The Journal of Clinical Endocrinology & Metabolism
Copyright 1999 by The Endocrine Society
A Critical Evaluation of Simple Methods for the Estimation of Free Testosterone in Serum
ALEX VERMEULEN, LIEVE VERDONCK, AND JEAN M. KAUFMAN
Laboratory for Hormonology and Department of Endocrinology, University Hospital Ghent, 9000Ghent, BelgiumABSTRACT
bioavailable T correlated significantly with the AFTC concentration;
The free and nonspecifically bound plasma hormone levels gener-
AFTC and FT values were almost identical under all conditions stud-
ally reflect the clinical situation more accurately than total plasma
ied, except during pregnancy. Values for aFT, however, were only a
hormone levels. Hence, it is important to have reliable indexes of these
fraction of either AFTC or FT, the fraction varying as a function of
fractions. The apparent free testosterone (T) concentration obtained
SHBG levels. Also, the FAI/AFTC ratio varied as a function of the
by equilibrium dialysis (AFTC) as well as the fraction of serum T not
SHBG levels, and hence, neither aFT nor FAI is a reliable index of
precipitated by 50% ammonium sulfate concentration (non-SHBG-T;
SHBG, sex hormone-binding globulin), often referred to as bioavail-
The FT value, obtained by calculation from T and SHBG as de-
able T, appear to represent reliable indexes of biologically readily
termined by immunoassay, appears to be a rapid, simple, and reliable
available T, but are not well suited for clinical routine, being too time
index of bioavailable T, comparable to AFTC and suitable for clinical
consuming. Several other parameters have been used without com-
routine, except in pregnancy. During pregnancy, estradiol occupies a
plete validation, however: direct immunoassay of free T with a labeled
substantial part of SHBG-binding sites, so that SHBG as determined
T analog (aFT), calculation of free T (FT) from total T and immuno-
by immunoassay overestimates the actual binding capacity, which in
assayed SHBG concentrations (iSHBG), and the free androgen index
pregnancy sera results in calculated FT values that are lower than
(FAI ϭ the ratio 100T/iSHBG). In the view of substantial discrepan-
AFTC. The nonspecifically bound T, calculated from FT, correlated
cies in the literature concerning the free or bioavailable T levels, we
highly significantly with and was almost identical to the values of
compared AFTC, FT, aFT, FAI, and non-SHBG-T levels in a large
non-SHBG-T obtained by ammonium sulfate precipitation, testifying
number of sera with SHBG capacities varying from low, as in hirsute
to the clinical value of FT calculated from iSHBG. (J Clin Endocrinol
women, to extremely high as in hyperthyroidism. All these indexes of
Metab 84: 3666 –3672, 1999)
TESTOSTERONE(T)circulatesinplasmanonspecifically Several methods have been used to estimate free or bio-
bound to albumin, specifically bound to sex hormone-
available T in plasma. FT can be measured by equilibrium
binding globulin (SHBG), and, in a small percentage, un-
dialysis [apparent FT concentration (AFTC)] (2), the method
bound. Although the free and nonspecifically bound hor-
of choice for measurement of the free fraction of steroids in
mone fraction in plasma (commonly referred to as the
vivo (3). Alternatively, the non-SHBG-bound, biologically
bioavailable fraction) reflects only partially the hormone
readily available fraction may be obtained by precipitation of
available at the cellular level in specific tissues where partial
SHBG-bound T with ammonium sulfate (non-SHBG-T) (4, 5).
dissociation of the specific steroid-protein complex may oc-
As both are rather time-consuming procedures, many re-
cur and inactive prohormones may be converted intracellu-
searchers use an indirect parameter of FT, the free androgen
larly to active hormones, there is good evidence that this
index (FAI), which is obtained as the quotient 100 T/SHBG
fraction reflects more accurately the clinical situation than
(6). FT can also be estimated directly by an immunoassay
method involving an analog ligand (aFT). The latter tech-
T bound to albumin represents the nonspecifically bound
nique is often considered the easiest and fastest method for
T fraction (1) and is linearly related to free T (FT): AT ϭ KaCa
measuring FT (3), but the values obtained with this analog
ϫ FT, where AT is the albumin-bound T, Ka is the association
ligand immunoassay are substantially lower than values ob-
constant of albumin for T, Ca is the albumin concentration,
and FT is the free T fraction. The sum of FT and AT (i.e. the
At equilibrium, binding of T to plasma proteins can be
non-SHBG-bound T) is for convenience often referred to as
represented by Eq I: T ϭ FT ϩ P1T ϩ P2T ϩ P3T. . . . . ϩPnT
bioavailable fraction in plasma. As the total serum T con-
(Eq I). For each protein, applying the law of mass action, we
centration is subject to variations in the concentration of the
have: [FT] ϩ [P] º [PT] or [FT] ϭ ([PT]/(K ϫ [P])), and in
binding proteins, it is not a reliable index of bioavailable T.
the presence of several binding proteins: [FT] ϭ ([P
For example, in hyperthyroidism T levels are high, whereas
bioavailable T levels are normal (2). Hence, there is a clinical
need for a reliable index of bioavailable androgens.
1T], [P2T], . . . [PnT] are the concentrations of T
bound to proteins 1, 2, . . . n, respectively; K1, K2, . . . Kn arethe association constants of proteins 1, 2, . . . n, respectively,
Received April 5, 1999. Revision received July 7, 1999. Accepted July
for T; and [P1], [P2], . . . [Pn] are the free binding sites on
protein 1, 2, . . . n, respectively. As the binding capacity of
Address all correspondence and requests for reprints to: Dr. J. M.
Kaufman, University Hospital, 9K12 IE Endocrinology, De Pintelaan
albumin is very high with respect to the concentration of T,
185, 9000 Ghent, Belgium. E-mail: jean.kaufman@rug.ac.be.
the following equation holds true for the ratio of albumin-
bound T (AT) to unbound T: (AT/FT) ϭ KaCa or AT ϭ (FT ϫ
T at 37 C and a value of 3.6 ϫ 104 L/mol for that of albumin for T (10);
this value is corrected for serum dilution, and AFTC in undiluted serum
aCa) (Eq III), where Ka equals 3.6 ϫ 104 L/Mol (9, 10); Ca is
the albumin concentration, i.e. Ϯ 43 g/L or (mol wt, 69,000) Ϯ
is then calculated. The interassay CV is 7.8%. The SHBG binding ca-pacities thus calculated from AFTC are further referred to as cSHBG.
6.2 ϫ 10Ϫ4 mol/L; and KaCa is Ϯ 22.
The same second degree equation (introduction; Eq IV) is used for
The only other protein that binds T is SHBG; binding of T
calculation of FT from serum total T and serum iSHBG measured by
to transcortin or orosomucoid is negligible (2). As the binding
IRMA. The FAI is calculated from total T and iSHBG: FAI ϭ (100 ϫ
of other steroid hormones normally present in plasma can be
T)/SHBG (Eq V), with both T and SHBG expressed in nanomoles per L(6).
omitted from the calculation (9), it follows: FT ϭ ([T] Ϫ (N ϫ
Direct estimation of serum free T by an analog ligand RIA (aFT) was
[FT]))/(Kt{SHBG Ϫ [T] ϩ N[FT]}) (Eq IV), where Kt is the
performed using a commercial kit from Diagnostic Products (Los An-
association constant of SHBG for T, and N ϭ K
geles, CA); the binding capacity of SHBG calculated from aFT using the
yields a second degree equation that can be solved either for
above-mentioned second degree equation is further referred to as
Non-SHBG-T was obtained from serum total T and determination of
Several immunoassay methods are available for measure-
the non-SHBG-bound T fraction by ammonium sulfate precipitation, the
ment of serum SHBG levels. If the concentration of immu-
method involving incubation with a tracer dose of [3H]T at 37 C, and
noassayable SHBG (iSHBG) is a reliable measure of SHBG
precipitation of SHBG-bound hormone with ammonium sulfate at a final
binding capacity (11), a reliable value of FT can thus easily
concentration of 50%, followed by centrifugation and counting of ra-dioactivity in the supernatant (5).
be calculated. Conversely, when the AFTC is determined byequilibrium dialysis, the SHBG binding capacity (cSHBG)
It is surprising that whereas calculation of free T from total
Sera from men were obtained from a randomly selected subgroup of
T and iSHBG is a simple and rapid procedure, the reliability
sera from healthy, ambulant men participating in a population study onthe influence of age on plasma T levels.
of the calculated FT has never been extensively studied by
Sera from postmenopausal women were obtained from women con-
comparing these values to those obtained by dialysis (AFTC).
sulting the menopause clinic for check-up. Pregnancy sera were ob-
Indeed, So¨dergaard et al. (11) calculated FT from SHBG and
tained in third trimester pregnancy. Sera were also obtained from hy-
albumin concentrations, but did not validate the method,
perthyroid subjects; hyperthyroidism was confirmed on the basis of
whereas Wilke and Utley (7) compared calculated FT values
suppressed TSH with elevated FT4 and FT3 levels. Finally, sera were also
obtained from a small group of women investigated for mild clinical
in women to data obtained by direct analog ligand immu-
noassay of FT (aFT), but not to AFTC. The manufacturer ofa kit for aFT measurement claims that aFT corresponds to
0.42 AFTC ϩ 9.8 pg/mL (r ϭ 0.67) in males and to 0.79 AFTC
ϩ 1.07 pg/mL (r ϭ 0.75) in women, but values obtained in
Correlation of results obtained by different methods was estimated
using the method of least square regression.
clinical routine appear to be substantially lower (7, 8).
It is thus not surprising that there are substantial discrep-
ancies in the literature concerning the free or non-SHBG-bound bioavailable T levels obtained by a variety of only
partially validated methods, as has recently been illustrated
In a first series of experiments we compared the FT values
(8). We decided therefore to compare the AFTC values, gen-
calculated from serum T using the iSHBG and the actual
erally considered as the index of choice for evaluation of free
measured albumin serum concentration to the AFTC deter-
T levels, to calculated FT and aFT levels as well as to the FAI
mined by equilibrium dialysis in a group of ambulant men
in sera from subjects with normal, low, and high SHBG
(n ϭ 28), aged 25– 80 yr, with serum T concentrations varying
binding capacities, respectively. Finally, we also compared
the non-SHBG-T obtained by the ammonium sulfate precip-
As can be seen in Fig. 1a, AFTC and FT values differed very
itation technique with the FT levels as well as the nonspe-
little; the mean values (Ϯsem) were 330 Ϯ 36.4 and 332 Ϯ 37.1
cifically bound T levels, calculated from T and iSHBG (i.e.,
pmol/L for AFTC and FT (FT ϭ 1.002 AFTC ϩ 0.877 pmol/
L), respectively, with a correlation coefficient of 0.987.
As calculation of FT from T and iSHBG requires an ad-
Materials and Methods
ditional measurement of the serum albumin concentration,
we studied the influence of variations of the albumin con-centration on the FT values. In a series of 30 ambulant sub-
Total serum T was measured by RIA using a commercial kit (Bio-
source Technologies, Inc., Fleurus, Belgium). Concentrations are given
jects, the mean albumin concentration was 6.29 Ϯ 0.4 (sd)
in nanomoles per L; for conversion to nanograms per dL, multiply by
10Ϫ4 mol/L (43.4 Ϯ 0.26 g/L), with as extremes 5.62 and
28.84. The interassay coefficient of variation (CV) is below 10% for the
7.01 ϫ 10Ϫ4 mol/L (38.8 and 48.4 g/L). Hence, we decided
to study the influence on FT of a variation in albumin con-
Serum concentrations of SHBG were measured by immunoradio-
metric assay (IRMA), using a commercial kit (Orion Dignostica, Espoo,
centration ranging from 5.8 –7.2 ϫ 10Ϫ4 mol/L (i.e., 40, 45,
Finland) and are referred to as iSHBG; the interassay CV is 8%.
and 50 g/L), respectively. In subjects (n ϭ 30) with an iSHBG
In our reference procedure for estimation of free T, AFTC is obtained
concentration varying between 13.3–91.6 nmol/L, the mean
from serum total T and the free T fraction determined by equilibrium
(Ϯse) FT was 322 Ϯ 37.1 pmol/L at the actual albumin
dialysis on diluted serum at 37 C with use of [3H]T as described pre-
concentration, 340 Ϯ 40.9 pmol/L, assuming an albumin
viously (2, 9); the SHBG binding capacity is calculated from AFTC usingthe above-mentioned second degree equation (see introduction; Eq IV),
concentration of 40 g/L, 320 Ϯ 37.8 pmol/L assuming an
taking a value of 1 ϫ 109 L/mol for the association constant of SHBG for
albumin concentration of 45 g/L, and 303 Ϯ 35.4 pmol/L
FIG. 1. Comparison of AFTC determined by equilibrium dialysis with FT calculated from T and SHBG by immunoassay (iSHBG) in a) 28ambulant men, aged 25– 80 yr, using the actual albumin concentration in the calculation of FT and AFTC; b) the same 28 men assuming a fixedalbumin concentration of 43 g/L (6.2 ϫ 10Ϫ4 mol/L) in the FT calculation; c) 32 postmenopausal women (assuming a fixed albumin concentrationof 43 g/L); d) 18 patients of either sex with hyperthyroidism (assuming a fixed albumin concentration of 43 g/L); and e) 16 serum samples obtainedduring third trimester pregnancy, using the actual albumin concentration in the FT calculation.
assuming a concentration of 50 g/L albumin. Using a value
(40 g/L albumin) to 7.80 nmol/L (50 g/L albumin). In view
of 3.6 ϫ 104 L/mol for the association constant of albumin for
of the relatively unimportant changes in FT, when the albu-
T, the calculated albumin-bound T varied from 7.14 nmol/L
min concentration varies by as much as 25%, we concluded
that for routine purposes FT could be calculated assuming an
albumin concentration of 43 g/L (6.2 ϫ 10Ϫ4 mol/L) if one
In the next series of experiments we compared the aFT
is not dealing with sera from patients with marked abnor-
levels as measured directly by analog ligand RIA to AFTC
malities in plasma protein composition, such as in nephrotic
levels. In a group of 28 men, mean (Ϯsem) aFT values (65.5 Ϯ
syndrome or cirrhosis of the liver, or with sera obtained
7.28 pmol/L) were only a fraction (one fifth) of AFTC values
during pregnancy, in which cases the actual albumin con-
(303 Ϯ 34.0 pmol/L; aFT ϭ 0.186 AFTC ϩ 4.38 pmol/L),
centration should be taken into account. As expected, the use
albeit there was a significant correlation between these val-
of a fixed albumin concentration in the calculations did not
ues (r ϭ 0.937; Fig. 2a). In sera from women (n ϭ 8), aFT
affect the observed correlation between FT and AFTC con-
(mean Ϯ sem, 4.85 Ϯ 0.832 pmol/L) was around 30% of AFTC
centrations; FT values were almost identical to AFTC levels
values (15.95 Ϯ 4.27 pmol/L). Our data furthermore suggest
(FT ϭ 1.008 AFTC Ϫ 0.632 pmol/L) with a correlation co-
that there exists a positive correlation (P Ͻ 0.05) between the
specific binding capacity for T (i.e. cSHBG) and the aFT/
A similar study was performed with sera from 32 post-
menopausal women with T concentrations ranging from
Using the aFT values to calculate the SHBG capacity
461-1553 pmol/L. The FT concentrations (mean Ϯ sem,
(aSHBG) in a normal population (n ϭ 34; T ranging from
11.4 Ϯ 1.04 pmol/L) were almost identical to AFTC con-
0.312– 46.7 nmol/L), we obtained a mean (Ϯsem) value of
centrations (11.1 Ϯ 1.04 pmol/L; FT ϭ 0.949 AFTC ϩ 1.14
2.75 Ϯ 0.16 ϫ 10Ϫ7 mol/L for aSHBG compared to 5.5 Ϯ
pmol/L; r ϭ 0.966; Fig. 1c). To assess whether the close
0.43 ϫ 10Ϫ8 mol/L cSHBG by the dialysis method (aSHBG ϭ
correlation between FT and AFTC persisted in subjects
2.1331 ϫ cSHBG ϩ 156 nmol/L; r ϭ 0.638).
with high SHBG, we performed a similar study using sera(n ϭ 18) from patients of either sex with hyperthyroidism
(SHBG, 41–204 nmol/L). Again, similar values were ob-
We next evaluated the reliability of the FAI (FAI ϭ 100
tained by both methods; mean (Ϯsem) FT and AFTC were
T/iSHBG) as a parameter of bioavailable T. This parameter,
142 Ϯ 30.5 and 146 Ϯ 31.6 pmol/L, respectively (FT ϭ 0.951
obtained in 28 healthy subjects, was highly significantly cor-
AFTC ϩ 3.16 pmol/L; r ϭ 0.982; Fig. 1d). A similar study
related with AFTC (r ϭ 0.848; FAI ϭ 0.132 AFTC ϩ 7.273; Fig.
was performed on sera from a small group of women with
3a). However, the ratio FAI/AFTC varied from 0.12– 0.26,
mild clinical hyperandrogenism (n ϭ 12); the mean (Ϯse)
indicating that in the individual case FAI is a rather unre-
AFTC and FT were 25.0 Ϯ 3.12 and 22.9 Ϯ 2.77 pmol/L,
liable index of bioavailable T. In 18 hyperthyroid subjects, the
respectively (FT ϭ 0.850 AFTC ϩ 1.56 pmol/L), with a
correlation coefficient was 0.946 (Fig. 3b), whereas the FAI/
AFTC ratio varied between 0.17– 0.39. The FAI/AFTC ratio
Finally, we compared FT levels to AFTC levels obtained
is negatively correlated with the number of free binding sites
for third trimester (weeks 24 –37) pregnancy sera (n ϭ 16). As
on SHBG (cSHBG Ϫ cSHBG-T; Fig. 3c).
the mean albumin concentration was only 32 Ϯ 1 g/L (4.6 Ϯ0.14 ϫ 10Ϫ4 mol/L), we used the actual albumin concentra-
tion in the calculation. The mean (Ϯsem) value for AFTC
As iSHBG concentrations are not necessarily identical to
(14.6 Ϯ 1.73 pmol/L) was significantly higher than that for
the concentrations of binding sites (cSHBG), we compared
FT (10.06 Ϯ 1.39 pmol/L; FT ϭ 0.713 AFTC Ϫ 0.217 pmol/L;
iSHBG to cSHBG values calculated assuming an albumin
concentration of 43g/L. In 28 normal men, the mean (Ϯsem)
FIG. 2. a, Comparison of AFTC determined by equilibrium dialysis (fixed albumin concentration) and aFT determined by analog-tracerimmunoassay in 28 ambulatory men. B, The ratio of aFT/AFTC as a function of the SHBG binding capacity calculated from AFTC (cSHBG)in 34 normal subjects of both sexes.
FIG. 3. Comparison of FAI and AFTC (fixed albumin concentration) in 28 ambulant men (a) and 18 hyperthyroid subjects of either sex (b) andthe correlation between the FAI/AFTC ratio (AFTC in pmol/L) and the free binding sites on SHBG (mmol/L) (c). The free binding sites arecalculated as the difference between the molar concentration of cSHBG and the molar concentration of SHBG-bound T: cSHBG– cSHBG-T.
cSHBG was 56.1 Ϯ 5.06 nmol/L compared to 48.04 Ϯ 4.48
specifically bound T (i.e. non-SHBG-bound T) in 24 subjects
nmol/L for iSHBG (SHBG ϭ 1.09 iSHBG ϩ 3.64 nmol/L; r ϭ
of either sex with T concentrations varying between 0.485–
0.963). In our series of hyperthyroid subjects (n ϭ 18), the
25.2 nmol/L. As shown in Fig. 4, the correlation coefficient
mean iSHBG concentration was 96.6 Ϯ 11.58 nmol/L in com-
for non-SHBG-T vs. FT was 0.974, with non-SHBG-T corre-
parison with a cSHBG of 93.9 Ϯ 10.98 nmol/L (0.962 cSHBG
sponding to approximately 20 times FT, whereas the non-
ϩ 6.238 nmol/L; r ϭ 0.912). A close correlation between
specifically bound T calculated from T and iSHBG, assuming
iSHBG (mean Ϯ sem, 54.25 Ϯ 5.30 nmol/L) and cSHBG
a fixed albumin concentration of 6.12 ϫ 10Ϫ4 mol/L, was Ϯ23
(49.96 Ϯ 6.32 nmol/L) was also observed in sera from women
times FT, with a similar correlation coefficient as for FT,
with clinical hyperandrogenism (r ϭ 0.977; iSHBG ϭ 0.820
nonspecifically bound T being a multiple of FT.
cSHBG ϩ 13.273 nmol/L), whereas in a group of 32 post-menopausal women, at low T levels the correlation coeffi-
Discussion
cient was only 0.806. In the third trimester pregnancy sera,
Measurement of AFTC by equilibrium dialysis at 37 C is
for which AFTC values were higher than FT values, cSHBG
probably the physiologically most representative method
values calculated from AFTC were correspondingly signif-
and the most exact for estimating free T (3) if care is taken that
icantly lower (mean Ϯ sem, 161.7 Ϯ 7.09 nmol/L) than iSHBG
the labeled tracer T used for measurement of the FT fraction
is highly purified. Indeed, impurities that do not bind toSHBG might significantly increase the apparent FT fraction.
An alternative to the use of labeled T is direct measurement
As at least part of the albumin-bound T is bioavailable, we
of T in the dialysate, provided the necessary very high sen-
compared non-SHBG-T determined by the ammonium sul-
sitivity of the RIA for T can be achieved (12). As the asso-
fate precipitation technique to FT and to the calculated non-
ciation constant (Kt) of SHBG for T (and hence the FT con-
ues very close, if not identical, to those obtained by equilib-rium dialysis (AFTC).
Whereas our data show that within the physiological
range of 40 –50 g/L (5.8 –7.2 ϫ 10Ϫ4 mol/L), the albuminconcentration does not significantly affect FT values, itshould be realized that this is only valid for these physio-logical concentrations. Moreover, there is good evidence thatat least part of the albumin-bound T might be bioavailable(1). Hence, when the albumin concentration is expected todeviate significantly from normalcy, the actual albumin con-centration should be determined, and FT and albumin-bound T calculated accordingly.
Such a situation exists during pregnancy. In our third
trimester sera, the mean albumin concentration was 32 Ϯ 1g/L. Hence, the actual albumin concentration had to be usedto calculate FT and cSHBG, and although the AFTC is higherthan that in nonpregnant women, the nonspecifically bound
FIG. 4. Correlation between non-SHBG-T as determined by the am-
T concentration is in the normal range. Bammann et al. (15)
monium sulfate precipitation method and FT calculated from T and
as well as Wilke and Utley (7) observed increased AFTC as
iSHBG in subjects of either sex covering a broad range of T levels.
well as FT levels in pregnancy serum, although, as theypointed out, increased androgenicity is not a normal clinical
centration) varies with temperature, strict control of the
feature of pregnancy. The normal concentrations for the bio-
temperature is required when applying these dialysis tech-
available, albumin-bound T may explain the absence of vir-
niques, even though according to So¨dergard and co-workers
ilization in the presence of increased FT and AFTC.
(11), a variation in Kt by as much as 30% would influence FT
As to the observed differences between AFTC and FT
to only a moderate degree. In any case, equilibrium dialysis
concentrations, and thus between cSHBG and iSHBG con-
is a rather laborious and time-consuming procedure, hence
centrations in pregnancy serum, this is the consequence of
the occupation of a substantial fraction of the SHBG-binding
Direct measurement of the FT concentration by an analog
sites by estradiol. Indeed, AFTC as determined by dialysis is
ligand immunoassay procedure is an attractive and simple
dependent upon the number of binding sites available for T.
alternative. Our study indicates, however, that the direct
In the presence of competing steroids in concentrations cor-
measurement of aFT by analog ligand RIA, although show-
responding to a substantial fraction of the T concentration,
ing a generally good correlation with AFTC, is not a reliable
cSHBG will be significantly lower than iSHBG, as the latter
index of FT; the aFT represents a variable fraction (20 – 60%)
measures all SHBG molecules regardless of whether they are
only of AFTC. This is confirmed by the fact that calculation
available for T binding. Knowing the estradiol concentration
of the SHBG binding capacity from aFT (aSHBG) yields val-
as well as the association constants of estradiol for albumin
ues that are multiples of all values previously reported in the
and SHBG, the concentration of estradiol bound to SHBG can
literature. Similar findings were reported by Wilke and Utley
be calculated. For term pregnancy, with an estradiol con-
(7) as well as by Rosner (8). Our data show, moreover, that
centration around 20 ng/ml (73.4 nmol/L), it can be calcu-
the aFT/AFTC ratio is SHBG dependent. Winters et al. (13)
lated that about 50 nmol/L SHBG is occupied by estradiol
and not available for T binding; this explains why the cal-
The FAI also appears not to be a reliable index of FT.
culated cSHBG is significantly lower than the iSHBG value
Indeed, the ratio FAI/AFTC, which should be constant if FAI
measured by IRMA, whereas FT is falsely lower than AFTC
reflects AFTC, varies in fact by as much as a factor of 2.5.
as a consequence of inclusion of the binding sites actually
Combining Eq III, IV, and V (see introduction and Materials
occupied by estradiol in the calculation of FT. and Methods), one obtains: FAI/FT ϭ (100[(SHBG Ϫ SHBG-T)-
The excellent correlation of non-SHBG-T with FT and with
K ϩ N])/SHBG. From the latter equation it follows that the
the calculated nonspecifically bound T, which is a multiple
FAI/AFTC ratio is correlated to the number of free binding sites
of FT, respectively, is a strong argument in support of the
on SHBG (Fig. 3c). The FAI/AFTC ratio will be high when the
validity of the calculated nonspecifically bound T as a pa-
number of occupied binding sites is small related to the SHBG
rameter of the bioavailable fraction of T. Non-SHBG-T mea-
binding capacity (e.g. in women), whereas the ratio will, con-
sured by ammonium sulfate precipitation was around 20
versely, be low when a substantial proportion of the binding
times the FT, whereas the calculated nonspecifically bound
sites is occupied (e.g. in adult men). Based on a similar calcu-
T, using an association constant of albumin for T of 3.6 ϫ 104
lation, Kapoor et al. (14) came to the conclusion that the FAI is
L/mol, was around 23 times the FT. The value for the as-
sociation constant, which is obtained using pure human al-
The overall excellent correspondence between FT and
bumin, might be slightly lower in serum due to the presence
AFTC levels indicates that FT is a reliable index of unbound
T. Calculation of FT from total T and immunoassayable
In any case, the calculated nonspecifically bound T reliably
SHBG represents a simple and rapid method that under all
conditions studied, except for pregnancy serum, yielded val-
In conclusion, this study shows that neither aFT nor FAI
is a reliable parameter of FT. The similar values of FT and
drogens and sex hormone binding globulin in the evaluation of hirsute pa-
AFTC as well as iSHBG and cSHBG obtained under various
7. Wilke TJ, Utley DJ. 1987 Total testosterone, free androgen index, calculated
physiopathological conditions, provided no competing ste-
free testosterone and free testosterone by analog RIA compared in hirsute
roids are present in a high concentration, show that the
women and in otherwise normal women with altered binding of sex hormone
calculated FT is a reliable index of FT, that calculated non-
binding globulin. Clin Chem. 33:1372–1375.
8. Rosner W. 1997 Errors in the measurement of plasma free testosterone. J Clin
specifically bound T reflects reliably non-SHBG-T (bioavail-
able T), and that immunoassayable SHBG is a reliable mea-
9. Vermeulen A. 1973 Testosterone in plasma: a physiopathological study. Verh.
Konikl Acad Geneesk Belg. 35:95–180.
10. Moll Jr GW, Rosenfield RL, Hecke JH. 1981 Estradiol-testosterone binding
interaction and free plasma estradiol under physiological conditions.J Clin
References
11. So¨dergard R, Ba¨ckstro¨m T, Shanbhag V, Carstensen H. 1982 Calculation of
1. Manni A, Partridge WM, Cefalu W, et al. 1985 Bioavailability of albumin
free and bound fractions of testosterone and estradiol 17 to human plasma
bound testosterone. J Clin Endocrinol Metab. 61:705–710.
protein at body temperature. J Steroid Biochem. 26:801– 810.
2. Vermeulen A, Stoica T, Verdonck L. 1971 The apparent free testosterone
12. Sinha-Hikim I, Arver S, Beall G, et al. 1998 The use of a sensitive equilibrium
concentration, an index of androgenicity. J Clin Endocrinol. 33:759 –767.
3. McCann DS, Kirkish LS. 1985 Evaluation of free testosterone in serum. J Clin
dialysis method for the measurement of free testosterone levels in healthy
cycling women and in human immunodeficiency virus-infected women. J Clin
4. 0Connor S, Baker HWG, Dulmanis A, Hudson B. 1973 The measurement of
sex steroid binding globulin by differential ammoniumsulfate precipitation. J
13. Winters SJ, Kelly DE, Goodpaster B. 1998 The analog free testosterone assay:
are the results in men clinically useful? Clin Chem. 44:2178 –2182.
5. Cumming DC, Wall SR. 1985 Non sex hormone binding globulin bound
14. Kapoor P, Luttzell BM, Williams D. 1993 The free androgen index is not valid
testosterone as a marker of hyperandrogenism. J Clin Endocrinol Metab.
for adult males. J Steroid Biochem Mol Biol. 45:325–326.
15. Bammann BL, Coulam CB, Jiang N. 1980 Total and free testosterone during
6. Mathur RS, Moody LO, Landgrebbe S, Williamson HO. 1981 Plasma an-
pregnancy. Am J Obstet Gynecol. 137:293–298.

Date of the report: 07.05.2002 Report number: 12 Veterinary drugs in poultry and rabbit – secondary investigation Joint campaign by the cantons Basel Country and Basel City (main laboratory) Background Antibiotics are used in cattle, pigs and poultry to promote performance (faster growth and improved feed conversion) as well as in the prophylaxis and treatment of infectious diseases. W

Keil Urogynecology 4500 E. 9th Ave. Ste. 420 Denver, CO 80220 Urodynamic Appointment: _________________________ Special Instructions: 1. Please come to our office with a full bladder. 2. If you have had a knee replacement, have mitral valve prolapse, or must take antibiotics prior to any dental or medical procedure, please take these antibiotics prior to your testing. Please notif