Association between testosterone and estradiol and age-related decline in physical function in a diverse sample of men

Abstract

Objectives: To examine the association between aging and physical function in men by testing a theoretically based model of aging, hormones, body composition, strength, and physical function with data obtained from men enrolled in the Boston Area Community Health/Bone (BACH/Bone) Survey.

Design: Cross-sectional, observational survey.

Setting: Population-based.

Participants: Eight hundred ten black, Hispanic, and white randomly selected men from the Boston area aged 30 to 79.

Measurements: Testosterone, estradiol, sex hormone-binding globulin, lean and fat mass, grip strength, and summated index of physical function (derived from walk and chair stand tests).

Results: Measures of grip strength and physical function declined strongly with age. For instance, 10 years of aging was associated with a 0.49-point difference (scale 0-7) in physical function. Age differences in total testosterone and estradiol concentrations were smaller than age differences in their free fractions. Weak or nonsignificant age-adjusted correlations were observed between hormones and measures of physical function, although path analysis revealed a positive association between testosterone and appendicular lean mass and a strong negative association between testosterone and total fat mass. Lean and fat mass, in turn, were strongly associated with grip strength and physical function, indicating the possibility that testosterone influences physical function via indirect associations with body composition.

Conclusion: The age-related decline in serum testosterone concentration in men has a weak association with physical strength and functional outcomes through its associations with lean and fat mass.

FIGURE 1

Path model depicting multiple paths from age to physical function. Hypothesized associations are depicted in the full model drawing (upper left). The model asserts that the association between age and physical function lies along many potential pathways, for instance through physical activity, which itself may influence body composition, strength, and physical function via direct and indirect means (upper right). Total estimated associations (displayed here accompanied by 95% confidence intervals) between age or testosterone and downstream outcomes may be decomposed into components accounted for by the different covariate groups (smaller drawings). The first row of these depicts the model’s decoupling of the influence of age on strength into components lying along the hormonal pathway (including portions mediated by hormones’ influence on body composition), physical activity (including portions mediated by the relation between activity and body composition), body composition (discounting the effects that pass through it via the hormonal and physical activity mechanisms), and the residual, “direct” association of age with strength that may be obtained from Table 5. In this way the model demonstrates that while much of the association between age and strength can be accounted for by the combined influences of hormones, activity, and body composition, a substantial direct association remains. All of these factors combined account for a relatively small proportion of the association between age and physical function. Meanwhile, the positive association between testosterone and strength as mediated by lean mass is partially offset by the negative associations mediated by fat mass, so that the overall relation between testosterone and strength is non-significant (lower left). At the same time, the bulk of the significant positive association between testosterone and physical function appears to be mediated by the negative association between fat mass and testosterone (lower right). Abbreviations: appendicular lean mass (ALM), estradiol (E2), fat mass (FM), luteinizing hormone (LH), testosterone (TT).

FIGURE 2

Physical function score versus age, with jitter applied for visual clarity. The close agreement between means by decade (grey circles, with confidence intervals), a GAM fit via penalized smoothing splines (black line), and a simple linear regression fit (grey line), indicates that linear regression provides a reasonable summary of the cross-sectional association between age and physical function.