Copas, AJ

13 Pages·2017·1.21 MB·English

ISSN2472-1972
Distribution of Salivary Testosterone in Men
and Women in a British General
Population-Based Sample: The
Third National Survey of
Sexual Attitudes and
Lifestyles (Natsal-3)
Brian G. Keevil,1 Soazig Clifton,4 Clare Tanton,4 Wendy Macdowall,5
Andrew J. Copas,4 David Lee,2 Nigel Field,4 Kirstin R. Mitchell,5,6
Pam Sonnenberg,4 John Bancroft,7 Cath H. Mercer,4 Anne M. Johnson,4
Kaye Wellings,5 and Frederick C. W. Wu3
1DepartmentofClinicalBiochemistry,UniversityHospitalSouthManchester,ManchesterAcademic
HealthScienceCentre,2CathieMarshInstituteforSocialResearch,
SchoolofSocialSciences,and3AndrologyResearchUnit,ManchesterCentre
ofEndocrinologyandDiabetes,ManchesterAcademicHealthScienceCentre,TheUniversityof
Manchester,ManchesterM139PL,UnitedKingdom;
4ResearchDepartmentofInfectionandPopulationHealth,University
CollegeLondon,LondonWC1E6BT,UnitedKingdom;
5DepartmentofSocialandEnvironmentalHealthResearch,LondonSchoolof
HygieneandTropicalMedicine,LondonWC1E7HT,UnitedKingdom;
6MedicalResearchCouncil/ChiefScientistOfficeSocialandPublicHealthSciencesUnit,
UniversityofGlasgow,GlasgowG40SF,UnitedKingdom;and
7KinseyInstitute,IndianaUniversity,Bloomington,Indiana47405
Introduction:Measurementofsalivarytestosterone(Sal-T)toassessandrogenstatusoffersimportant
potentialadvantagesinepidemiologicalresearch.Theutilityofthemethoddependsontheinterpretationof
theresultsagainstrobustlydeterminedpopulationdistributions,whicharecurrentlylacking.
Aim:Todetermineage-specificSal-Tpopulationdistributionsformenandwomen.
Methods: Morning saliva samples were obtained from participants in the third National Survey of
SexualAttitudesandLifestyles,aprobabilitysamplesurveyoftheBritishgeneralpopulation.Sal-Twas
measuredusingliquidchromatography-tandemmassspectrometry(LC-MS/MS).Linearandquantile
regressionanalyseswereusedtodeterminetheage-specific2.5thand97.5thpercentilesforthegeneral
population(1675menand2453women)andthepopulationwithhealthexclusions(1145menand1276
women).
Results:Inthegeneralpopulation,themeanSal-Tlevelinmendecreasedfrom322.6pmol/Lat18years
ofageto153.9pmol/Lat69yearsofage.Inwomen,thedecreaseinthegeometricmeanSal-Tlevelwas
from39.8pmol/Lat18yearsofageto19.5pmol/Lat74yearsofage.Theannualdecreasevariedwith
age,withanaverageof1.0%to1.4%inmenand1.3%to1.5%inwomen.Forwomen,the2.5thpercentile
fell below the detection limit (,6.5 pmol/L) from age 52 years onward. The mean Sal-T level was
approximately6timesgreaterinmenthaninwomen,andthisremainedconstantovertheagerange.
TheSal-Tlevelwaslowestformenandhighestforwomeninthesummer.Theresultsweresimilarfor
thegeneralpopulationwithexclusions.
Abbreviations: b,lowerbound;BMI,bodymassindex;b ,upperbound;LC-MS,liquidchromatography-tandemmassspectrometry;
l u
Natsal,NationalSurveyofSexualAttitudesandLifestyles;Sal-T,salivarytestosterone;SD,standarddeviation;SHBG,sexhormone
bindingglobulin.
Received21October2016 January2017|Vol.1,Iss.1
Accepted14December2016 doi:10.1210/js.2016-1029|JournaloftheEndocrineSociety|14–25
Downloaded from https://academic.oup.com/jes/article-abstract/1/1/14/2890811
by London School of Hygiene & Tropical Medicine user
on 15 January 2018
doi:10.1210/js.2016-1029|JournaloftheEndocrineSociety|15
Conclusions:Toourknowledge,thisisthefirststudytodescribethesex-andage-specificdistributions
forSal-TinalargerepresentativepopulationusingaspecificandsensitiveLC-MS/MStechnique.The
presentdatacaninformfuturepopulationresearchbyfacilitatingtheinterpretationofSal-Tresultsasa
markerofandrogenstatus.
ThisarticlehasbeenpublishedunderthetermsoftheCreativeCommonsAttributionLicense(CC
BY;https://creativecommons.org/licenses/by/4.0/),whichpermitsunrestricteduse,distribution,and
reproductioninanymedium,providedtheoriginalauthorandsourcearecredited.Copyrightforthis
articleisretainedbytheauthor(s).
Freeform/KeyWords:saliva,testosterone,liquidchromatography-tandemmassspectrometry,
LC-MS,population
The use of saliva in the investigation of testosterone status is attractive because sample
collection is convenient, requires minimal training, and can be easily undertaken at home.
Measurementofsalivarytestosterone(Sal-T),therefore,offersgreatpotentialinfacilitating
epidemiological and biomedical research at the population level.
Most testosterone circulating in the blood is bound to sex hormone binding globulin
(SHBG)andalbumin,renderingonlyasmallfree(unbound)fraction(~1%to2%)[1]ableto
diffuse across capillaries into target tissues, where it exerts biological activity. Direct
measurementofserumfreetestosterone(thereferencestandard)istechnicallychallenging
and expensive; hence, serum free testosterone is usually derived from mathematical
formulasusingassociationconstantsoftestosteronewithitsbindingproteins[2].However,
the relationship of calculated serum free testosterone to directly measured free testos-
terone and the clinical significance have not been universally accepted [3]. Testosterone
circulatinginthebodyreadilydiffusesacrosscapillariesandsalivaryducts,resultingina
salivary fraction containing free unbound testosterone [4]. Measurement of Sal-T might
thus provide an alternative to measuring serum total testosterone, free testosterone, or
bioavailabletestosteroneintheassessmentofandrogenstatus.Concernshavebeenraised
regardingthereliabilityofSal-Tmeasurementusingimmunoassaymethods[5].However,
recentmethodologicaladvanceshaveallowedSal-Ttobereliablyandaccuratelymeasured
usingmorespecificandsensitiveliquidchromatographytandemmassspectrometry(LC-
MS/MS)[6–8].Inbothmenandwomen,Sal-Tcorrelateswellwiththecalculatedserumfree
testosteronelevel[8]anddoesnotcorrelatewithSHBG[9].AhighcorrelationbetweenSal-
T and serum free testosterone measured by equilibrium dialysis in both men and women
has also been confirmed; however, a substantial systematic positive bias was present in
women, which might reflect the influence of salivary protein binding to the lower female
concentrations of Sal-T [8]. Whether Sal-T can be a surrogate for circulating free testos-
terone or a valid measure of tissue bioavailable testosterone can now be investigated
further.
ApplicationofSal-Tmeasurementsfortheassessmentofandrogenstatusinmenand
women is critically dependent on the interpretation of results against rigorously de-
termined age-specific population distributions. Using relatively small numbers of
samples from hospital personnel or clinic attenders for this purpose is convenient but
problematic owing to inherent selection bias and inadequate statistical power. Pop-
ulation ranges based on probability samples are more representative of the general
population. They have only become available recently for serum testosterone mea-
surementsinbothmen[10]andwomen[11]and,asyet,havenotbeenwidelyused.The
presentstudyaimedtodeterminetheage-specificpopulationdistributionsforSal-Tina
large sample of adult men and women from the general population in Britain using a
highly sensitive and specific LC-MS/MS method. We have provided population distri-
butions for the general population with exclusions (excluding those with self-reported
medical conditions or using medications that can alter the testosterone levels) and the
generalpopulationacrossthefullagerangetomaximizeitsusefulnessforabroadrange
of research studies.
Downloaded from https://academic.oup.com/jes/article-abstract/1/1/14/2890811
by London School of Hygiene & Tropical Medicine user
on 15 January 2018
16|JournaloftheEndocrineSociety|doi:10.1210/js.2016-1029
1. Methods
A. Study Population
The third National Survey of Sexual Attitudes and Lifestyles (Natsal-3) is a stratified
probability sample survey of 15,162 men and women aged 16 to 74 years and resident in
Britain, which used the postcode address file as its sampling frame. Participants were
interviewed between September 2010 and August 2012 using computer-assisted personal
interviewing,includingacomputer-assistedself-interviewforthemoresensitivequestions.
Theresponseratewas57.7%.Fulldetailsofthemethodsusedhavebeendescribedpreviously
[12, 13].
Aftertheinterview,asubsampleofmenandwomenaged18to74years,whodidnot
regularly work night shifts, were invited to provide a saliva sample to test for tes-
tosterone, without a return of the results. Consenting participants were given a self-
collectionpackandaskedtoprovidetheirsamplebefore10AMtominimizethediurnal
variation in testosterone [7]. They were asked not to brush their teeth, eat, or chew
before giving the sample and to spit directly into a plain polystyrene tube. The saliva
sampleswere posted to the laboratory, where they were prepared and frozenat 280°C
untilanalysis[7].Onreceiptofthesample,theparticipantsweresenta£5voucherasa
token of appreciation.
Of 13,431 participants aged 18 to 74 years who did not regularly work at night, 9170
wereinvitedtoprovideasalivasample.Atotalof4591sampleswerereceivedandmatched
tothesurveydata(50.1%ofthoseinvitedtoprovideasample).Ofthesamples,463(10.1%)
were excluded (insufficient volume, n = 154; sample discolored or bloody, n = 91; sample
recorded as taken after 10:30 AM, n = 34; .5 days between the sample being taken and
receivedbythelaboratoryorintervalunknownbecausedateofcollectionmissing,n=172;
andnottestedbecauseoferror,n=12),leaving4128samples(45.0%)withatestosterone
result.
Theanalysisforthegeneralpopulationincludedall4128participants(1675menand2453
women) with usable testosterone results. To generate the distribution for a general pop-
ulationwithexclusions(thosewhodidnotreporthealthconditionsortakingmedicationthat
can influence testosterone levels), 530 men and 1177 women were excluded from analysis
(individualscouldbeexcludedfor.1reason).Thereasonsincludedprostatecancer(13men),
prostateenlargement(90men),prostatesurgery(20men),andpolycysticovaries(35women).
Theexclusionsalsoincludedtreatmentofanyofthefollowinginthepreviousyear:cancer,25
men and 49 women; thyroid conditions, 27 men and 183 women; testicular or pituitary
conditions, 16 men; and ovarian or pituitary conditions, 23 women. Also, we excluded par-
ticipantsiftheyweretakingprescriptionmedicationforepilepsy(15menand15women),had
undergone hysterectomy and were taking hormone replacement therapy (to indicate oo-
phorectomy;181 women), and becauseofunpromptedreporteduse oftestosterone(1 man).
We did not ask participants directly regarding their use of testosterone. Additionally, we
excluded363menwithabodymassindex(BMI),18.5or.30kg/m2and118womenwitha
BMI,18.5or.40kg/m2.Womenreportingthecurrentuseofeitherhormonalreplacement
therapy(62women)orhormonalcontraception(pill,intrauterinedevice,injections,implants,
orpatch;535women)andthosewhowerecurrentlypregnant(42women)werealsoexcluded.
Finally,thosewhodidnotansweranyoftheabovequestionswereexcluded(42menand134
women),leaving1145menand1276womenfortheanalysisofthegeneralpopulationwith
exclusions.
B. Measurements
The LC-MS/MS Sal-T assay was developed using strict validation criteria [7, 14]. Sample
preparationusingliquid–liquidextractionentailedaddingsample(200mL),D -testosterone
5
internalstandard(10mL;340pmol/L),andmethyl-tert-butylether(1mL).Aftervortexing
Downloaded from https://academic.oup.com/jes/article-abstract/1/1/14/2890811
by London School of Hygiene & Tropical Medicine user
on 15 January 2018
doi:10.1210/js.2016-1029|JournaloftheEndocrineSociety|17
for 5 minutes, the organic layer was transferred and evaporated and the residue recon-
stituted with a 500-mL/L methanol mobile phase (80 mL) and transferred to a 96-well
microtiter plate.
Liquid chromatography was performed with an ACQUITY Ultra Performance Liquid
Chromatography system coupled to a Xevo TQ-S mass spectrometer (Waters Corporation,
Manchester,UK)operatedinpositiveionizationmode.Thelowerlimitofquantificationwas
6.5 pmol/L, and the assay was linear to $52,000 pmol/L. The interassay coefficient of var-
iation6standarddeviation(SD)andbiaswas12.9%61.7%and1.2%;9.8%62.5%and0.4%;
and4.5%612.0%and1.9%ataconcentrationof12.9,26.0,and260pmol/L,respectively.The
intra-assaycoefficientofvariation6SDandbiaswas9.5%61.3%and0.8%;5.5%61.6%and
12.6%; and 2.1% 6 6.2% and 11.1% at a concentration of 12.9, 26.0, and 260 pmol/L, re-
spectively. Recovery was 104% (range, 98.3% to 108.9%) [7].
C. Statistical Analysis
StatisticalanalyseswereperformedusingSTATA,version13.1(StataCorp,CollegeStation,
TX),accountingforthecomplexsurveydesign(stratification,clustering,andweightingofthe
sample). We applied 2 weights when analyzing the data. The survey weight corrected for
unequal probability of selection and differential response (by age, sex, and region) to the
survey itself, and the saliva weight corrected for unequal probability of selection and dif-
ferentialresponsetothesalivasample.Anumberoffactorswereassociatedwithprovidinga
sample,includingageatinterview,ethnicity,generalhealth,andsexualfunctionmeasured
usingtheNatsalsexualfunctionquestionnaire[15].Thefulldetailsoftheseweightsandtheir
calculation have been previously reported [12].
We used 2 statistical approaches to estimate the 2.5th to 97.5th percentiles for the pop-
ulation distributions for Sal-T levels in men and women: linear regression and quantile
regression,aspreviouslyreportedforcalculatingtheserumtestosteronereferenceranges[10,
11]. Both analyses were performed to produce the distribution limits for the general pop-
ulation and the general population with exclusions.
Linear regression, as a parametric technique, can be unduly affected by extreme values.
Therefore, very high Sal-T values were censored such that for each 10-year age group
stratifiedbysex,valuesgreaterthanthe99thpercentilewerereplacedbythe99thpercentile
(17menand26women).The99thpercentilevaluesrangedfrom587.4pmol/Lintheyoungest
mento352.6pmol/Lintheoldestmenandfrom233.2pmol/Lintheyoungestwomento104.6
pmol/L in the oldest women, respectively. The Sal-T data for men were approximately
normallydistributed;however,thedistributionforwomenwasskewed.Thus,thevalueswere
transformedonthenaturallogscaleforanalysisandback-transformedtogeneratethefinal
population distribution limits.Becausequantile regression isa nonparametric approach, it
wasnotnecessarytocensortheextremehighvaluesofSal-Tortotransformthedataforthe
women.
Threemen(allaged.60years,allincludedinthegeneralpopulationwithexclusions)and
76women(distributedacrossthe18to74-year agerange,33ofwhomwere included inthe
general population with exclusions) had Sal-T levels less than the limit of detection (,6.5
pmol/L).Intervalregressionwasused,assigningthesecasestotherangeof0to6.5pmol/Lfor
the linear regression for men. The lower bound for the women was set as 0.5 to allow the
values to be log transformed. For quantile regression, these cases were assigned a value of
3.25 pmol/L (one-half the limit of detection).
Forbothmenandwomen,theSDofSal-Twasnotconstantwithage.Therefore,afterfitting
the linear regression for the mean values, we calculated the SD of the Sal-T levels for each
year of age and used these values as the outcome in a second linear regression analysis to
predicttheSDasafunctionofage.Thepredicted2.5thand97.5thpercentilesforeachyearof
age were calculated as the predicted mean Sal-T minus the predicted SD for that age
multipliedbythelowerbound(b)andthepredictedmeanplusthepredictedSDmultipliedby
l
theupperbound(b ),respectively,withb andb selectedsuchthatacrossallages,2.5%ofthe
u l u
Downloaded from https://academic.oup.com/jes/article-abstract/1/1/14/2890811
by London School of Hygiene & Tropical Medicine user
on 15 January 2018
18|JournaloftheEndocrineSociety|doi:10.1210/js.2016-1029
populationhadtestosteronevalueslessthanthelowerboundsand2.5%ofthepopulationhad
testosterone values greater than the upper bounds. We tried different values for each
multiplier,b andb ,startingwith1.96,whichcorrespondedtothenormaldistribution,and
l u
iteratively increasing or decreasing the values until we achieved the desired coverage. For
meninthegeneralpopulation,thevalueswereb of2.00andb of2.30,andforwomeninthe
l u
generalpopulation,theywereb of2.11andb of1.96.Thevaluesforthemeninthegeneral
l u
population with exclusions were b of 2.09 and b of 2.25, and for women, b of 2.10 and b
l u l u
of 1.96.
Formen,theSDofSal-Tdecreasedwithageuptoapoint(fromapproximatelyage70years)
and increased again in the oldest age group. We were unable to adequately model this in-
creaseintheSDtoaccuratelycalculatethe2.5thand97.5thpercentiles(whicharebasedon
theSD)andconsequentlytruncatedthepopulationdistributionanalysisformenatage69.No
equivalent increase in the SD was found among older women; therefore, the data are pre-
sentedforthefullagerange,18to74years.Truncationwasnotnecessaryfortheanalysisof
the mean testosterone levels, including associations with seasonal changes.
To allow for a possible nonlinear relationship between Sal-T and age, we explored 2 dif-
ferent functions of age (in addition to a linear function) in both the linear and the quantile
regressionanalyses:aquadraticfunctionandarestrictedcubicsplinefunction.Forthelatter,
3knotswerespecifiedatthe10th,50th,and90thpercentilesofage(thedefaultplacementfor
3knots).Thepopulationdistributionproducedbythemodelsusingthequadraticandcubic
splinefunctionswassimilar;therefore,weoptedtousethesimplerquadraticfunctioninthe
finalmodels.Forwomen,theanalyseswereperformedonlog-transformeddataandthedata
were back-transformed; therefore, the geometric mean values are presented.
To assess the seasonal variation in testosterone, the mean (geometric mean for women)
testosteroneand95%confidenceintervalswereplottedbyseasonforthegeneralpopulation,
and linear regression was used to test for differences. Each season was defined as winter
(December, January, and February), spring (March, April, and May), summer (June, July,
and August), and autumn (September, October, and November). To explore potential geo-
graphical differences, the participants were grouped into 3 broad regions of residence:
Scotland and North of England, Midlands and Wales, and East and South of England (in-
cluding London).
D. Ethics Statement
The OxfordshireResearch Ethics CommitteeAapprovedNatsal-3 (reference no.09/H0604/
27).Allparticipantsprovidedwritteninformedconsentforanonymizedtestingofthesaliva
samples, without a return of the results.
2. Results
Distributionsofthemean6SDandmedianandinterquartilerangeoftheSal-Tlevelsinthe
generalpopulationby10-yearagegrouparelistedinTable1andforthegeneralpopulation
withexclusionsinSupplementalTable1.TheSal-Tlevelsforbothmenandwomenshoweda
distinct age-related decline, with a clear demarcation in the mean levels between men and
women.ThemeanSal-Tconcentrationwasapproximately6timesgreaterinthementhanin
thewomen;thisrelationshipremainedconstantoverthe6decadesstudiedinboththegeneral
population and the general population with exclusions (Table 1; Supplemental Table 1).
The Sal-Tdistributions accordingtothelinear and quantileregression analyses formen
andwomeninthegeneralpopulationareshowninFig.1.Forbothmenandwomen,thelinear
and quantile regression analyses produced similar population distributions. Supplemental
Table2showstheage-specificvaluesforthe2.5thand97.5thpercentilesofthedistribution
for the general population produced by the linear regression (those produced by quantile
regression analysis not shown). The Sal-T distributions for men and women in the general
populationwithexclusionsareshowninSupplementalFig.1,andthevaluesforthe2.5thand
Downloaded from https://academic.oup.com/jes/article-abstract/1/1/14/2890811
by London School of Hygiene & Tropical Medicine user
on 15 January 2018
doi:10.1210/js.2016-1029|JournaloftheEndocrineSociety|19
Table1. MeanandMedianSalivaryTestosteronebyAgeGroupandSexinGeneralPopulation
Sal-T(pmol/L) Denominator
Variable Mean6SD Median(IQR) UWt Wt
Men
18–24 314.86111.6 314.9(246.3–384.1) 187 244
25–34 266.76102.5 264.6(198.6–325.9) 249 335
35–44 232.6691.5 229(178.4–285.3) 244 376
45–54 207.5680.2 203.2(155.3–248.9) 305 397
55–64 174.4664.7 175.9(130.6–214.9) 347 350
65–69 157.6658.5 152.0(119.3–190.1) 194 153
Women
18–24 51.1645.1 39.2(21.7–65.6) 247 268
25–34 42.6632 37.1(24.3–49.6) 441 403
35–44 41.1631.7 32.4(21–50.7) 425 414
45–54 33.9628.5 26.6(17.9–40.5) 451 430
55–64 27.6618.6 22.9(15.3–35.8) 462 368
65–74 27.5620.2 23.2(14.8–33.2) 427 284
Abbreviations:IQR,interquartilerange(25to75thpercentiles);UWt,unweighted;Wt,weighted.
97.5thpercentilesofthedistributionarelistedinSupplementalTable2.Forwomen,the2.5th
percentile fell below the limit of detection (,6.5 pmol/L) from age 52 years onward in the
generalpopulationandage54yearsonwardinthegeneralpopulationwithexclusions;thus,
these data are not provided.
TherangeofSal-Tvaluesgreaterthanthe97.5thpercentileamongwomenaged,55years
was wide; however, for those aged .55 years, most of the high values were clustered just
abovethe97.5thpercentileline(Fig.1).Althoughdetailedinformationonmenstrualphase
was not collected, our questionnaire enabled the identification of women who had provided
salivasampleswithin7daysofstartingtheirlastmenstrualperiod(presumedearlyfollicular
phase). Very few of the high values in the premenopausal women were among those in the
early follicular phase (data not shown), suggesting that they might ay reflect mid-cycle
testosterone peaks [16].
Forthefullagerangeexamined,themeanSal-Tlevelsdecreasedbyapproximately50%to
60% in both the general population with exclusions and the general population of men and
women. Because our models of the association between Sal-T and age included a nonlinear
functionofage,thepredictedyear-by-yeardeclineintestosteronevariedbyage.Formenin
thegeneralpopulation,thepredicteddecreaseintheaverageSal-Tlevelforeachyearofage
was1.3%to1.5%.Thepredicteddeclinebetweenage18and19was1.4%(range,322.6to318.0
pmol/L),betweenage45and46was1.5%(range,216.9to212.7pmol/L),andbetweenage68
and 69 was 1.3% (range, 156.0 to 153.9 pmol/L). For women in the general population, the
predicteddecreaseintheaverageSal-Tlevelforeachyearofagewas1.0%to1.4%.Thedecline
betweenage18and19was1.0%(range,39.8to39.4pmol/L),betweenage45and46was1.4%
(range,28.9to28.5pmol/L),andbetweenage73and74was1.0%(range,19.7to19.5pmol/L).
SeasonaldifferencesinthemeanSal-Tlevelswereobserved(P,0.0001forbothmenand
women;Fig.2);however,thesedifferedbysex,withthelowestlevelsinthesummerformen
andthehighestlevelsinthesummerforwomen.Wefoundnoassociationsbetweenthemean
Sal-T level and the broad geographical region (P = 0.2432; Fig. 2).
3. Discussion
To our knowledge, the present study is the first to establish age-specific population distri-
butions for LC-MS–analyzed Sal-T in men and women from a large general population
sample.Ourfindingsshowedonlyaminoroverlapbetweentheage-specificmaleandfemale
Downloaded from https://academic.oup.com/jes/article-abstract/1/1/14/2890811
by London School of Hygiene & Tropical Medicine user
on 15 January 2018
20|JournaloftheEndocrineSociety|doi:10.1210/js.2016-1029
Figure1. Distributionofsalivarytestosteronein(A)menand(B)womeninthegeneral
population.Curvescreatedusinglinearregression(solidline)forthefittedmean(men)or
geometricmean(women)forthe2.5thand97.5thpercentilesandquantileregression(dashed
line)forthemedian,2.5thpercentile,and97.5thpercentile.Observedvalues(x)for1526men
and2543womendisplayed.
populationdistributions,mirroringthoseseenwithserumtestosterone,andlendingsupport
tothevalidity ofourSal-T measurements.The finding ofsixfold greaterSal-T levelsinthe
mencomparedwiththewomenwasalsosimilartothatobservedforserumtestosterone[17],
reflecting the markedly greater daily testosterone blood production rate in men [18]. A
distinct age trend in Sal-T levels was observed in both sexes. The rate of cross-sectional
declineinSal-TwithagewassimilartothedeclineinSal-Twithageinothersmallerstudies
ofmen[19–21]andwomen[20]butgreaterthanthereporteddeclineinserumtestosteronein
men [22–24] and women [11, 25, 26].
The age-associated decline in serum testosterone has been implicated in a variety of
physiologicalchangesinagingmen[27,28].However,thishasbeendisputedbysome[29,30],
who have suggested that the apparent decline is largely due to comorbidity, with healthy
elderlymenshowinglittlechangeintheircirculatingtestosteronelevels.Althoughwefound
that men aged .45 years who had not reported any of the exclusion health conditions had
slightlygreaterlevelsofSal-Tcomparedwiththewholesample,theSal-Tlevelsinthesemen
hadneverthelessdecreasedbyone-halffromage18to69years.Thissuggeststhatthewidely
reportedserumtotalandfreetestosteronedecreasesduringthelifecourseof17%and35%,
Downloaded from https://academic.oup.com/jes/article-abstract/1/1/14/2890811
by London School of Hygiene & Tropical Medicine user
on 15 January 2018
doi:10.1210/js.2016-1029|JournaloftheEndocrineSociety|21
Figure2. Mean(men)andgeometricmean(women)salivarytestosterone(pmol/L)byseason
[(A),men;(B)women]andregion[(C)men;(D)women]inthegeneralpopulation.
respectively[22–24],mightunderestimatetheaging-associateddeclineintesticularfunction
or testosterone bioavailability at the tissue level. An important corollary of this compelling
age trend in Sal-T is to reinforce the view that a Z-score approach, using age-specific pop-
ulation ranges, might be more appropriate and physiologically meaningful than the pre-
viously preferred testosterone score approach using comparisons with ranges derived from
young (age ,40 years) healthy men [10].
Inpremenopausalwomen,weobservedsomeextremehighvaluesofSal-T,extendingfar
above the 97.5th percentile, which possibly reflected the midmenstrual cycle peaks in tes-
tosterone[16,26,31].Wedidnotcollectdetailedinformationonthemenstrualphase;thus,we
were unable to control for variations in testosterone across the menstrual cycle in our
analysis.Inbroadagreementwiththeserumtestosteronelevelsfromotherlargepopulation-
based studies [11, 25, 26], we found that the decline in Sal-T in women was steepest in the
early reproductive years and subsequently flattened out in midlife. In agreement also with
the serum testosterone findings from other studies [11, 26, 32], we did not observe a sub-
stantial effect of the menopausal transition on Sal-T levels. The percentage of change in
serumtestosteronepreviouslyfoundinhealthywomenaged20to60yearswas30%[11].In
contrast, the percentage of change in Sal-T in our study was ;60% for a similar age range.
Thus,justasinmen,theage-relateddecreaseinSal-Tlevelsinwomenwhodidnotreportany
of the exclusion health conditions was appreciably greater than that observed for serum
testosterone.Theprincipalsourcesofandrogensinpostmenopausalwomenaretheadrenal
gland and the ovary [33]. An increase in free testosterone could also arise from a relative
decrease in SHBG compared with testosterone, a finding consistent with the trend of de-
creasing SHBG across the menopausal transition [34].
TheseasonalvariationinSal-Tobservedinmenshowedtheoppositetrendtothatseenin
women, with an increase in the summer and a decrease in the winter in women. Previous
studiesexaminingseasonalvariationsinserumtestosteronelevelsinmenandwomenhave
yieldedinconsistentresults,witheithernoseasonalvariationfound[35]orwithpeaklevels
Downloaded from https://academic.oup.com/jes/article-abstract/1/1/14/2890811
by London School of Hygiene & Tropical Medicine user
on 15 January 2018
22|JournaloftheEndocrineSociety|doi:10.1210/js.2016-1029
foundinthewinter[36]orthesummer[37].IntheonlySal-Tstudy,peaklevelswerefoundin
October and December for the women and men, respectively [38]. Although statistically
significant,themagnitudeoftheobservedseasonaldifferencesinSal-Twasrelativelysmall
(~20pmol/Linmenand~8pmol/Linwomen),andthevariationmightnotbebiologicallyor
clinicallyimportant.Giventheseinconsistencies,wedonotbelieveitwouldbeappropriateto
provide separate Sal-T population distributions stratified by season.
Natsal-3isbroadlyrepresentativeoftheBritishpopulation,includingintermsofethnicity
[13],butwasnotdesignedspecificallytoexamineethnicvariationsintestosterone.Wefound
noassociationwiththebroadgeographicalregions,whichisperhapsunsurprisinggiventhat
Britainisasmallcountryintermsofareaandpreviousresearchintogeographicalvariation
has been on a global scale [39].
Thestrengthsofthepresentstudyarethelargegeneralpopulationsamplesize,thestate-
of-the-artLC-MS/MSmeasurementofSal-T,andtherigorousstatisticalanalysistechniques.
To enable the fullest application in future investigations, we established population distri-
butions,notonlyfortheentiregeneralpopulation,butalsoafterexclusionofconditionsand
medications that can affect Sal-T levels. This ensured applicability of the presented in-
formation to a wide range of epidemiological and biomedical studies in the future.
Thepresentstudyalsohadsomelimitations.Thehealthconditionswereself-reported,and
single morning saliva samples cannot account for intraindividual variations resulting from
circhoral,diurnal,andcircannualrhythms.Thelackofaccurateinformationonthetimingof
samples in relation to the menstrual cycle and clinical information on the presence of
polycystic ovarian syndrome among women could have introduced added “noise” in the
distributions.Althoughoursamplewassimilartothecensuswithrespecttoethnicity,health,
andmaritalstatusafterweighting[12,13],justaswithanygeneralpopulationsurvey,our
dataweresusceptibletosomeparticipationbiases.Forinstance,individualsinresidentialor
nursing care were not included in the sampling frame, and poor health could have affected
subjects’ willingness to participate (i.e., our population distributions for the general pop-
ulationmightrefertoaslightlyhealthiersamplethanthetrueBritishgeneralpopulation).
Thefinalresponseratetothesalivastudywas45%;therefore,thesalivadatawereweighted
during analysis to minimize the potential for a nonresponse bias [12].
Age- and sex-specific population distributions are important as a baseline against which
otheranalysesandresearchstudiescanbecompared.Thearrayofbackgroundinformation,
in particular, with respect to age and BMI, will be important when considering important
researchquestions,suchasthevariationsinSal-Tatthepopulationlevelwithrespecttothe
frequency of sexual activity at the extremes of the age spectrum, sexual satisfaction, and
number of sexual partners. Some of these questions are being addressed in our ongoing
analyses.
ThepresenteddatadescribethedistributionofSal-Tinthegeneralpopulationaspartofa
largestudytoinvestigatethedeterminantsofvariationsinsexuallifestyleandpracticesin
menandwomen.Theinformationisnotintendedtobeappliedtotheclinicalsetting(without
further stringent clinical evaluation), particularly with respect to hormone replacement
therapyforolderindividuals.TheverycleardeclineinSal-Tlevelswithagelendssupportto
theviewthatlowertestosteronelevelsareaphysiologicalchangeandargueagainsttheuseof
hormone replacement therapy for older individuals.
We have determined age-specific population distributions for Sal-T in a large, represen-
tative population of men and women using a highly specific and sensitive LC-MS/MS
technique. The relative simplicity of saliva collection has important implications for large
population-based studies, in which serum collection has been impractical or too expensive.
These population data, which can be harmonized with those from other laboratories using
validated LC-MS/MS methods, provide a benchmark for ensuring the appropriate in-
terpretationandcomparisonsofSal-Tresultsinfutureresearch.Anessentialstephasnow
beentakentoallowtheapplicationofSal-Tlevelsininvestigatingthepotentialimportanceof
androgen exposure in many aspects of sexual behavior and general health in largescale
population surveys of men and women.
Downloaded from https://academic.oup.com/jes/article-abstract/1/1/14/2890811
by London School of Hygiene & Tropical Medicine user
on 15 January 2018
doi:10.1210/js.2016-1029|JournaloftheEndocrineSociety|23
Acknowledgments
Theauthorsexpresstheirappreciationforthecontributionstothisworkofouresteemedcolleaguethelate
Dr.MichaelWallace.Wealsothankthestudyparticipants,theteamofinterviewersfromNatCenSocial
Research,operations,andcomputingstafffromNatCenSocialResearch,andthestudyfunders.Natsal-3
isacollaborationamongUniversityCollegeLondon,theLondonSchoolofHygieneandTropicalMedicine,
NatCen Social Research, Public Health England (formerly the Health Protection Agency), and the
UniversityofManchester.WegratefullyacknowledgetheimportanttechnicalcontributionsofHalina
McIntyreandAnneKelly,DepartmentofClinicalBiochemistry,RoyalInfirmary,Glasgow,Scotland,UK,
andPhilipMacdonald,Departmentof ClinicalBiochemistry, UniversityHospital SouthManchester,
ManchesterUK.
Address all correspondence to: Brian Keevil, MSc, Department of Clinical Biochemistry, Uni-
versity Hospital South Manchester, Southmoor Road, Manchester M23 9LT, UK. E-mail: brian.
[email protected].
ThepresentstudywassupportedbygrantsfromtheMedicalResearchCouncil(GrantG0701757)and
theWellcomeTrust(Grant084840),withcontributionsfromtheEconomicandSocialResearchCouncil
and Department of Health. S.C. is supported by the National Institute for Health Research (NIHR
Research Methods Programme, Fellowships and Internships; Grant NIHR-RMFI-2014-05-28). N.F. is
supportedbyanacademicclinicallectureship.SinceSeptember2015,K.M.hasbeencorefundedbythe
UKMedicalResearchCouncil,MedicalResearchCouncil/ChiefScientistOfficeSocialandPublicHealth
Sciences Unit, University of Glasgow (Grant MC_UU_12017-11). The views expressed in this publi-
cationarethoseoftheauthorsandnotnecessarilythoseoftheNationalHealthService,theNational
InstituteforHealthResearch,ortheDepartmentofHealth.
DisclosureSummary:A.M.J.hasbeenagovernoroftheWellcomeTrustsince2011.F.C.W.W.hasacted
asaconsultantforBayer-Schering,EliLilly,andBesinsHealthcareandparticipatedinadvisoryboard
meetings and lectured on their behalf; has received lecture fees from Bayer-Schering and Besins
Healthcare;andreceivedgrantsupport(2010–2014)fromBayerScheringAGandBesinsHealthcare.
Theremainingauthorshavenothingtodisclose.
ReferencesandNotes
1.HammondGL,WuTS,SimardM.Evolvingutilityofsexhormone-bindingglobulinmeasurementsin
clinicalmedicine.CurrOpinEndocrinolDiabetesObes.2012;19(3):183–189.
2.VermeulenA,VerdonckL,KaufmanJM.Acriticalevaluationofsimplemethodsfortheestimationof
freetestosteroneinserum.JClinEndocrinolMetab.1999;84(10):3666–3672.
3.HackbarthJS,HoyneJB,GrebeSK,SinghRJ.Accuracyofcalculatedfreetestosteronediffersbetween
equationsanddependsongenderandSHBGconcentration.Steroids.2011;76(1-2):48–55.
4.ViningRF,McGinleyRA,SymonsRG.Hormonesinsaliva:modeofentryandconsequentimplications
forclinicalinterpretation.ClinChem.1983;29(10):1752–1756.
5.Davison S. Salivary testing opens a Pandora’s box of issues surrounding accurate measurement of
testosteroneinwomen.Menopause.2009;16(4):630–631.
6.MacdonaldPR,OwenLJ,WuFC,MacdowallW,KeevilBG;NATSALTeam.Aliquidchromatography-
tandemmassspectrometrymethodforsalivarytestosteronewithadultmalereferenceintervalde-
termination.ClinChem.2011;57(5):774–775.
7.KeevilBG,MacDonaldP,MacdowallW,LeeDM,WuFC,TeamN;NATSALTeam.Salivarytestos-
teronemeasurementbyliquidchromatographytandemmassspectrometryinadultmalesandfemales.
AnnClinBiochem.2014;51(Pt3):368–378.
8.FiersT,DelangheJ,T’SjoenG,VanCaenegemE,WierckxK,KaufmanJM.Acriticalevaluationof
salivarytestosteroneasamethodfortheassessmentofserumtestosterone.Steroids.2014;86:5–9.
9.KeevilBG,FiersT,KaufmanJM,MacdowallW,CliftonS,LeeD,WuF.Sexhormone-bindingglobulin
has no effect on salivary testosterone [published online ahead of print April 26, 2016]. Ann Clin
Biochem.doi:10.1177/0004563216646800.
10.BhasinS,PencinaM,JasujaGK,TravisonTG,CovielloA,OrwollE,WangPY,NielsonC,WuF,Tajar
A, Labrie F, Vesper H, Zhang A, Ulloor J, SinghR, D’AgostinoR, VasanRS. Referencerangesfor
testosterone in men generated using liquid chromatography tandem mass spectrometry in a
community-basedsampleofhealthynonobeseyoungmenintheFraminghamHeartStudyandapplied
tothreegeographicallydistinctcohorts.JClinEndocrinolMetab.2011;96(8):2430–2439.
Downloaded from https://academic.oup.com/jes/article-abstract/1/1/14/2890811
by London School of Hygiene & Tropical Medicine user
on 15 January 2018