Population variation in age-related decline in male salivary testosterone

Abstract

BACKGROUND: Age-related declines in free and bioavailable testosterone are frequently reported for Western populations, but interpopulation variation in this pattern has not previously been investigated. METHODS: Salivary testosterone was measured using a consistently applied protocol on morning samples collected from men in four populations (USA, Congo, Nepal, and Paraguay) representing different geographical, ecological, and cultural settings. RESULTS: Mean testosterone levels varied significantly between the four populations. The mean testosterone differences between populations were greatest for young men (aged 15–30 years) and insignificant for older men (aged 45–60 years). The slope of age-related decline in testosterone was significant in the USA and Congolese participants, but not in the Nepalese or Paraguayan participants. CONCLUSIONS: Age patterns of testosterone decline vary between populations primarily as a result of variation in the peak levels attained in young adulthood. The potential consequences of this variation for other aspects of male health deserve investigation.

ageing, population variation, salivary testosterone, testosterone

Introduction

Although formerly controversial, an age-related decline in free and bioavailable testosterone in healthy males is now a frequently reported pattern in Western populations (Ferrini and Barrett-Connor, 1998; Vermeulen et al., 1999; Harman et al., 2001). Age-related declines in testosterone levels in males have been associated with changes in body composition, including increases in fat mass, decreases in muscle mass, and decreases in bone mineral density (Denti et al., 1999; Vermeulen et al., 1999; Kenny et al., 2000; van den Beld et al., 2000). Reflecting these linkages to body composition, age-related declines in male testosterone are also associated with elevated risk of type II diabetes (Stellato et al., 2000; Haffner et al., 1996), coronary heart disease (Simon et al., 1997; De Pergola et al., 1997), ischemic stroke (Jeppesen et al., 1996) and osteoporosis (Francis, 1999; Snyder et al., 1999a). Low testosterone levels in older men are also associated with low libido, erectile dysfunction, and depression (Barrett-Connor et al., 1999; Schweiger et al., 1999; Wang et al., 2000). On the other hand, high testosterone levels have been associated with elevated risk and progress of prostate cancer (Crawford, 1992; van Tinteren and Dasio, 1993). Testosterone supplementation of older men has been shown to result in increases in lean-body mass, decreases in fat mass, increases in bone mineral density, and improvements in mood and sexual function (Bhasin and Tenover, 1997; Snyder et al., 1999a,b; Winters, 1999).

Despite evidence of its clinical importance, relatively little is known about population variation in age patterns of testosterone. This contrasts strikingly with the increasing awareness of population variation in female ovarian steroid profiles and its relationship to disease risk (Ellison et al., 1993; O'Rourke and Ellison, 1993; Ellison, 1999; Jasienska et al., 2000; Jasienska and Thune, 2001). There are scattered reports of male testosterone levels from non-Western, non-clinical populations, but the comparability of these data is limited by differences in sample collection, handling, and assay procedures (Guerra-Garcia et al., 1969; Smith et al., 1975; Christiansen, 1991a,b; Beall et al., 1992; Campbell, 1994).

This report presents data on age variation in male salivary testosterone values from four populations spanning broad genetic, ecological, and life style differences. All the samples were collected using the same protocols and were assayed in the same laboratory using consistent methods. The data are used for a preliminary test of the hypothesis that significant population variation exists in the pattern of age-related decline in male testosterone. By analogy with findings for women (Ellison et al., 1993), we expect that non-Western males will have lower levels of testosterone in adulthood and slower rates of decline in testosterone with increasing age.

Materials and methods

Participants in this study represent males from four populations that are geographically, genetically, ecologically, and culturally distinct: Lese horticulturalists from the Ituri Forest, Democratic Republic of the Congo (formerly Zaïre), (Bentley et al., 1993; n = 33); Tamang agropastoralists from central Nepal (n = 39; Ellison and Panter-Brick, 1996); Ache foragers from southern Paraguay (Bribiescas, 1996; n = 45) and residents of Massachusetts, USA (n = 106). Lese, Tamang, and Ache participants represent all available willing males in the corresponding study areas. The USA participants were recruited by public advertisement, thus the potential for self-selection bias should be noted. Age data were derived from detailed demographic interviews in the Congo, Nepal, and Paraguay and from self-report in the USA.

In all cases participants provided between one and five morning saliva samples collected within 2 h of waking from 1–5 days according to established collection protocols which minimize the risk of contamination with serum or gingival fluid (Lipson and Ellison, 1989). The samples were preserved with sodium azide until they could be transferred to the laboratory and frozen at –20°C. Testosterone levels have been found to be stable under these collection and handling procedures for up to 6 months (Lipson and Ellison, 1989). Testosterone values were determined for each sample by a tritium-based radioimmunoassay according to published protocols (Ellison et al., 1989). The sample values for individual participants were averaged to provide mean values for each subject.

The age pattern of testosterone within each population was analysed using simple linear regression. Sample sizes were too small to justify higher order curve fitting. Population variation in average testosterone values is assessed by analysis of variance and by multiple linear regression with age, population, and age multiplied by population interaction as independent variables. For the purposes of the multiple regression population was re-coded as a binary variable, USA and non-USA.

Results

The mean salivary testosterone values for men in the four populations were significantly different (ANOVA, P = 0.0002, Table I) although they fall within the broad normal range (100–1000 pmol/l) for adult males in all cases. However, the differences are not consistent across age groups and show convergence with age. Population differences are greatest among men under the age of 30 years (ANOVA, P = 0.0001) and are non-significant for men 45–60 years.

A linear regression of testosterone on age for the combined data from all four populations indicates a highly significant negative relationship (y = 332 – 2.0x, P = 0.0001). The individual population regressions of testosterone against age are significant in the USA (y = 418 – 3.4x, P = 0.0001) and Congo (y = 346 – 2.3x, P = 0.03), but not in Nepal (y = 286 – 1.5x, P = 0.39) or Paraguay (y = 218 – 0.7x, P = 0.50). A multiple linear regression model with age, population (recoded as USA and non-USA), and age×population interaction as independent variables indicates significant effects of all three variables (y = 549 – 5.2×Age – 131.5×Population + 1.7×Age×Population, multiple R = 0.47, P = 0.0001. Significance of individual variables: Age, P = 0.0001; Population, P = 0.0002; Age×Population, P = 0.05).

Discussion

The data presented here provide the first opportunity to consider population variation in the relationship of male testosterone to age. All the data were collected according to the same protocols and analysed with the same assay methods in the same laboratory. This common methodology helps to control for many of the sources of variation that may confound the comparison of results from different studies. The most important general observation from these data is that age-related decline in free testosterone, as represented by salivary levels, is not a uniform characteristic of all populations. In particular, the rate of decline with age varies significantly between populations and in some populations (i.e. Nepal and Paraguay) does not achieve statistical significance, although this may be due to the sample size limitations in this study.

The testosterone levels of older men (>45 years) tend to converge for all the populations studied while the levels of younger men (<30 years) vary significantly. This suggests that variability in the pattern of testosterone decline with age, results from population variation in the reproductive physiology of young males rather than in population differences in the rate of reproductive senescence among older males. If so, research into the mechanisms underlying age patterns of male testosterone should focus on the determinants of young adult levels as well as the causes of age-related decline.

Because population differences in age patterns of testosterone have not previously been appreciated, the implications of such variation for patterns of health risk have not been considered. If, for example, set-points for muscle anabolism and bone mineral density are established relative to testosterone levels in young adulthood, a steeper decline in testosterone from higher young-adult levels might result in more rapid age-related changes in male body composition, bone mineral density, and related health risks. Higher testosterone exposure throughout adulthood in populations with high young-adult levels may also lead to greater prostate cancer risk. These possibilities provide additional motivation for increased efforts to study global variation in male gonadal function to complement studies of Western populations.

Table I.

Mean (pmol/l) salivary testosterone levels (T) from men in four populations, for total sample and for three age groups.

Population	All ages			15 to < 30 years			15 to < 45 years			45–60 years
	T (pmol)	SEM	n	T (pmol)	SEM	n	T (pmol)	SEM	n	T (pmol/l)	SE	n
P-value is given for one-way ANOVA; NS = not significant.
USA	259	10	106	335	20	24	288	17	31	238	14	26
Congo	268	12	33	286	15	17	250	19	11	247	34	5
Nepal	240	13	37	251	18	22	224	19	14	225	0	1
Paraguay	192	12	45	197	25	15	187	14	21	192	36	8
P-value	0.0002			0.0001			0.0004			NS

7

To whom correspondence should be addressed at: Department of Anthropology, Peabody Museum, Harvard University,11 Divinity Avenue, Cambridge, MA 02138 USA. E-mail: