recommendations to
physicians ordering and using androgen assays:
• Know the type and quality of the assay that is being used
and the properly established and validated reference in-
tervals for that assay. Reference intervals should be es-
tablished by each laboratory in collaboration with endo-
crinologists, using well-defined and characterized
populations.
• In the absence of other information, direct assays (those
performed on whole serum) perform poorly at low T con-
centrations (i.e. in women, children, and hypogonadal
men) and should be avoided. Assays after extraction and
chromatography, followed by either MS or immunoassay,
are likely to furnish more reliable results and are currently
preferred.
• Assays for T may behave differently in controls and af-
fected individuals, perhaps reflecting differences in the
endocrine milieu of patients.
• Most assays will distinguish between T concentrations
in classic hypogonadism and those in normal men.
Serum TT, preferably obtained on more than one morn-
ing sampling, is the recommended screening test for
hypogonadism.
• Assuming a high-quality assay and well-defined reference
intervals, a serum TT, preferably drawn during the early
follicular phase of the menstrual cycle, is recommended as
the initial test in seeking out androgen-producing tumors
in women.
• Calculated FT, using high-quality T and SHBG assays with
well-defined reference intervals, is the most useful, clin-
ically sensitive marker of hyperandrogenemia in women
and can be used in concert with clinical end points in the
diagnosis and follow-up of such patients.
• In the absence of pituitary insufficiency, the use of T assays
in the evaluation of sexual dysfunction or fatigue in adult
women is not supported by published evidence and is
strongly discouraged.
• In children, reference intervals must be adjusted for gen-
der, age, and stage of adolescent development and must
be specific for the assay method, until a universal standard
is available.
• FT measurements in children are of limited value. Eval-
uations of androgen excess, virilization, intersex disor-
ders, or contrasexual maturation are the only indications
for T measurement in girls. Several indications exist for T
measurements in boys, including assessment of gonadal
failure, disorders of sexual development or puberty, and
monitoring response to treatment.
Acknowledgments
Received August 24, 2006. Accepted October 30, 2006.
To whom correspondence and reprint requests should be addressed:
William Rosner, M.D., Department of Medicine, St. Luke’s/Roosevelt
Hospital Center, 1000 Tenth Avenue, AJA 403, New York, New York
10019. E-mail: wr7@columbia.edu.
Disclosures: W.R. has previously consulted for Solvay. R.J.A. received
lecture fees from Columbia Laboratories. R.A. consults for Procter and
Gamble, Merck, and Quest Diagnostics. P.M.S. consults for Diagnostic
Systems Laboratories and Diagnostic Products Corp. and received lec-
ture fees from Fujirebio, Abbott Diagnostics, Bayer Diagnostics, and
Roche Diagnostics and grant support (2005 to present) from Roche
Diagnostics. H.R. has nothing to declare.
References
1. Matsumoto AM, Bremner WJ 2004 Serum testosterone assays—accuracy mat-
ters. J Clin Endocrinol Metab 89:520 –524
2. Azziz R, Carmina E, Dewailly D, Diamanti-Kandarakis E, Escobar-Morreale
HF, Futterweit W, Janssen OE, Legro RS, Norman RJ, Taylor AE, Witchel SF
2006 Position statement: criteria for defining polycystic ovary syndrome as a
predominantly hyperandrogenic syndrome: an Androgen Excess Society
Guideline. J Clin Endocrinol Metab 91:4237– 4245
3. Wierman ME, Basson R, Davis SR, Khosla S, Miller KK, Rosner W, Santoro
N 2006 Clinical practice guideline: androgen therapy in women: an Endocrine
Society Clinical Practice Guideline. J Clin Endocrinol Metab 91:3697–3710
4. Dorgan JF, Fears TR, McMahon RP, Aronson FL, Patterson BH, Greenhut SF
2002 Measurement of steroid sex hormones in serum: a comparison of radio-
immunoassay and mass spectrometry. Steroids 67:151–158
5. Kushnir MM, Rockwood AL, Roberts WL, Pattison EG, Bunker AM, Fitzger-
ald RL, Meikle AW 2006 Performance characteristics of a novel tandem mass
spectrometry assay for serum testosterone. Clin Chem 52:120 –128
6. Rauh M, Groschl M, Rascher W, Dorr HG 2006 Automated, fast and sensitive
quantification of 17
␣
-hydroxy-progesterone, androstenedione and testoster-
one by tandem mass spectrometry with on-line extraction. Steroids 71:450 – 458
7. Taieb J, Mathian B, Millot F, Patricot MC, Mathieu E, Queyrel N, Lacroix I,
Somma-Delpero C, Boudou P 2003 Testosterone measured by 10 immuno-
assays and by isotope-dilution gas chromatography-mass spectrometry in sera
from 116 men, women, and children. Clin Chem 49:1381–1395
8. Van Uytfanghe K, Stockl D, Kaufman JM, Fiers T, Ross HA, De Leenheer
AP, Thienpont LM 2004 Evaluation of a candidate reference measurement
procedure for serum free testosterone based on ultrafiltration and isotope
dilution-gas chromatography-mass spectrometry. Clin Chem 50:2101–2110
9. Wang C, Catlin DH, Demers LM, Starcevic B, Swerdloff RS 2004 Measure-
ment of total serum testosterone in adult men: comparison of current labo-
ratory methods versus liquid chromatography-tandem mass spectrometry.
J Clin Endocrinol Metab 89:534–543
10. Thienpont LM, Van NB, Stockl D, Reinauer H, De Leenheer AP 1996 De-
termination of reference method values by isotope dilution-gas chromatog-
raphy/mass spectrometry: a five years’ experience of two European Reference
Laboratories. Eur J Clin Chem Clin Biochem 34:853– 860
11. Cawood ML, Field HP, Ford CG, Gillingwater S, Kicman A, Cowan D, Barth
JH 2005 Testosterone measurement by isotope-dilution liquid chromatogra-
phy-tandem mass spectrometry: validation of a method for routine clinical
practice. Clin Chem 51:1472–1479
12. Chang YC, Li CM, Li LA, Jong SB, Liao PC, Chang LW 2003 Quantitative
measurement of male steroid hormones using automated on-line solid phase
extraction-liquid chromatography-tandem mass spectrometry and compari-
son with radioimmunoassay. Analyst 128:363–368
13. Fiet J, Gosling JP, Soliman H, Galons H, Boudou P, Aubin P, Belanger A,
Villette JM, Julien R, Brerault JL 1994 Hirsutism and acne in women: coor-
dinated radioimmunoassays for eight relevant plasma steroids. Clin Chem
40:2296–2305
14. Stanczyk FZ, Cho MM, Endres DB, Morrison JL, Patel S, Paulson RJ 2003
412 J Clin Endocrinol Metab, February 2007, 92(2):405– 413 Rosner et al. • Testosterone Assays
Limitations of direct estradiol and testosterone immunoassay kits. Steroids
68:1173–1178
15. Boots LR, Potter S, Potter D, Azziz R 1998 Measurement of total serum
testosterone levels using commercially available kits: high degree of between-
kit variability. Fertil Steril 69:286–292
16. Sikaris K, McLachlan RI, Kazlauskas R, de Kretser D, Holden CA, Han-
delsman DJ 2005 Reproductive hormone reference intervals for healthy fertile
young men: evaluation of automated platform assays. J Clin Endocrinol Metab
90:5928–5936
17. Rosner W 1999 Sex hormone-binding globulin. In: Knobil E, Neill JD, eds.
Encyclopedia of reproduction. San Diego: Academic Press; 471– 475
18. Mendel CM 1989 The free hormone hypothesis: a physiologically based math-
ematical model. Endocr Rev 10:232–274
19. Mendel CM 1992 The free hormone hypothesis. Distinction from the free
hormone transport hypothesis. J Androl 13:107–116
20. Chang WY, Knochenhauer ES, Bartolucci AA, Azziz R 2005 Phenotypic
spectrum of polycystic ovary syndrome: clinical and biochemical character-
ization of the three major clinical subgroups. Fertil Steril 83:1717–1723
21. Westphal U 1971 Steroid-protein interactions. New York: Springer-Verlag;
1–567
22. Adachi K, Yasuda K, Fuwa Y, Goshima E, Yamakita N, Miura K 1991 Mea-
surement of plasma free steroids by direct radioimmunoassay of ultrafiltrate
in association with the monitoring of free components with [14C]glucose. Clin
Chim Acta 200:13–22
23. Sinha-Hikim I, Arver S, Beall G, Shen RQ, Guerrero M, Sattler F, Shikuma
C, Nelson JC, Landgren BM, Mazer NA, Bhasin S 1998 The use of a sensitive
equilibrium dialysis method for the measurement of free testosterone levels in
healthy, cycling women and in human immunodeficiency virus-infected
women. J Clin Endocrinol Metab 83:1312–1318
24. Harman SM, Metter EJ, Tobin JD, Pearson J, Blackman MR 2001 Longitu-
dinal effects of aging on serum total and free testosterone levels in healthy men.
Baltimore Longitudinal Study of Aging. J Clin Endocrinol Metab 86:724 –731
25. Hogervorst E, Bandelow S, Combrinck M, Smith AD 2004 Low free testos-
terone is an independent risk factor for Alzheimer’s disease. Exp Gerontol
39:1633–1639
26. Sutton-Tyrrell K, Wildman RP, Matthews KA, Chae C, Lasley BL, Brockwell
S, Pasternak RC, Lloyd-Jones D, Sowers MF, Torrens JI 2005 Sex hormone-
binding globulin and the free androgen index are related to cardiovascular risk
factors in multiethnic premenopausal and perimenopausal women enrolled in
the Study of Women Across the Nation (SWAN). Circulation 111:1242–1249
27. Emadi-Konjin P, Bain J, Bromberg IL 2003 Evaluation of an algorithm for
calculation of serum “bioavailable” testosterone (BAT). Clin Biochem 36:591–
596
28. Morley JE, Patrick P, Perry III HM 2002 Evaluation of assays available to
measure free testosterone. Metabolism 51:554–559
29. Vermeulen A, Verdonck L, Kaufman JM 1999 A critical evaluation of simple
methods for the estimation of free testosterone in serum. J Clin Endocrinol
Metab 84:3666–3672
30. Rosner W 1975 Some theoretical considerations regarding filter disk assays.
Anal Biochem 67:422–427
31. Kapoor P, Luttrell BM, Williams D 1993 The free androgen index is not valid
for adult males. J Steroid Biochem Mol Biol 45:325–326
32. Miller KK, Rosner W, Lee H, Hier J, Sesmilo G, Schoenfeld D, Neubauer G,
Klibanski A 2004 Measurement of free testosterone in normal women and
women with androgen deficiency: comparison of methods. J Clin Endocrinol
Metab 89:525–533
33. Rosner W 2001 An extraordinarily inaccurate assay for free testosterone is still
with us. J Clin Endocrinol Metab 86:2903
34. Sodergard R, Backstrom T, Shanbhag V, Carstensen H 1982 Calculation of
free and bound fractions of testosterone and estradiol-17
␣
to human plasma
proteins at body temperature. J Steroid Biochem Mol Biol 16:801– 810
35. Rinaldi S, Geay A, Dechaud H, Biessy C, Zeleniuch-Jacquotte A, Akhmed-
khanov A, Shore RE, Riboli E, Toniolo P, Kaaks R 2002 Validity of free
testosterone and free estradiol determinations in serum samples from post-
menopausal women by theoretical calculations. Cancer Epidemiol Biomarkers
Prev 11:1065–1071
36. Ly LP, Handelsman DJ 2005 Empirical estimation of free testosterone from
testosterone and sex hormone-binding globulin immunoassays. Eur J Endo-
crinol 152:471–478
37. Davis SR, Davison SL, Donath S, Bell RJ 2005 Circulating androgen levels
and self-reported sexual function in women. JAMA 294:91–96
38. Santoro N, Torrens J, Crawford S, Allsworth JE, Finkelstein JS, Gold EB,
Korenman S, Lasley WL, Luborsky JL, McConnell D, Sowers MF, Weiss G
2005 Correlates of circulating androgens in mid-life women: the study of
women’s health across the nation (SWAN). J Clin Endocrinol Metab 90:4836–
4845
39. Davis SR, van der Mooren M, van Lunsen RH, Lopes P, Ribot J, Rees M,
Moufarege A, Rodenberg C, Buch A, Purdie DW 2006 The efficacy and safety
of a testosterone patch for the treatment of hypoactive sexual desire disorder
in surgically menopausal women: a randomized, placebo controlled-trial.
Menopause 13:387–396
40. Miller KK, Biller BM, Beauregard C, Lipman JG, Jones J, Schoenfeld D,
Sherman JC, Swearingen B, Loeffler J, Klibanski A 2006 Effects of testos-
terone replacement in androgen-deficient women with hypopituitarism: a
randomized, double-blind, placebo-controlled study. J Clin Endocrinol Metab
91:1683–1690
41. Vermeulen A 2005 Hormonal cut-offs of partial androgen deficiency: a survey
of androgen assays. J Endocrinol Invest 28:28 –31
42. Sam S, Legro RS, Essah PA, Apridonidze T, Dunaif A 2006 Evidence for
metabolic and reproductive phenotypes in mothers of women with polycystic
ovary syndrome. Proc Natl Acad Sci USA 103:7030 –7035
43. Arlt W, Justl HG, Callies F, Reincke M, Hubler D, Oettel M, Ernst M, Schulte
HM, Allolio B 1998 Oral dehydroepiandrosterone for adrenal androgen re-
placement: pharmacokinetics and peripheral conversion to androgens and
estrogens in young healthy females after dexamethasone suppression. J Clin
Endocrinol Metab 83:1928–1934
44. Forest MG, Sizonenko PC, Cathiard AM, Bertrand J 1974 Hypophyso-go-
nadal function in humans during the first year of life. 1. Evidence for testicular
activity in early infancy. J Clin Invest 53:819 –828
45. Forest MG, Cathiard AM, Bertrand JA 1973 Total and unbound testosterone
levels in the newborn and in normal and hypogonadal children: use of a
sensitive radioimmunoassay for testosterone. J Clin Endocrinol Metab 36:
1132–1142
46. Stanczyk FZ 2006 Androgen measurements: methods, interpretation and lim-
itations. In: Azziz R, Nestler JE, Dewailly D, eds. Androgen excess disorders
in women. 2nd ed. Totowa, NJ: Humana Press
47. Morris PD, Malkin CJ, Channer KS, Jones TH 2004 A mathematical com-
parison of techniques to predict biologically available testosterone in a cohort
of 1072 men. Eur J Endocrinol 151:241–249
JCEM is published monthly by The Endocrine Society (http://www.endo-society.org), the foremost professional society serving the
endocrine community.
Rosner et al. • Testosterone Assays J Clin Endocrinol Metab, February 2007, 92(2):405– 413 413
