The authors state that they have no conflicts of interest.
Published online on December 29, 2008
This study investigates determinants of peak bone mass (PBM) in healthy men, focusing on effects and interactions of parameters reflecting mechanical loading and sex steroids. Healthy male siblings (n = 677; 25–45 yr) were recruited in a cross-sectional, population-based study. Physical activity score was assessed by a self-reported questionnaire. Cross-sectional muscle area (CSMA) and bone parameters of radius (4% and 66% site) and tibia (66% site) were assessed using pQCT. Peak torque of biceps and quadriceps muscles was assessed by isokinetic dynamometry. Serum testosterone (T) and estradiol (E2) levels were measured using immunoassays; free hormone fractions were calculated. Relations between indices of bone strength, CSMA, muscle strength, and sex steroids were studied using linear mixed-effects modeling. Physical activity, CSMA, and muscle strength were positively associated with indices of bone strength, except for volumetric BMD (vBMD). After controlling for age, weight, and height, free E2 levels were positively associated with trabecular and cortical vBMD, negatively associated with endosteal circumference at the radius, and positively associated with cortical vBMD at the tibia. In addition, positive interactions between physical activity and serum E2 concentrations were observed for bone size at the tibia. No associations between free T levels and pQCT bone parameters were found. In this population of healthy men at the age of PBM, parameters reflecting mechanical loading are confirmed as important determinants of bone size. E2, but not T, levels are positively associated with vBMD and modulate the impact of physical activity on bone size at the tibia.
Osteoporosis is a common disorder affecting both sexes, with a lifetime risk of sustaining a fragility fracture at age 50 yr estimated at 40–50% in women and 13–22% in men.(1) The age-specific fracture risk in men is about one half that in women, which results in part from the achievement of a higher bone mass, in particular of larger bone size during growth.(2) Because bone mass and strength in the elderly are highly dependent on peak bone mass (PBM),(3,4) it is important to understand the determinants of their main components and to identify risk factors for impaired PBM.
According to the mechanostat theory,(5) mechanical loading is one of the most important determinants of bone strength. This is reflected by strong and consistent associations between indices of bone strength and parameters reflecting physiologic mechanical loading, such as physical activity level,(6,7) muscle strength,(8–10) muscle mass,(11) or size.(12–14) Moreover, the effects of mechanical loading on bone seem to persist after the pubertal period.(15,16) Conversely, disorders primarily affecting muscles are associated with low bone mass.(17,18)
Sex steroid hormones also play an important role in pubertal bone development and in maintenance of bone mass during aging.(19–21) In men, this is reflected by associations of serum sex steroid levels with BMD,(22–32) bone size,(22,26,27) longitudinal changes in BMD,(23–25,33) and fracture prevalence.(28,34) Moreover, it has been hypothesized that the setpoint of the bone mechanostat, and thus the relation between mechanical loading and bone strength, is modulated by estrogens.(35,36) Previously, we have shown that idiopathic osteoporosis in men mainly results from deficient acquisition of bone mass and size during maturation(37) and that these patients and their affected sons are characterized by relatively lower free estradiol (FE2) levels compared with healthy controls.(38) Although some results are conflicting, associations of serum estradiol (E2) levels with indices of bone strength generally seem to be stronger than with testosterone (T) levels, and bioavailable hormone fractions correlate better than total hormone levels.(24,26,28,33,38)
Most of these observations were obtained by BMD measurements using DXA. Fewer studies have used QCT,(7–10,12–14,26,27) which allows assessment of both bone geometry and volumetric BMD (vBMD). The aim of this study was to analyze vBMD and geometric bone parameters, assessed by pQCT, in relation to parameters reflecting mechanical loading, sex steroid levels, and their interaction in a population-based sample of healthy male siblings at the age of PBM. We will discuss E2 as the major sex steroid determinant of aBMD in these men and expand these observations to vBMD and geometric bone parameters.
MATERIALS AND METHODS
Participants were recruited from the population registries of three semirural to suburban communities around Ghent, Belgium. Men (n = 12,446) 25–45 yr of age were contacted by direct mailing, briefly describing the study purpose and asking whether they would be willing to participate and fulfilled the inclusion criterion of having a brother within the same age range also willing to participate (maximal age difference between brothers was set at 12 yr). The response rate was 30.2%. Finally, a sample of 768 young healthy men, fulfilling the primary inclusion criterion of having an eligible brother, agreed to participate. This population-based, cross-sectional study was designed to investigate the determinants of PBM in men, focusing on general lifestyle, sex hormone status, body composition, and genetic background. The study was approved by the ethics review board of the Ghent University Hospital (Belgium). All participants gave written informed consent for participation in this study, were in good health, and completed questionnaires pertaining to medical history, current and past smoking habits, alcohol consumption, dietary intake of calcium, and physical activity during the previous year. Alcohol consumption was recorded as the number of alcoholic beverages a week. Calcium intake was estimated by a food questionnaire on dairy products accounting for the number of standard portions per week. Physical activity was assessed by recording the weekly frequency of both recreational and/or professional activities and was scored using the questionnaire as proposed by Baecke et al. (39)
After exclusion, 677 healthy young men were included for analysis, consisting of 296 pairs of brothers. Sixty-four men were included as single participants, when their brother could not participate in the study, 19 men were included as a third brother, and 2 as a fourth brother in a family. For hormonal analyses, three subjects were additionally excluded for temporary use of medications affecting androgen status. Exclusion criteria were defined as disorders or treatments affecting body composition, bone metabolism, or sex hormone status: hypogonadism, hypothyroidism, cystic fibrosis, malabsorption or eating disorders, disorders of collagen metabolism or bone development, chronic renal failure, alcohol abuse, and autoimmune rheumatoid disease or the current or prolonged use in the past of glucocorticosteroids, androgens, and anti-androgens, vitamin D supplements, insulin, thyroxin, and previous or current use of anti-epileptic drugs. Data on the genetic determination and heritability of sex steroid levels in this population have been reported previously.(40,41)
Anthropometry, muscle strength, and areal BMD
Participants had their body weight measured to the nearest 0.1 kg on a calibrated balance scale in light indoor clothing without shoes. Standing height was measured to the nearest 0.1 cm using a wall-mounted Harpenden stadiometer (Holtain, Crymuch, UK). Body mass index (BMI) was calculated as the weight (kg) divided by the height in meters squared (m2). Isokinetic peak torque of biceps and quadriceps muscles was assessed at the dominant limbs using an isokinetic dynamometer (Biodex, New York, NY, USA). Grip strength at the dominant hand was measured using an adjustable hand-held standard grip device (JAMAR hand dynamometer; Sammons & Preston, Bolingbrook, IL, USA). Their maximum performance was assumed to best reflect the current status and the history of their musculoskeletal adaptation. Areal BMD (aBMD) at the lumbar spine and proximal femur (total hip region) of the nondominant limb was measured using DXA with a Hologic QDR-4500A device (software version 11.2.1; Hologic, Bedford, MA, USA). The CV% was <1% as calculated from daily spine phantom measurements.
Cross-sectional muscle area and volumetric bone parameters
Volumetric bone parameters were determined at the dominant forearm and lower leg (midshaft, 66% of bone length from distal) by pQCT (XCT2000 scanner; Stratec, Pforzheim, Germany). At the distal end of the nondominant forearm (metaphysis, 4% of bone length from distal) a cross-sectional image was taken to determine trabecular vBMD. Single tomographic slices of 2.0 mm thickness were taken at a voxel size of 0.4 mm. Imaging and the calculation of numerical values were performed using the manufacturer's software package (version 5.4). The cross-sectional area (CSA) of the radius/tibia was determined after detecting the outer bone contour at a threshold of 280 mg/cm3. At the distal radius, 55% of this cross-sectional bone area was peeled of to separate trabecular bone from the cortical shell. For determining cortical vBMD, the threshold was set at 710 mg/cm3, whereas for trabecular bone, it was set at 180 mg/cm3. Cortical thickness was estimated using the endosteal and periosteal circumferences. The cross-sectional moment of inertia (CSMI) was calculated along the latero-lateral, anterior-posterior, and polar axes, based on the cortical bone seen in the CT image at the specific threshold. The dubbed strength strain index (SSI), reflecting bone strength more generally than the CSMI,(42) was calculated as latero-lateral, anterior-posterior, and polar SSI. Muscle CSA (CSMA) was estimated using a threshold below water equivalent linear attenuation set at 0.22/cm. This threshold eliminated skin and fat mass with lower linear attenuation in the cross-sectional slice. From the remaining area, bone area was subtracted, showing the muscle at its maximum CSA.
Venous blood samples were obtained between 8:00 and 10:00 a.m. after overnight fasting, and serum was stored at −80°C until batch analysis. Commercial radioimmunoassays were used to determine serum levels of total T, sex hormone binding globulin (SHBG; Orion Diagnostica, Espoo, Finland), and E2 (Clinical Assay; DiaSorin, Saluggia, Italy; according to a modified protocol that doubles the serum amount).(23) Serum non–SHBG-bound T (bioavailable T), free testosterone (FT), non–SHBG-bound E2 (bioavailable E2), and free E2 (FE2) were calculated from serum total T (TT), total E2 (TE2), SHBG, and albumin concentrations using a previously validated equation derived from the mass action law.(43,44) Intra- and interassay CVs were <10% and 15% for all measurements, respectively.
Continuous variables were described in terms of mean ± SD if their distribution was normal according to the Kolmogorov-Smirnov test and in terms of median and first and third quartiles otherwise. Linear mixed-effects modeling with random intercepts and a simple residual correlation structure for both random (family structure) and fixed effects (variables measured within siblings) was used to evaluate cross-sectional relationships in our study population, taking the interdependence of measurements within siblings into account. These models were constructed in subsequent steps: first, based on a priori hypotheses, associations between muscle size and strength, physical activity, sex steroid levels, and volumetric bone parameters were explored. Modulation of the effects of mechanical loading by sex steroid status was explored by implementation of interaction terms (nonstandardized regression coefficients are reported because there was no indication of multicollinearity). Second, age, body weight, and height were included as covariates based on explorative univariate associations. Finally, the effect of possible confounders on these models was explored by entering calcium intake, SHBG levels, smoking status, and alcohol consumption. Parameters of fixed effects were estimated using maximum likelihood estimation and reported as regression coefficients (β) with their respective 99% CIs. In addition, decomposition of fixed effects into within- and between-family components was used to estimate the within-family variance explained,(45) which is reflected by R2 statistics for different models.(46) When necessary, analysis was done on logarithmically transformed or standardized data. Validity of the models was assessed by exploring normality of distribution of the residuals. Estimates for fixed effects were considered statistically significant at p < 0.01; all p values were two-tailed. All analyses were performed with the software package SAS 9.1.3, Service Pack 4 (SAS Institute, Cary, NC, USA).
General characteristics and associations
General characteristics and hormonal levels are shown in Table 1. All participants were of white origin. Generally, they were well educated, and the majority performed a profession with no or little manual labor. Serum sex steroid and gonadotropin concentrations were in the normal range for healthy young men.(41) Descriptives of DXA bone parameters of the lumbar spine and total hip and pQCT bone parameters of the radius and tibia and CSMA of the dominant limbs are given in Table 2.
Table Table 1.. General Characteristics and Sex Steroid Levels of Study Participants (n = 677)
Table Table 2.. Descriptives of DXA and pQCT Bone Parameters
In multivariate analyses, adjusting for total body weight and height, age was positively associated with bone area at the lumbar spine and total hip and cortical bone area and periosteal circumference at the radius but not the tibia. Except for weak negative associations between age and aBMD at the lumbar spine and total hip and trabecular vBMD at the radius, no associations with other bone parameters were observed (p = 0.005–0.029, data not shown). As reported previously,(41) negative associations were found between age and TT and FT levels, whereas no association between age and E2 or SHBG levels was observed (data not shown).
These analyses also showed that both body weight and height were positively related to cortical bone area and BMC, as well as to cortical thickness, peri- and endosteal circumference, and SSI in the three dimensions at both the radius and tibia (data not shown). Furthermore, cortical vBMD at the tibia was inversely related to total body weight (β = −0.36 [−0.58; −0.14], p < 0.001). Because age, height, and weight were associated with indices of bone strength and with sex steroid levels,(41) they are included as covariates in all further analyses exploring contributions of parameters reflecting mechanical loading and sex steroid levels to bone parameters.
Sex steroid levels as determinants of areal bone parameters
For TT levels, trends toward positive associations with BMC at the lumbar spine and total hip were found (β = 0.007 [−0.001; 0.013], p = 0.038 and β = 0.004 [−0.001; 0.009], p = 0.025, respectively), whereas FT levels seemed to be associated with total hip BMC and aBMD (β = 0.175 [−0.033; 0.383], p = 0.031 and β = 0.003 [−0.001; 0.007], p = 0.045, respectively). Serum E2 levels showed stronger positive associations with lumbar spine BMC and aBMD (TE2: β = 4.8 [−0.2; 9.8], p = 0.012 and β = 0.056 [0.003; 0.109], p = 0.005, respectively; FE2: β = 4.0 [−0.7; 8.7], p = 0.033 and β = 0.050 [0.007; 0.093], p = 0.010, respectively) and with aBMD at the total hip (TE2: β = 0.051 [−0.003; 0.105], p = 0.016 and FE2: β = 0.053 [0.001; 0.105], p = 0.008). In multivariate analyses including SHBG and total E2 or T levels, E2 levels remained positively associated with lumbar spine BMC and with aBMD at both sites, whereas TT levels were no longer significantly associated with areal bone parameters.
Parameters reflecting mechanical loading and volumetric bone parameters
Muscle size (CSMA) was strongly and positively associated with cortical bone area and BMC, cortical thickness, periosteal circumference, and SSI in the three dimensions at both the radius and tibia (Tables 3 and 4). At the radius, CSMA was positively related to endosteal circumference, and a weak negative association with cortical vBMD was found. Maximal grip strength and peak torque of both biceps and quadriceps muscles were positively associated with CSMA (p = 0.001–0.010, data not shown), cortical bone area, and BMC, as well as with cortical thickness, periosteal circumference, and SSI in the three dimensions at both sites. No associations with cortical vBMD were observed, whereas grip strength and biceps peak torque were positively associated with trabecular vBMD and endosteal circumference at the radius, respectively. Finally, physical activity score was positively associated with cortical bone area, periosteal circumference, and SSI at both sites. At the radius, a trend toward a weak negative association with cortical vBMD and a positive association with endosteal circumference was observed, whereas at the tibia, physical activity was positively associated with cortical thickness. Both muscle size and physical activity were positively associated with the ratio of cortical over total bone area at the tibia (β = 0.0008 [0.0001; 0.0015], p < 0.001 and β = 0.013 [0.001; 0.025], p = 0.003, respectively), indicating a stronger association with cortical than with total bone size. When including CSMA, grip strength or peak torque, and level of physical activity in the same model, CSMA and grip strength or peak torque remained associated with geometric bone parameters. The level of physical activity remained an independent predictor of geometric bone parameters at the tibia but not at the radius.
Table Table 3.. Associations Between Geometric and Volumetric Bone Parameters at the Radius, Parameters Reflecting Mechanical Loading, and Hormonal Variables
Table Table 4.. Associations Between Geometric and Volumetric Bone Parameters at the Tibia, Parameters Reflecting Mechanical Loading, and Hormonal Variables
Sex steroid levels and volumetric bone parameters
Total T levels tended to be positively associated with cortical bone area, periosteal circumference, and SSIp at the radius (p = 0.049, p = 0.026, and p = 0.019, respectively) and were positively associated with peri- and endosteal circumference and SSIp at the tibia (p = 0.004 and p = 0.015, respectively; Tables 3 and 4; Fig. 1). However, TT levels were no longer associated with bone size indices once SHBG was included in the same model (data not shown), in agreement with the lack of associations between FT levels and geometric or volumetric bone parameters at either site.
Both TE2 and FE2 levels were positively associated with cortical vBMD at the radius and tibia (p = 0.002 and p = 0.004, respectively), whereas a trend toward a positive association between FE2 and trabecular vBMD was observed at the distal radius (p = 0.026). At the midshaft radius, we observed positive trends between E2 levels and cortical thickness (p = 0.045 and p = 0.058 for TE2 and FE2, respectively) and a negative trend between FE2 and endosteal circumference (p = 0.023). In addition, a trend for a positive association between FE2 levels and the ratio of cortical over total bone area (β = 0.028 [−0.001; 0.057], p = 0.013) was found. No associations between E2 levels and cortical bone area, BMC, periosteal circumference, or SSI at either site were observed. These associations between bone parameters and E2 levels did not alter when adjusting for SHBG levels, and when using non–SHBG-bound (bioavailable) T or E2 instead of FT and FE2 in these analyses, results were essentially similar (data not shown). No relations between serum sex steroid concentrations and muscle size or strength were observed after controlling for age, body weight, and height (data not shown), except for weak negative associations between FT, TE2, and FE2 levels and quadriceps peak torque (β = −1.1 [−2.3; 0.1], p = 0.017; β = −16.5 [−33.3; 0.3], p = 0.012; and β = −21.2 [−37.4; −5.0], p = 0.001; respectively). Furthermore, we found no evidence for interactions between sex steroid levels and age (data not shown). As for all analyses performed, additionally adjusting for smoking, alcohol consumption, or daily calcium intake did not alter our results (data not shown).
Models with age, weight, and height as covariates and T levels as a predictor explained 3.7% of the within-family variance in cortical vBMD at the radius and 11.9% at the tibia, with individual contributions by T of <0.1%. For models with E2 levels, this was 4.9% at the radius and 15.6% at the tibia, with individual contributions by E2 of 1.9% and 4.3%, respectively. For geometric bone parameters, these numbers were 5.8–13.2% at the radius and 13.4–20.3% at the tibia for models with T levels, with individual contributions by T of <0.1–0.2% and <0.1–0.3%, respectively. For models with E2 levels, this was 5.7–16.9% at the radius and 13.6–22.9% at the tibia, with individual contributions by E2 of 0.5–3.8% and 0.4–3.7%, respectively. In general, the explained variance was higher at the tibia compared with the radius.
Interactions between parameters reflecting mechanical loading and sex steroids
To explore whether the observed associations between parameters reflecting mechanical loading and volumetric bone parameters were modulated by sex steroid levels, interaction terms between physical activity, muscle size, grip strength or peak torque, and serum E2 or T levels were modeled (Tables 3 and 4). In these analyses, positive interactions between the level of physical activity and serum E2 concentrations were observed on cortical bone area, periosteal circumference, and SSI in the three directions at the tibia but not the radius. This interaction was not found for trabecular or cortical vBMD, and additionally adjusting for smoking, alcohol consumption, daily calcium intake, muscle size, grip strength, or peak torque did not alter these observations (data not shown). These results indicate a higher impact of physical activity on bone size at the tibia in men with higher serum E2 levels. No interactions between E2 levels and muscle size or strength, or between T levels and parameters reflecting mechanical loading, were observed. We explored the possibility that these observed interactions between physical activity and E2 levels were specifically caused by an interaction toward CSMA, grip strength, or peak torque. In a model adjusting for age, weight, and height, physical activity was significantly positively related to CSMA, grip strength, and peak torque. However, neither CSMA nor grip strength or peak torque was significantly associated with E2 levels or with an interaction term between physical activity and E2 levels (data not shown).
In this study of healthy male siblings at the age of PBM, we found that parameters reflecting mechanical loading, defined by the level of physical activity, CSMA, grip strength, or peak torque, were major determinants of bone size and mineral content at both the radius and tibia. In agreement with the mechanostat theory as proposed by Frost(5) and previous reports on the relation between mechanical loading and bone size,(7,10,12) we showed that larger bones and consequently greater indices of bone strength in males are positively associated with parameters reflecting mechanical loading by muscles. Furthermore, consistent with the finding of estrogen as the major sex steroid associated with aBMD at the lumbar spine and total hip in these men, we observed a strong positive association between (F)E2 levels and cortical vBMD at the radius and tibia, as well as some trends for associations with cortical thickness, endosteal circumference, and the ratio of cortical over total bone area at the radius. Together with our previous finding of lower FE2 levels in men with idiopathic osteoporosis and their affected sons,(38) this suggests a significant role of estrogen in the acquisition of bone mass and geometry in men.
The important contribution of parameters reflecting mechanical loading to cortical bone size and mineral content observed in our study participants is not surprising. In previous studies, it was found that physical activity was associated with BMC and aBMD,(6,15) but also with cortical bone size and trabecular vBMD.(7) Grip strength has been shown to be associated with cortical vBMD and cortical bone size,(10) as well as with indices of bone strength,(8–10) and muscle size is known to correlate with bone size and mineral content.(12–14) We expand these reports with the observation of strong positive associations between isokinetic peak torque of biceps and quadriceps muscles and cortical bone size in men at the age of PBM. Moreover, the positive relationships between parameters reflecting mechanical loading and periosteal circumference at both limbs, and with endosteal circumference at the radius clearly support the concept of mechanical determination of periosteal bone apposition (modeling). Furthermore, cortical vBMD at the tibia was not associated with indicators of mechanical loading and inversely related to total body weight, whereas cortical vBMD at the radius seemed to be negatively associated with muscle size and level of physical activity. In this regard, previous studies also failed to find a relation between level of physical activity and cortical vBMD.(7) Several mechanisms may be put forward to explain these latter findings. One possible explanation might be that higher mechanical loading could lead to more intracortical remodeling, increased cortical porosity, and subsequently lower mean material density of cortical bone.(47)
The finding of serum E2 and not T levels as the dominant sex steroid associated with BMD in men is in agreement with previous cross-sectional and longitudinal reports,(22–24,26,29,33) although we should note that participants of most of these studies were of older age. For instance, Khosla et al.(26) found similar observations in middle-aged and elderly, but not in young men. In contrast with our results, Lorentzon et al.(27) found that, in a population of younger men, FT was a positive predictor of cortical bone size, whereas FE2 was negatively associated with cortical bone size at both the radius and tibia. However, these younger men might not yet have attained cortical PBM,(48) which could explain the positive associations of serum T levels with bone size. In our study, a trend for associations between FT levels and total hip aBMD and BMC might indicate some effect of T on bone size but not necessarily on vBMD, because aBMD and BMC are influenced by size. However, in agreement with the lack of associations between FT levels and cortical bone size, positive associations between TT serum levels and bone size were nonsignificant when adjusting for SHBG levels. Furthermore, within our age group, no evidence of modulation the effect of sex steroids on bone parameters by age was found.
Taken together, it seems that, whereas normal T and E2 levels are important for attaining an adequate PBM during adolescence and young adulthood, as well as for preventing increased bone loss with aging, variation of T levels within the normal range at age of PBM is of lesser importance. In this context, it should be pointed out that differences in serum T within the normal range reflect in part differences in androgen sensitivity and feedback setpoint and thus not simply interindividual differences in exposure to androgen action,(40) which might explain the lack of associations between T levels and indices of bone strength in this study. Furthermore, most of the literature reporting T effects on bone homeostasis pertains to drastic changes around puberty or to situations of (partial) hypogonadism and its treatment, whereas this study concerns the different situation of healthy young men with serum T levels in the normal range.
The important role of estrogens in bone homeostasis is generally acknowledged. Besides positive associations between serum E2 levels and BMD, we found a trend toward a negative association between FE2 levels and endosteal circumference at the radius and a similar nonsignificant trend at the tibia. This is in agreement with the putative role of estrogens in the mechanostat theory, where they are considered to lower the modeling threshold on endosteal bone surfaces,(36) and with the results by Khosla et al. in elderly men.(26) As a result, estrogens would promote a greater bone mass for the same muscle mass, as observed in women compared with men from puberty until menopause.(10,11,13,14)
In addition, positive interactions between serum E2 levels and physical activity level on cortical bone area, periosteal circumference, and SSI in the three directions were observed at the tibia. This suggests that endogenous levels of E2 might increase cortical bone area by modulating the effect of physical activity on the periosteal bone surface. One could speculate that estrogens also lower the threshold for the bone modeling process on the periosteal bone surface because of mechanical loading in men, as has been suggested by Vanderschueren et al.(20) This modulation by estrogens has also been observed for the remodeling threshold in women compared with men, explaining both the larger accumulation of bone mass with regard to muscle size in pubertal girls compared with boys and the increased rate of postmenopausal bone loss.(11,36) However, in our study, we found no evidence of modulation of the impact of muscle size or strength on bone parameters by sex steroids in these healthy males. Albeit not identical, the finding of modulation of physical activity impact by estrogens is in line with a previous report by Lorentzon et al.(49) In addition, in mice deficient of estrogen receptors (ERs), it has been shown that periosteal (and endosteal in case of ER-α deficiency) apposition caused by mechanical loading was less efficient compared with wildtype mice.(50)
We recognize that our study has some limitations. First, observations within brothers are not completely independent from each other. However, all analyses in this study were performed using linear mixed-effects modeling with random intercepts to account for this interdependence. Furthermore, cross-sectional population-based studies are inherently prone to certain bias (e.g., cohort effects and selection bias, the so-called “healthy worker effect”); in our cohort, this was reflected by relatively high levels of education and a low percentage of manual labor. This could certainly have biased our study population, but may in fact even strengthened our results, by avoiding a large variation in bone mass, physical activity, and sex steroid levels caused by larger differences in environmental factors such as lifestyle. Obviously, cross-sectional studies do not allow one to draw causative conclusions. The strength of this study is that it was performed in a well-defined sample of healthy men, using a population-based approach. The use of pQCT allows to differentiate trabecular from cortical bone and material from structural parameters of bone strength. Bone parameters obtained by pQCT in our study population are of equal magnitude as previously reported in young healthy men of comparable age.(51) Because the associations found between age and BMD were nonsignificant or very weak, we can assume that our study participants generally achieved PBM at the time of the study. Age was positively associated with cortical bone area and periosteal circumference at the radius, reflecting continuing periosteal apposition (modeling) in these adult men.(19) Because all study subjects were younger than 46 yr of age and healthy at the time of the study, we can assume the absence of relevant interference from degenerative alterations of the skeletal system.
In conclusion, the results of this study indicate that mechanical loading is one of the strongest determinants of bone size at the age of PBM and thus of utmost importance for achieving optimal bone strength. In addition, our findings confirm that E2 is the dominant sex steroid in relation to BMD and also seems to be associated with cortical bone size in men. Moreover, E2 seems to reinforce positive effects of physical activity on bone size. In contrast, variations of T levels within the normal range do not seem to have a substantial influence on material or structural bone parameters in healthy men at age of PBM. These observations support the notion that both mechanical loading and E2 metabolism play a crucial role in the acquisition of bone mass and geometry in men.
The authors thank Prof. Dr. D. Cambier and Prof. Dr. E. Witvrouw for help with the isokinetic muscle strength measurements, Prof. Dr. S. Vansteelandt for statistical consulting, and K. Toye, R. De Muynck, H. Myny, M. Masschelin, and I. Bocquaert for excellent technical assistance. This work was supported by grants from the Flemish Fund for Scientific Research [Fonds Wetenschappelijk Onderzoek (FWO) Vlaanderen Grants G.0331.02 and G.0404.00]. V.B. and G.V.B are PhD students funded by Bijzonder Onderzoeksfonds of Ghent University and the Flemish Fund for Scientific Research, respectively.