salivary testosterone and body composition among ariaal males

Salivary Testosterone and Body Composition AmongAriaal Males

BENJAMIN CAMPBELL,1* MARY T. O’ROURKE,2 AND SUSAN F. LIPSON2

1Department of Anthropology, Boston University, Boston, Massachusetts 022152Department of Anthropology, Harvard University, Cambridge, Massachusetts 02138

ABSTRACT To determine if testosterone is negatively related to acute and/or chronic nutri-tional status among men in a subsistence society, saliva samples and anthropometric measureswere compared among nomadic and settled Ariaal pastoralists of northern Kenya. Fifty-sixnomadic men and 62 settled men facing drought conditions, estimated ages 22–96 years, weresampled. Measures included height, weight, four skinfolds, and %body fat by bioelectric impe-dance (BIA). Saliva samples were assayed for testosterone using radioimmunoassay. Overall, bothbody mass index (BMI) (avg.¼ 17.8̄� 6.0 kg/m2) and salivary testosterone (T) levels (avg. AM value¼ 176.8̄ � 74.8 pmol/l) were low compared to values fromWestern populations. Comparison of thetwo subpopulations revealed no significant difference in height, weight, BMI, or lean body mass.However, nomadic males exhibited significantly smaller skinfolds. Evening, but not morning,salivary T values differed between the subpopulations. Age-related changes in body compositionincluded a significant decline in BMI with age, related to loss of body fat, but with little change inlean body mass. Age-related declines in BMI and %body fat were more pronounced among thenomadic males. AM salivary T values declined with age; again, the decline was significantly greateramong nomadic males. PM salivary T levels showed no significant decline with age. When con-trolled for residence and age, salivary T was positively related to %body fat and WHR ratio, butnot lean body mass. These results provide evidence that salivary T is related to acute nutritionalstatus among males in an energetically stressed subsistence population, in accordance with lifehistory theories of somatic allocation. Am. J. Hum. Biol. 15:697–708, 2003. # 2003 Wiley-Liss, Inc.

Studies of men in non-Western popula-tions have consistently described lowerlevels of testosterone compared to similarsamples of men in Western populations (seeCampbell and Leslie, 1995; Bribiescas, 2001,for reviews). These results span severalcontinents, including Africa (Christensen1991a,b; Worthman and Konner, 1987),South America (Beall et al., 1992; Bribiescas,1996), and Asia (Ellison and Panter-Brick,1996) and are based on assays in a varietyof different media including blood (Beallet al., 1992; Campbell, 1994), urine (Guerra-Garcia et al., 1969), and saliva (Bentley et al.,1993; Ellison and Panter-Brick, 1996).As such, lower testosterone levels amongmen in non-Western populations may beconsidered a robust phenomenon worthy ofexplanation.

It is generally agreed that the lowered Tlevels exhibited by non-Western populationsare not indicative of impaired reproductivefunction, i.e., sperm production (Campbelland Leslie, 1995; Bribiescas, 2001), but thefunctional significance of lower testosteronelevels remains a matter of debate. Ellison(2001) has suggested that lower levels of

testosterone among males in subsistencesocieties reflect the effects of chronic under-nutrition on the development of males.Bribiescas (1996, 2001) has argued thatlower T levels among males in subsistencesocieties represent an accommodation ofsomatic maintenance to limited energy avail-ability. Reduced testosterone production willlead to less stimulation of muscle tissue andreduced muscle mass. Because muscle is anenergetically costly tissue to maintain(Bribiescas, 2001), reduced growth of musclemass would result in total energy savingsunder conditions of limited nutrition.

Data on testosterone, nutritional status,and body composition in subsistence popula-tions needed to directly test the hypothesisthat variation in testosterone levels are a

� 2003 Wiley-Liss, Inc.

Contract grant sponsor: the African Studies Center atBoston University.

*Correspondence to: Benjamin Campbell, Department ofAnthropology, Boston University, 232 Bay State Road,Boston, MA 02215. E-mail: [email protected]

Received 28 February 2003; Accepted 12 May 2003

Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/ajhb.10203

AMERICAN JOURNAL OF HUMAN BIOLOGY 15:697–708 (2003)

function of both current and long-termnutritional status are largely unavailable.Existing data demonstrate 1) delayed onsetof testosterone increases during pubertyamong males in subsistence societies(Campbell, 1994; Zemel, 1993); 2) onset ofpuberty in males is related to fat stores(Vismos and Marti-Henneberg, 2002); 3) arelationship between adult height and tes-tosterone in some populations (Ellison andPanter-Brick, 1996); and 4) differences insalivary testosterone (T) across a set of sub-sistence populations are most apparentamong adult males age 20–30 (Ellison et al.,2002). Taken together, these findingssuggest that both the timing of increases intestosterone and their final adult level maybe a developmental function of overallnutritional status.

On the other hand, evidence for a directrole of current nutritional status on testo-sterone levels among adult males in non-Western populations has been hard toobtain. Bribiescas (1997) reported a relation-ship between salivary T and differencesin body mass index (BMI) with changingeconomic conditions among the Ache ofParaguay. However, results from Lese villa-gers of the Ituri forest actually show adecrease in salivary T after, as compared toduring, the hunger season despite improvednutritional status (Bentley et al., 1993). Inaddition, salivary T levels among males inNepal show little seasonal change despitesignificant changes in energy balance(Ellison and Panter-Brick, 1996).

We chose to investigate the role of nutri-tional status in salivary T levels among maleAriaal pastoral nomads of northern Kenya.Like other pastoral nomads the Ariaal havebeen reported to have low caloric intake andpoor nutritional status (Shell-Duncan andYung, in press), making them a prime casefor studying the relationship between nutri-tional status and testosterone among males.In addition, there are two distinct subgroupsof Ariaal. Nomads are primarily camel her-ders in the desert, while the settled popula-tion are cattle herders and farmers at ahigher altitude. The resulting differences inlifestyles, including subsistence activities,diet, and disease exposure, add additionalfactors to the comparison between the twogroups (Fratkin, 1998).

The fact that the nomads were experien-cing a food shortage due to drought at thetime of data collection meant that we could

test the impact of acute undernutrition inaddition to chronic undernutrition on sali-vary T under conditions similar to theimpact of environmental fluctuation duringhuman evolution. We hypothesized that, ifsalivary T is related to current energetic sta-tus, nomads would exhibit lower levels ofsalivary T related to differences in body fat.Furthermore, if salivary T is related tochronic nutritional status, we hypothesizedthat nomads would exhibit lower levelsof salivary T related to less muscle massand/or shorter heights. In addition, basedon earlier results linking maximum testo-sterone values and age-related declines in T(Ellison et al., 2002), we expect the settledgroup to show a more pronounced age-related decline in salivary T.

SUBJECTS AND METHODS

Field site

The Ariaal are pastoral nomads inhabitingboth upland and lowland regions around theNdoto Mountains in Marsabit district,Kenya. First appearing in oral history inthe 1880s, they are derived from groups ofpoor Rendille and Samburu who bandedtogether to build up their herds in the moun-tains. Culturally, they still exhibit featuresof both Rendille and Samburu, includingSamburu age-set rituals and Rendille annualcamel blessings (Fratkin, 1998).

In terms of subsistence, the Ariaal herdcamel, cattle, goats, and sheep, which theydepend on for nutrition, in the form of milk,blood, and meat. Milk is the staple, providingas much as 75% of the calories and 90% ofprotein intake during the wet season(Fratkin, 1998). As the rainfall decreases,greater amounts of blood and meat are con-sumed and animals are sold for maize meal,making it in effect the dry season substitutefor milk (Fratkin, 1998).

Comparisons of the nutritional statusamong the settled and nomadic subpopu-lations have concentrated on women andchildren. Recent work by Shell-Duncan andYung (in press) reported no difference in BMIor triceps skinfold between nomadic womenand those living in small towns. On theother hand, nomadic children showed lowerrates of malnourishment during the droughtyear of 1992 due to higher consumptionof milk compared to their town counter-parts, although there is little difference in

698 B. CAMPBELL ET AL.

measures of morbidity between the twogroups (Fratkin et al., 1999; Nathan et al.,1996). Similar information on the nutri-tional status of men is not available.

Study design

A total of 118 men, 62 from the settledvillage of Karare and 56 from the nomadicvillage of Lewogoso, participated in thestudy. The age of the participants was esti-mated from an identity card, when available,as well as membership in age sets. Age setsare based on male circumcision ceremonies,well-documented at 14 years (Fratkin, 1998).Membership in an age set is a highly impor-tant aspect of identity and unlikely tobe forgotten. Estimated ages ranged from22–96. Every effort was made to get the bestage estimates possible, including comparisonwith those of similar but better-establishedages. Still, ages above 80 must be consideredwith some skepticism.

Permission to conduct the study wasobtained from the Kenyan government andthe local chiefs. Oral consent was obtainedfrom each subject. Subjects were paid a smallhonorarium for their participation.

Data collection

Nutritional status was assessed usingstandard anthropometric measures, includ-ing height, weight, mid upper arm, waist,and hip circumference. Six skinfold mea-sures were obtained: triceps, subscapular,mid-axillary, periumbilical, suprailliac, andmidcalf. Body fat was determined using abioelectric impedance (BIA) body fat scale(Tanita model BF-315). Derived measuresinclude BMI calculated as (wgt (kg)/hgt2 (m)),total lean mass ((1-%body fat)*weight),and waist to hip ratio (WHR), calculated aswaist circumference/hip circumference.

In addition to anthropometric data, menwere asked what they had to eat in the prior2 days. Some of the nomadic men weredrinking acacia tea, a fall-back food itemeaten during drought times when food isscarce. Ethnographic interviews revealedthat the nomads were dependent on foodaid and complained of not having enough toeat. In contrast, none of the settled malescomplained about lacking food.

Saliva samples were obtained from eachsubject both in the morning (7–9 AM) andthe evening (3–6 PM) using standard collec-

tion methods (Ellison, 1988). Samples werecollected en masse under the supervision ofthe investigator. Subjects were ask to refrainfrom eating or chewing tobacco prior to sam-ple collection. In addition, they rinsed theirmouth out with water immediately beforesample collection. Sodium azide was addedto each tube as a preservative after the sam-ple was collected.

Samples were transported to PeterEllison’s laboratory at Harvard Universitywhere testosterone was determined using amodified application of the I125 double anti-body kit produced by Diagnostic SystemsLaboratories (Webster, TX). Sample andstandard reactions were run in duplicate.Substrate (150 ml) was pipetted into borosili-cate tubes, 100 ml of sample and 50ml of buf-fered saline or, for the standard reactions, a400 pg/ml standard concentration was addedin volumes of 2.5, 12, 30, 75, and 150 ml, withvolumes of buffered saline adjusted to yield150 ml total volume. Antiserum, diluted 1:3(100 ml), and undiluted tracer (200 ml) wereadded to sample and standard tubes.Reactions incubated overnight for at least18 hours, after which precipitating reagent(400 ml) was added, tubes were centrifuged,and aspirated. The assays were sensitive to14 pmol/L T.

Statistical analysis

Differences in nutritional status, bodycomposition, and salivary T between settledand nomadic groups were compared usingt-tests. All available data points were usedfor these comparisons. In order to ensure thatvarying sample sizes for individual measuresdid not bias the results, analyses were alsorun on values from only those individualswith complete data. Similar statistical differ-ences were obtained. Thus, means based onall data points for individual measures arereported.

To assess age-related changes in salivary Tlevels and nutritional status, ordinary least-squares regression (OLS) models were usedwith both residence and age as predictors.An age by residence interaction was alsotested to determine if the slope of age-relatedchange differed between the two groups.

In order to determine if measures of nutri-tional status and body composition couldaccount for differences in salivary T levelsbetween the two subpopulations, OLSregression models were run with age and

SALIVARY TESTOSTERONE IN ARIAAL MEN 699

residence as controls and each of five anthro-pometric measures added separately. Thesemeasures include BMI, because of its use asa standard measure of nutritional status,percent body fat as a measure of wholebody fat, and lean mass as a measure oftotal nonfat weight. WHR was considered ameasure of abdominal fat, which has beenrelated to variation in testosterone(Jorgensen et al., 1996). Finally, height wasalso tested because it has been related tovariation in testosterone in at least twostudies (Ellison and Panter-Brick, 1996;Jamison et al., 1993).

RESULTS

Impact of chewing tobacco on salivarytestosterone

In order to determine if the tobacco thatforms a constant companion for men has ademonstrable effect on salivary T, 10 menspit into test tubes immediately before andafter chewing a wad of tobacco for 10 minutes.Table 1 shows the individual testosteronevalues before and after chewing tobacco.Individual values vary substantially.Average value for salivary T before chewingtobacco was 178.2̄ � 94.3 pmol/l compared to202.5̄ � 101.6 pmol/l after chewing. A pairedt-test indicates a significant difference insalivary T values before and after chewing(P ¼ 0.053). The magnitude of the averagedifference is sufficiently large to warrantconcern about the immediate impact oftobacco chewing on salivary T levels.However, given sample collection techniquesthis effect may be less important than more

long-term effects of chronic tobacco use on T(see Discussion).

Salivary testosterone

Individual salivary T values ranged from14 (lower limit of assay sensitivity) to422 pmol/l with an overall average of 176.8�̄ 74.8 pmol/l in the morning and 178.0¯�71.4 pmol/l in the afternoon. Overall, thevalues obtained here are similar to thosereported for non-Western populations(Bentley et al., 1993; Beall et al., 1992) andbelow those for a Western sample using avariety of different assays techniques(Dabbs et al., 1995), including results fromthe same assay (Gray, 2003; Burnham et al.,2003).

On the other hand, the fact that AM valueswere only 10.5 pm/L higher when comparedto PM values for the same individual (pairedt-test, t ¼ 0.74; P ¼ 0.46; n ¼ 106) is unusualcompared to previous results demonstratingdiurnal variation in both US (Dabbs, 1990),and non-Western populations (Gray, 2003).This lack of diurnal variation may reflectboth the low overall levels of salivary T inthis population, leaving less room fordecline, and the fact that PM sample werecollected very soon after 3:00 PM, minimizingthe amount of diurnal decline.

Subpopulation comparison

Table 2 compares average anthropometricand hormonal measures for the nomadic andsettled men. In terms of body composition,the overall sample is very lean, with an aver-age %body fat of 7̄ � 3.0%. Average BMI was17.8¯� 6.0kg/m2, less than the 18.5 kg/m2

considered a cut-off for chronic energy defi-ciency (Ferro-Luzzi et al., 1992). Averageheight, on the other hand, was 170̄ � 6.0 cm,placing it near the norm compared withother African populations, although wellbelow that of other pastoral nomadic groupssuch as the Masai (Eveleth and Tanner,1990).

Comparison of the two subpopulationsindicates that they do not exhibit a signifi-cant difference in height, hip circumference,WHR, BMI, or lean body mass. On theother hand, nomads exhibited significantlylower %body fat and smaller skinfolds,including triceps, subscapular, midaxillary,and suprailliac skinfold. The lack of a differ-ence in average height suggests similarity in

TABLE 1. Effects of tobacco: salivary T values before andafter chewing

Salivary testosterone (pmol/L)

ID # Before (n ¼ 10) After (n ¼ 10)

1 121 842 109 1233 129 1884 361 4105 325 3306 145 1497 144 1348 98 1649 121 181

10 229 262

Average* 178.2̄ � 94.3 202.5̄ � 101.6

P ¼ 0.053 using a paired t-test.


long-term growth and health status betweenthe two populations. However, nomads dohave less adipose stores, indicating pooreracute nutritional status compared to thesettled sample. This is the result of anongoing drought and poor food availabilityat the time of the study, as revealed in theethnographic interviews.

Comparison of salivary T between the twosubpopulations revealed a significant differ-ence in PM, but not AM, salivary T values.There were slight differences in the age dis-tribution of the two populations. Controllingfor these differences using multiple regres-sion does not substantially change theresults as presented and the t-test is usedfor the sake of clarity in exposition.

Age-related changes in body composition andsalivary testosterone

Figure 1 shows age-related changes inselected measures of nutritional status andbody composition among settled and nomadicmen. BMI shows a clear decline with age inboth subpopulations. However, the slope ofthe age-related decline in the nomadic popu-lation is steeper, as confirmed by a significantage by nomadic residence interaction term

(nomadic residence std. coeff. ¼ 0.337, P ¼0.16; age std. coeff. ¼ 0.079, P ¼ 0.76; noma-dic residence*age std. coeff. ¼ –0.698, P ¼0.046). %Body fat shows a strong declinewith age among the nomadic, but not thesettled, population. Regression analysis con-firms the strong difference in slopes betweenthe two subpopulations as well as marginalmain effects of age and residence (nomadicresidence std. coeff. ¼ 0.588, P ¼ 0.059;age std. coeff. ¼ 0.537, P ¼ 0.062; nomadicresidence*age std. coeff. ¼ –1.094, P ¼ 0.007).

In contrast to BMI and %body fat, leanbody mass shows relatively little age-relateddecline in either population. Regression ana-lysis shows only a marginal effect of age(nomadic residence std. coeff. ¼ –0.126, P ¼0.21; age std. coeff. ¼ –0.180, P ¼ 0.075). Noris there a difference in the two slopes, asindicated by the lack of a significant interac-tion term. Waist circumference shows a lackof age-related change altogether, althoughthere is a marginal difference between thetwo populations (nomadic residence std.coeff. ¼ –0.166, P ¼ 0.076; age beta ¼–0.075, P ¼ 0.42). An age * nomadic resi-dence interaction term was not significant.

Figure 2A shows AM salivary T values byage for both settled and nomadic males.

TABLE 2. Comparison of anthropometric and salivary T values in nomadic and settled males

Nomads SettledVariable Mean̄ � S.D. Mean̄ � S.D. P-value

Sample SizeAge (yrs)

n ¼ 5646.8̄ � 14.3

n ¼ 6149.6̄ � 18.7 N.S.

Height (cm) 170.1̄ � 6.0 171.0̄ � 6.4 N.S.Weight (lbs) 112.6̄ � 12.2 116.1̄ � 14.7 N.S.Waist (mm) 69.6̄ � 4.0 70.9̄ � 4.8 N.S.Hip (mm) 84.8̄ � 4.1 85.9̄ � 4.9 0.09WHR .83̄ � .04 .83̄ � .05 N.S.MUAC (mm) 22.9̄ � 1.9 23.6̄ � 2.2 N.S.BMI 17.6̄ � 1.6 18.0̄ � 1.8 N.S.Lean Body Mass (lbs)a 107.3̄ � 8.2 109.5̄ � 9.1 N.S.%Body Fata 6.2̄ � 3.1 7.7̄ � 3.8 0.04Triceps Skinfold (mm) 3.6̄ � 1.0 4.5̄ � 1.6 0.001Subscapular Skinfold (mm) 6.1̄ � 1.3 6.9̄ � 1.9 0.02SupraIliiac Skinfold (mm) 3.3̄ � .78 3.7̄ � .82 0.001Mid Axillary Skinfold (mm) 4.2̄ � .79 4.8̄ � 1.14 0.001Periumbilical Skinfold (mm) 4.8̄ � .95 5.8̄ � 2.0 0.001AM T (pmol/L)b 173.0̄ � 80.3 179.6̄ � 65.4 N.S.PM T (pmol/L)b 160.3̄ � 67.0 187.2̄ � 68.3 0.04

P value based on result of t-test comparing means of two subpopulations. Adjusting for age using multiple regression has little effect onsignificance of subpopulation differences.aSample size for %body fat and hence lean body mass is 49 and 52 for nomads and settled males, respectively.bSample size for AM and PM T is 46 and 52 for nomads and settled male, respectively.In order to be sure that missing values for %body fat, lean body, mass, and salivary T do not bias the significance of subpopulationcomparisons, t-tests were run on values from only individuals with complete data. They reveal no difference in the significancedifference between variables between subpopulation comparisons shown, here.


Multiple regression indicates no mean differ-ence between the two subpopulations, but anage-related decline is apparent overall(nomadic residence std. coeff. ¼ –0.074, P ¼0.436; age std. coeff. ¼ –0.315; P < 0.01). Theslope of the age-related decline is signifi-cantly greater in the nomadic population(age*nomadic residence std. coeff. ¼ –0.776;P ¼ 0.047). Figure 2B shows PM salivary Tvalues by age for both settled and nomadicmales. Multiple regression indicates differ-ent mean levels between the two subpopula-tions, but no overall age-related decline inPM T values (nomadic residence std. coeff. ¼–0.197, P ¼ 0.04; age std. coeff. ¼ 0.039,P ¼ 0.69). There is no significant differencein the slope of age-related changes betweenthe two subpopulations.

Relationship of salivary T and bodycomposition

Table 3a shows the relationship of AM sali-vary T and five anthropometric measures ofnutritional status and body composition,each tested separately. When controlling forage and nomadic residence in separate linearregression models, body fat and BMI areeach significant positive predictors of AM Tlevels. Furthermore, with the addition ofbody fat and BMI (in separate models), ageis no longer a significant predictor (as shownto be in the previous section), suggestingthat age-related declines in salivary T are aconsequence of loss of body fat with age. Incomparison, WHR and lean body mass arenot significant predictors of salivary T, while

Fig. 1. Age-related changes in body composition among Ariaal males.A: BMI. For the nomads R2 ¼ 0.281; P< 0.001.For the settled males R2 ¼ 0.092; P ¼ 0.02. B: Lean mass. For the nomads R2 ¼ 0.082; P ¼ 0.04. For the settled malesR2 ¼ 0.011; P ¼ 0.47. C: Waist circumference. For the nomads R2 ¼ 0.013; P ¼ 0.42. For the settled males R2 ¼ 0.00;P ¼ 0.97. D: Percentage body fat. For the nomads R2 ¼ 0.303; P < 0.001. For the settled males R2 ¼ 0.000; P ¼ 0.96.


age remains a significant predictor in thesemodels, suggesting that age-related declinesin T are not a function of changes in leanmass. Similar results were obtained forheight.

The addition of the age*nomadic residenceinteraction makes little difference to thecoefficients of the anthropometric correlates,and has been left out the analysis for ease ofinterpretation.

Table 3b shows the relationship of PM sali-vary T and five anthropometric measures ofnutritional status and body composition.Again, controlling for age and nomadic resi-dence, body fat, BMI, and WHR are all sig-nificant and positive predictors of PM T levelsin separate models, while lean body massand height show no significant association.Note that nomadic residence is not a signifi-cant predictor of salivary T, except in thecase of height. Since nomadic residence wasa significant predictor when in the modelwith age alone (as shown in the previous

section), differences in body fat, but notheight, can be inferred to be responsible forsubpopulation differences in PM salivary T.

Table 4 shows a final multivariate modelpredicting AM and PM salivary T. Both WHRand %body fat are included to determine ifthey are independent predictors of salivaryT. For AM salivary T, both %body fat andWHR are independent and significant posi-tive predictors when controlling for age andresidence. Age is also a significant negativepredictor. Table 4b shows similar models forPM salivary T, with similar results except forage, which is not a significant predictor, asexpected given the results of the models inTable 3.

DISCUSSION

Our results indicate that among Ariaalmen, whose BMIs are indicative of chronicenergetic stress, salivary T levels are lowcompared to those of men from Westernpopulations. Furthermore, we found thatsalivary T was related to BMI, as previouslyreported for the Lese of Zaire (Bentley et al.,1993). However, our finding that individualvariation in salivary T is positively related tovariation in %body fat and WHR, but notlean body mass, is novel. Furthermore, levelsof salivary T decline more rapidly with ageamong the nomads as does %body fat, butnot lean body mass. These findings are con-sistent with the hypothesis that lower levelsof testosterone among males in subsistencesocieties reflect poor nutritional status(Ellison, 2001), but fail to substantiate arelationship between testosterone andmuscle mass. These results have importantimplications for male life history theory(Bribiescas, 2001), suggesting that thesomatic effects of T may be more closelyrelated to survivability than to reproductiveeffort.

Tobacco and salivary T

Our results suggest that tobacco chewingdoes have an acute effect on salivary T.However, given that we asked men not toeat or chew tobacco prior to collecting salivasamples, we believe we minimized anyimmediate impact of chewing tobacco onour results. More important, however,other studies have demonstrated thatsmokers exhibit significantly higher serumtestosterone levels compared to nonsmokers.

Fig. 2. Age-related changes in salivary testosteroneamong Ariaal males. A: AM salivary T. For the nomadsR2 ¼ 0.18; P < 0.001. For the settled males R2 ¼ 0.38;P < 0.19. B: PM salivary T. For the nomads R2 ¼ 0.016;P ¼ 0.38. For the settled males R2 ¼ 0.024; P ¼ 0.23.


(Tamimi et al., 2001; Fischer et al., 1997)Thus, our overall salivary T values may bea conservative estimate of how low salivaryT is among Ariaal males. Given the ubiquityof tobacco chewing among the Ariaal, it isunlikely that differences in the frequency oftobacco chewing between the two subpopula-tions would account for differences in aver-age salivary T values, although chewingtobacco during the day could help to accountfor the relatively small difference betweenAM and PM T values observed in our data.

Body composition and salivary T

The positive association of %body fatand WHR with both AM and PM salivaryT levels among Ariaal men runs counter tofindings from Western populations in whichBMI, body fat, and WHR are negativelyrelated to testosterone levels (Jorgensenet al., 1996), particularly in obese men (Allenet al., 2002). In Western populations withmore than adequate energy availability,intrinsic testosterone levels play a role in

TABLE 3. Anthropometric predictors of salivary testosterone

a) AM salivary testosterone

Overall regression model

n 100 84 84 101 101Adj. R2 0.11 0.11 0.04 0.09 0.02

Predictors Std. Coeff. Std. Coeff. Std. Coeff. Std. Coeff. Std. Coeff.

Nomadic �0.042 �0.041 �0.044 �0.045 �0.066Age �0.195þ �0.168 �0.269* �0.276** �0.309**BMI 0.220*%Body fat 0.295*Lean mass 0.008WHR 0.147Height �0.063

þP < 0.1; *P < 0.05; **P < 0.01.

b) PM salivary testosterone


n 110 95 95 110 110Adj. R2 0.12 0.12 0.04 0.09 0.02

Predictors Std. coeff. Std. coeff. Std. coeff. Std. coeff. Std. coeff.

Nomadic �0.151 �0.161 �0.206 �0.147 �0.197*Age �0.178þ 0.129 0.101 �0.062 �0.037BMI 0.351***%Body fat 0.310**Lean mass 0.161WHR 0.261**Height (m) 0.006

þP < 0.1; *P < 0.05; **P < 0.01; ***P < 0.001.

TABLE 4. Multivariate predictors of salivarytestosterone

a) AM salivary testosterone


n 84Adj. R2 0.16

Predictors Std. coeff.

Nomadic 0.034Age �0.227*%Body fat 0.234*WHR 0.242*

þP < 0.1; *P < 0.05; **P < 0.01.

b) PM salivary testosterone


n 95Adj. R2 0.14

Variables Std. coeff.

Residence �0.157Age 0.089Body Fat 0.263*WHR 0.184þ

þP < 0.1; *P < 0.05; **P < 0.01.


body composition. Low testosterone levelsare related to lower levels of muscle develop-ment and energy usage, resulting in deposi-tion of caloric intake as fat (Smith, 1996;Haffner et al., 1994). In contrast, among theAriaal, who experience chronic undernutri-tion, the direction of causality appears to bereversed, and testosterone reflects nutri-tional status. Undernutrition means bothlower energy intake and less adipose tissue,both of which may have an impact on sali-vary T levels, as discussed below.

Because salivary T reflects free orunbound testosterone (Wang et al., 1981;Navarro et al., 1986), the relationshipbetween acute nutritional status and testo-sterone levels reported here may representtwo related effects; reduced production oftestosterone, or decreased bioavailability oftestosterone because of elevated SHBGlevels. Chronic undernutrition is relatedto reduced responsiveness of the testes togonadotropin stimulation (Bribiescas, 2001;Smith, 1975), and low blood sugar has beenrelated to reduced LH stimulation of thetestes (Oltmann et al., 2001). In addition,previous findings among the Turkana, pas-toral nomads of northern Kenya, indicatednormal T levels, but elevated SHBG amongmen (Campbell, unpubl. data) with BMIssimilar to those of the Ariaal.

Both of these potential mechanisms couldbe the result of altered insulin production,which is thought to stimulate testosteroneproduction and suppress SHBG production(Haffner, 1996). Low BMI has been asso-ciated with lower levels of fasting insulin(Strain et al., 1994; Tymchuk et al., 1998)and low birthweight with reduced insulinsecretion in young males (Jensen et al.,2002). In turn, insulin levels have beeninversely related to SHBG levels in bothcross-sectional samples (Pasquali et al.,1995; Vermeulen et al., 1996) and experi-mental studies (Katsuki et al., 1996).

Salivary T and age

In men from Western populations, adipos-ity increases with age (Garn, 1994; Stini,1994) and the loss of muscle mass is consid-ered a hallmark of senescence (Lambertset al., 1997). In addition, age is also associatedwith increasing levels of SHBG (Longcopeet al., 1990), leading to a faster decline in freeas opposed to total serum T (Nahoul andRoger, 1990; Tenover, 1997). However, our

results indicate that nomadic Ariaal menexhibit a different age-related pattern ofchanges in body composition. BMI declineswith age because of a decrease in adiposity,while lean mass remains relatively constant.Thus, declines in salivary T with age aremore closely related to declines in adiposity,rather than lean muscle mass.

The independent relationship of age andAM salivary T levels is consistent with sen-escent declines in GnRH production andgonadotropin stimulation of the testes asdocumented in Western clinical studies(Veldhuis, 1999; Vermeulen and Kaufman,1995). Because GnRH production is increasedat night, age-related declines in hypothalamicpituitary function are hypothesized to bemore evident in the morning. In contrast,afternoon testosterone levels may reflectresponsiveness to activities during the day,including subsistence activity (Worthmanand Konner, 1987), food intake (Habito andBall, 2001), and behavioral challenges (Mazurand Booth, 1998; Cohen et al., 1996; Suarezet al., 1998), thereby obscuring the underly-ing age-related decline in GnRH stimulation.Hence, the lack of a significant relationshipbetween PM salivary T and age.

Subpopulation comparison

The fact that nomadic Ariaal men exhib-ited smaller skinfold, but were no shorterthan their settled counterparts, nor hadslightly lower lean body mass, suggestsdifferences in acute, but not chronic,nutritional status between the two sub-populations. The results of our multivariateanalyses showing that BMI and %bodyfat, but not height, replace residence as asignificant predictor of PM salivary T suggestthat differences in current energy status,and not long-term growth, explain differ-ences in salivary T between the two sub-populations. This conclusion finds supportin the fact that the two subpopulations sharea common history until the early 1960s,when lowland members of the Ariaal settledin the highland region surrounding Karareas part of a missionary-led scheme (ethno-graphic notes). Thus, differences in the age-related decline in salivary T among the oldermen are unlikely to be due to developmentaldifferences during childhood and adolescentdevelopment.

Other possible explanations of subpopulationdifferences in salivary T include differences in


psychosocial stress (Nilsson et al., 1995;Christiansen et al., 1985), activity patterns(Worthman and Konner, 1987), infection andinflammatory processes (Campbell et al., 2001),as well as dietary differences (Hill et al., 1979,1980; Christiansen, 1991a), all of which,excluding psychosocial stress, have beenrelated to testosterone levels in African popu-lations. Unfortunately, we are unable to testany of these possibilities directly.

Comparison with other populations

Our finding of lower afternoon salivaryT among nomadic relative to settled malescontrasts with those from the Turkana,where nomadic males showed slightly higherblood testosterone level than their settledcounterparts, despite poorer nutritionalstatus (Campbell et al., 2001). However, theTurkana results were based on whole blood,which reflects total testosterone ratherthan salivary T, which reflects free T levels(Baxendale et al., 1982; Navarro et al., 1986).Among the Turkana, SHBG, as well as testo-sterone levels, were also higher among thenomads than the settled males (Campbell,unpubl. results), suggesting that free T wasnot elevated, more consistent with ourresults here.

Further comparison of our finding with thosefrom nonpastoralist populations (Bentley et al.,1993; Bribiescas, 1997; Ellison and Panter-Brick, 1996), who have not reported an asso-ciation between body fat and salivary T, maybe limited by the field measures of nutri-tional status. BMI may be a somewhatmisleading measure among groups in aridregions, because of differences in relativelength of legs and torso (Norgan, 1994). Inaddition, the Tanita scale used in this studymay give biased estimates of body fat for non-Western populations, since it determinesbody composition using BIA and equationscalibrated on North American referencedata.

However, comparison among men fromboth Zimbabwe and Siberia shows that mea-sures of %body fat based on the Tanita scaleand those derived from skinfolds are highlyrelated (Campbell et al., n.d. unpubl. ms.).Thus, measures of body fat based on BIAmay provide an adequate approximation ofbody fat for investigation the relationship ofsalivary T and body composition across non-Western populations of particular interest toanthropologists.

SUMMARY

Our results suggest a clear relationshipbetween nutritional status and salivary Tamong Ariaal males. We attribute differ-ences in PM salivary T among nomadic andsettled population to acute undernutritionamong the nomads, which may have effectson both testosterone production and bioa-vailability. Furthermore, the impact ofacute undernutrition appears to result inmore pronounced loss of adipose tissueamong the older nomadic men, thus height-ening an age-related decline in AM salivary T.On the other hand, given that the nomadicmen have significantly smaller skinfolds, butshow significantly lower salivary T levelsonly in the afternoon, our results are similarto early studies in suggesting that malereproductive function is relatively insen-sitive to fluctuations in current energy avail-ability (Panter-Brick and Ellison, 1996).Thus, lower levels of salivary T amongAriaal males relative to Western samplesmay reflect development of the hypothala-mic pituitary under poor long-term nutri-tional conditions, a part of male life historythat deserves further research in its ownright.

ACKNOWLEDGMENTS

We thank Daniel Lemoille and KoreaLeala for help in collecting the data; PeterGray, Rick Bribiescas, William Lukas, andthe Reproductive Ecology Reading Groupfor comments and encouragement. We espe-cially thank the people of Karare andLewegoso without whom this work wouldnot have been possible.

LITERATURE CITED

Allen NE, Appleby PN, Davey GK, Key TJ. 2002.Lifestyle and nutritional determinants of bioavailableandrogens and related hormones in British men.Cancer Causes Control 13:353–363.

Baxendale PM, Jacobs HS, James VHT. 1982. Salivarytestosterone: relationship to unbound plasma testos-terone in normal and hyperandrogenic women. ClinEndocrinol 16:595–603.

Beall CM, Worthman CM, Stallings J, Strohl KP,Brittenham GM, Barragan M. 1992. Salivary testos-terone concentration of Aymara men native to 3600m.Ann Hum Biol 19:67–78.

Bentley GR, Harrigan AM, Campbell BC, Ellison PT.1993. Seasonal effects on salivary testosterone levelsamong Lese males of the Ituri forest, Zaire. Am J HumBiol 5:711–717.


Bribiescas RG. 1996. Testosterone levels among Achehunter/gatherer men; a functional interpretation ofpopulation variation among adult males. Hum Nat7:163–188.

Bribiescas RG. 1997. Testosterone as a proximate deter-minant of somatic energetic energy allocation inhuman males: evidence from Ache men of easternParaguay. Ph.d. thesis, Harvard University.

Bribiescas RG. 2001. Reproductive ecology and life historyof the human male. Yrbk Phys Anthropol 44:148–176.

Burnham TC, Flynn-Chapman J, Gray PB, McIntyre MH,Lipson SF, Ellison PT. 2003. Men in committed,romantic relationships have lower testosterone. HormBehav (in press).

Campbell KL. 1994. Blood, urine, saliva and dip-sticks:experiences in Africa, New Guinea, and Boston. In:Campbell KL, Wood JW, editors. Human reproductiveecology: interactions of environment, fertility andbehavior. Ann NY Acad Sci 709:312–330.

Campbell BC, Leslie PW 1995. Reproductive ecology ofhuman males. Yrbk Phys Anthropol 38:1–26.

Campbell BC, Lukas W, Campbell KL. 2001. Reproduc-tive ecology of male immune function and gonadal func-tion. In: Ellison PT, editor. Reproductive ecology andhuman evolution. Hawthorne, NY: Aldine de Gruyter.p 159–178.

Christiansen K. 1991a. Sex hormone levels, diet, andalcohol consumption in Namibian Kavango men.Homo 42:43–62.

Christiansen K. 1991b. Serum and saliva sex hormonelevels in !Kung San men. Am J Phys Anthropol 86:36–44.

Christiansen K, Knussmann R, Couwenbergs C. 1985.Sex hormones and stress in the human male. HormBehav 19:426–440.

Cohen D, Bowdle BF, Nisbett RE, Schwartz N. 1996. Insult,aggression and the Southern culture of honor: an ‘‘experi-mental ethnography.’’ J Pers Soc Pyschol 70:945–960.

Dabbs JM Jr. 1990. Salivary testosterone measure-ments, reliability across hours, days, weeks. PhysiolBehav 48:83–86.

Dabbs JM Jr., Campbell, BC, Gladue BA, Midgley AR,Navarro MA, Read FR, Susman EJ, Swinkles LMJW,Worthman C. 1995. Reliability of salivary testosteronemeasurements: a multicenter evaluation. Clin Chem41:1581–1584.

Ellison PT. 1988. Human salivary steroids: methodo-logical considerations and applications in physicalanthropology. Yrbk Phys Anthropol 31:1115–1142.

Ellison PT. 2001. On fertile ground: a natural historyof human reproduction. Cambridge, MA: HarvardUniversity Press.

Ellison PT, Panter-Brick C. 1996. Salivary testosteronelevels among Tamang and Kami males of centralNepal. Hum Biol 68:955–965.

Ellison PT, Bribiescas RG, Bentley GR, Campbell BC,Lipson SF, Panter-Brick C. 2002. Population variationin age-related decline in male salivary testosterone.Hum Reprod 17:3251–3253.

Eveleth PB, Tanner JM. 1990. Worldwide variationin human growth, 2nd ed. New York: CambridgeUniversity Press.

Ferro-Luzzi A, Sette S, Franklin M, James WPT. 1992. Asimplified approach of assessing adult chronic energydeficiency. Eur J Clin Nutr 46:173–186.

Fisher CL, Mannino DM, Herman WH, Frumkin H.1997. Cigarette smoking and thyroid hormone levelsin males. Int J Epidemiol 26:972–977.

Fratkin E. 1998. Ariaal pastoralists of Kenya: survivingdrought and development in Africa’s arid lands.Boston: Allyn & Bacon.

Fratkin EM, Roth EA, Nathan MA. 1999. When nomadssettle: the effects of commodization, nutritional change,and formal education on Ariaal and Rendille pastoral-ists. Curr Anthropol 40:729–735.

Garn SM. 1994. Fat, lipid, and blood pressure changes inadult years. In: Crews DE, Garruto RM, editors.Biological anthropology and aging. New York: OxfordUniversity Press. p 301–320.

Gray PB. 2003. Marriage, parenting and testosteronevariation among Kenyan Swahili Men. Am J PhysAnthropol (in press).

Guerra-Garcia R, Velasquez A, Coyotupa J. 1969. A testof endocrine gonadal function in men: urinary testo-sterone after the injection of HCG. II. A differentresponse of the high altitude native. J Clin EndocrinolMetabol 29:179–182.

Habito RC, Ball MJ. 2001. Postprandial changes in sexhormones after meals of different composition.Metabolism 50:505–511.

Haffner SM. 1996. Sex hormone-binding protein, hyper-insulinemia, insulin resistance and noninsulin-depen-dent diabetes. Hormone Res 45:233–237.

Haffner SM, Karhapaa P, Mykkanen L, Laakso M. 1994.Insulin resistance, body fat distribution, and sexhormones in men. Diabetes 43:212–219.

Hill P, Wynder EL, Garbaczewski L, Garnes H, WalkerARP. 1979. Diet and urinary steroids in black andwhite North American men and black South Africanmen. Cancer Res 39:5101–5105.

Hill P, Wynder EL, Garbaczewski L, Garnes H, WalkerARP. 1980. Environmental factors, hormone status,and prostatic cancer. Prev Med 9:657–666.

Jamison CS, Meier RJ, Campbell BC. 1993. Dermato-glyphic asymmetry and testosterone levels in normalmales. Am J Phys Anthropol 90:185–198.

Jensen CB, Storgaard H, Dela F, Holst JJ, Madsbad S,Vaag AA. 2002. Early differential defects of insulinsecretion and action in 19-year-old Caucasian menwho had low birth weight. Diabetes 51:1271–1280.

Jorgensen JOL, Vahl N, Hansen TB, Fisker S, Hagen C,Christansen JS. 1996. Influence of growth hormonesand androgens on body composition in adults. HormRes 45:94–98.

Katsuki A, Sumida Y, Murashima S, Fujii M, Ito K,Tsuchihashi K, Murata K, Yano Y, Shima T. 1996.Acute and chronic regulation of serum sex hormone-binding globulin levels by plasma insulin concentra-tions in male noninsulin-dependent diabetes mellituspatients. J Clin Endocrinol Metabol 81:2515–2519.

Lamberts SWJ, van den Beld AW, van der Lely A. 1997.The endocrinology of aging. Science 278:419–424.

Longcope C, Feldman HA, McKinley JB, Araujo AB.2000. Diet and sex hormone binding globulin. J ClinEndocrinol Metabol 85:293–296.

Mazur A, Booth A. 1998. Testosterone and dominance inmen. Behav Brain Sci 21:353–364.

Nahoul K, Rodgers M. 1990. Age-related decline ofplasma bioavailable testosterone in adult men. JSteroid Biochem 35:293–299.

Nathan MA, Fratkin E, Roth EA. 1996. Sedentism andchild health among Rendille pastoralists of northernKenya. Soc Sci Med 43:503–515.

Navarro MA, Juan L, Bonnin MR. 1986. Salivary testos-terone: relationship total and free testosterone inserum. Clin Chem 32:231–232.

Nilsson PM, Moller L, Solstad K. 1995. Adverse effects ofpsychosocial stress on gonadal function and insulinlevels in middle-aged males. J Intern Med 237:479–486.

Norgan NG. 1994. Interpretation of low body mass indices:Australian aborigines. Am J Phys Anthropol 94:229–237.


Oltmann KM, Fruehwalkd-Schultes B, Kern W, Born J,FehmHL, Peters A. 2001. Hypoglycemia, but not insu-lin, acutely decreases LH and T secretion in men.J Clin Endocrinol Metabol 86:4913–4919.

Pasquali R, Casimirri F, De Iasio R, Mesini P, Boschi S,Chierici R, Flamia R, Biscotti M, Vicennati V. 1995.Insulin regulates testosterone and sex hormone-bind-ing globulin concentrations in adult normal weight andobese men. J Clin Endocrinol Metabol 80:654–658.

Shell-Duncan B, Yung S. 2003. The maternal depletiontransition in Northern Kenya: The effects of settlement,development and disparity. Soc Sci Med (in press).

Smith SR. 1996. The endocrinology of aging. EndocrinolMetab Clin N Am 13:215–229.

Smith SR, Chetri MK, Johanson J, Radfar N, Migeon CJ.1975. The pituitary-gonadal axis in men with protein-calorie malnutrition. J Clin Endo Metab 41:60–69

Stini WA. 1994. Nutrition and aging: Intraindividualvariation. In Biological Anthropology and Aging:Perspectives on Human Variation over the Life Span.D.E. Crews and R.M. Garruto (eds.) Oxford UniversityPress. New York. Pp. 232–271.

Strain B, Zumoff B, Rosner W, Pi-Sunyer X. 1994. Therelationship between serum levels of insulin and sexhormone binding globulin in men: the effect of weightloss. J Endocrinol Metabol 79:1173–1176.

Suarez EC, Kuhn CM, Schanbery SM, Williams RB Jr,Zimmerman EA. 1998. Neuroendocrine, cardiovascu-lar, and emotional responses of hostile men; the role ofinterpersonal challenge. Psychosom Med 60:78–88.

Tamimi R, Mucci LA, Spano E, Lagiou A, Benetou V,Trichopoulos D. 2001. Testosterone and oestradiol inrelation to tobacco smoking, body mass index, energyconsumption and nutrient intake among adult men.Eur J Cancer Prev 10:275–280.

Tenover JS. 1997. Testosterone and the aging male.J Androl 18:103–106.

Tymchuk CN, Tessler SB, Aronson WJ, Barnard RJ.1998. Effects of diet and exercise on insulin, sex hor-mone-binding globulin and prostate-specific antigen.Nutr Cancer 31:127–131.

Veldhuis JD, Iranmanesh A, Mulligan T, Pincus SM.1999. Disruption of the young-adult synchronybetween luteinizing hormone release and oscillationsin follicle-stimulating hormone, prolactin, and noctur-nal penile tumescence (NPT) in healthy older males.J Clin Endocrinol Metabol 84:3498–3505.

Vermeulen A, Kaufman JM. 1995. Aging of the hypothala-mic-pituitary-testicular axis in men. Horm Res 43:25–28.

Vermeulen A, Kaufman JM, Giagulli VA. 1996. Influenceof some biological indexes on sex hormone-bindingglobulin and androgen levels in aging or obese males.J Clin Endocrinol Metabol 81:1821–1826.

Vismos BC, Marti-Henneberg C. 2000. Puberty beginswith a characteristic body fat mass in each sex. EurJ Clin Nutr 54:203–208.

Wang C, Plymate S, Nieschlag E, Paulsen CA. 1981.Salivary testosterone in men: further evidence of adirect correlation with free serum testosterone. J ClinEndocrinol Metabol 53:1021–1024.

Worthman CM, Konner MJ. 1987. Testosterone levelschanges with subsistence hunting effort in !KungSan men. Pyschoneuroendocrinology 12:449–458.

Zemel BS, Worthman CM, Jenkins C. 1993. Differencesin endocrine status associated with urban rural pat-terns of growth and maturation in Bundi (Gendespeaking) adolescents of Papua New Guinea. In:Schell LM, Smith MT, Bilsborough A, editors. Urbanecology and health in the Third World. New York:Cambridge University Press. p 38–60.


salivary testosterone and body composition among ariaal males

Documents