LIN.
CHEM. 41/11,
Reliability Fames ead,6
1581-1
584
(1995)
of Salivary
M. Dabbs Elizabeth
#{149} Enzymes
and
Testosterone
Jr.,”t#{176}Ben J. Susman,7
Protein
Markers
Measurements:
C. Campbell,2 Brian A. Gladue,3 Leon M. J. W. Swinkels,8 and
The reliability of salivary testosterone assays was evalutted by nine laboratories in four countries. Each laboraory used its own RIA procedures to assay samples from t set of 100 male and 100 female subjects. Agreement
imong ange
the laboratories reported
161-76). ndicated
by
on mean
Read
(Ann
scores
was within
N Y Acad
Sci
the
1 993;694:
Overall agreement by the intraclass
on individual scores, as correlation coefficient cornacross laboratories, was r = 0.87
A Multicenter
A. Rees Midgley,4 Carol M. Worthman9
Subjects
Evaluation
Miguel
A. Navarro,5
Graham
F.
and Methods
Saliva samples were collected from 100 male and 100 female undergraduates at Georgia State University, following a protocol approved by the Institutional Review Board. Each subject chewed a stick of sugar-free gum to stimulate the flow of saliva and deposited 15-20 mL into centrifuged
a 20-mL polyethylene to remove debris,
vial. The saliva 1.5-mL volumes
and
was from
)uted within subjects or men and r = 0.78 for women. Mean agreement )etween each laboratory and the combined set of all )ther laboratories (via Fisher’s Z-transformation) was r = ).61 for men and r = 0.58 for women. We take these atter values to be the best estimates of the average eliability of laboratories in their ordering of individual am pies.
each subject were producing a total
pipetted of 1800
into samples.
and
samples
were
ndexing
six of them assayed the women’s samples. Laboratories 2, 3, and 7 chose not to assay women’s samples because they had no experience assaying them or needed larger sample volumes. Of the resulting 1500 assays, 13 were not completed for various reasons. Before statistical analysis, a log transformation was applied to normalize the testosterone distributions. Six scores were discarded because they fell >3 SD above the means for their laboratories. The procedure for discarding outliers was performed once, using means and SDs for the log-transformed data from each laboratory. Log-transformed scores were used in all statistical analyses, but means and variances are reported below in untransformed raw score units, because readers are more familiar with raw score units and find them easier to
100
women’s
;ex-related
radioimmunoassay/interlaboratory
comparison!
differences
The ease of sample collection has led to a growing iumber of studies involving salivary testosterone mea;urements (1). Subjects will participate readily, and esearch can be carried out in diverse settings and )opulations. Salivary measures are less well known ;han serum measures, however, and researchers and eviewers have been skeptical about their use. Salivary ;estosterone is a well-established marker for circulatng free testosterone concentrations (2-5), but ques;ions remain about the reliability of measurements. Phe present study addressed the issue of agreement mong laboratories in assaying identical sets of saliva amples. Identical saliva samples were sent to laboratories with established records of conducting RIAs of salivary ;estosterone. Each laboratory used its own procedure ;o conduct the assays. The study examined agreement imong laboratories in their mean values for men and women and in their ordering of individual scores iround the respective means.
Department
of Psychology,
Georgia
State
University,
Atlanta,
Table
1. General
assay
Carolina
Population
Center,
Chapel
Hi!!,
1
charcoal 2
laboratory.
3
125I coated-tube
4
125
125!
7
3H
kit from Farmos ether-ethylacetate glycol separation.
1251
Diagnostica extraction,
(Bucks,
(Turku,
chromatography,
modified modified
extraction,
CLINICAL
Finland),
and polyethylene UK), ether
extraction,
separation.
Products 125
kit from
kit from ICN, no extraction. Amersham, ether extraction,
from paper
PA.
9
Diagnostic no extraction. Diagnostic
CA),
no extraction.
3H from Amersham
6
kit from
antibody Corp.,
charcoal
8
antibody
Corp. (Los Angeles,
coated-tube
125
extraction,
separation.
Products
5
e-mail
at each
ICN (Costa Mesa, CA), ether
125I from
NC.
University of Cincinnati, Cincinnati, OH. University of Michigan, Ann Arbor, MI. Hospital Princeps d’Espanya, Barcelona, Spain. 6 Tenovus Cancer Research Centre, Cardiff, UK. Pennsylvania State University, University Park, 8 Katholieke Universiteit, Nijmegen, The Netherlands. Emory University, Atlanta, GA. 10 Author for correspondence. Fax 404-651-1391; )
[email protected]. Received December 20, 1994; accepted July 18, 1995.
procedures
Laboratory
iA. 2
vials, men’s
The references include representative from the authors at these laboratories (6-17). laboratories assayed the men’s samples, and
Products 1
separate of 100
laboratory.
publications
All nine Terms:
Sets
frozen and shipped on participating laboratories of the present paper). of the RI.A procedure at
frozen CO2 to each of nine (associated with the authors Table 1 shows characteristics each
nine
charcoal
coated-tube
kit from
Corp., double kit from 2nd
Binax
by
Diagnostic
ether extraction. (S. Portland, ME), ether
antibody
CHEMISTRY,
followed separation.
separation.
Vol.
41, No. 11, 1995
1581
Table
2. Testosterone measurements laboratory.
800
at each
700
Women
Men
-j
Mean
Laboratory
(SD), pmol/L
CV, %
370 (1 33)bc
3.8
Mean
CV, %
(SD), pmol/L
2
357
(167)
8.5
3
410
(lgl)d
8.1
4
240
(95)
4.7
5
272
(99)
6
309
(111)
7
393
(181)d
8
324
(129)
6.3 8.6
55 (24)
12.9
9
382
(1
3.9
48 (28)
8.5
Measurements
600
E
6.3
55 (20)
0
500 C
5.1
0 400
6.9
300
89(21) 75 (25) 101
2.8
42)cd
are based
on 94-100
(0
0
(47)
I.-
cases
each.
SD refers
200 100
to variation
across subiects in each assay. CV refers to variation between pairs of sample duplicates within each assay. Superscript letters indicate results of statistical comparisons among means. Means within each sex that do not share a common superscript are significantly different from each other (NeumanKeuls test, P
