Cortical Consolidation due to Increased Mineralization and Endosteal Contraction in Young Adult Men: A Five-Year Longitudinal Study

Abstract

Context:

Peak bone mass is an important factor in the lifetime risk of developing osteoporosis. Large, longitudinal studies investigating the age of attainment of site-specific peak bone mass are lacking.

Objective and Main Outcome Measures:

The main outcome measures were to determine the site-specific development of peak bone mass in appendicular and axial skeletal sites and in the trabecular and cortical bone compartments, using both dual x-ray absorptiometry and peripheral computed tomography.

Design, Setting, and Population:

In total, 833 men [aged 24.1 ± 0.6 yr (mean ± sd)] from the original population-based Gothenburg Osteoporosis and Obesity Determinants Study (n = 1068) were included in this follow-up examination at 61.2 ± 2.3 months. Areal bone mineral density (aBMD) was measured with dual x-ray absorptiometry, whereas cortical and trabecular volumetric bone mineral density and bone size were measured by peripheral computed tomography at baseline and at the 5-yr follow-up.

Results:

During the 5-yr study period, aBMD of the total body, lumbar spine, and radius increased by 3.4, 4.2, and 7.8%, respectively, whereas a decrease in aBMD of the total hip of 1.9% was observed (P < 0.0001). Increments of 2.1 and 0.7% were seen for cortical volumetric bone mineral density of the radius and tibia, respectively (P < 0.0001), whereas cortical thickness increased by 3.8% at the radius and 6.5% at the tibia due to diminished endosteal circumference (radius 2.3% and tibia 4.6%, P < 0.0001).

Conclusion:

aBMD decreased at the hip but increased at the spine and radius, in which the increment was explained by continued mineralization and augmented cortical thickness due to endosteal contraction in men between ages 19 and 24 yr.

Peak bone mass (PBM), defined as the amount of bone present in the skeleton at the end of its maturation process (1), is an important factor in the lifetime risk of developing osteoporosis (2). Genetics is the major determinant of PBM (3) because genetic factors can explain about 70% of the variance in bone phenotype at any particular age and phase of life (4). Other important determinants are physical activity, calcium intake, and endocrine factors, such as sex hormones and GH (5). The timing of PBM has not yet been clarified, with previous studies presenting discrepant results (6–10), although most suggest that bone mass does not increase significantly after the third decade. In a longitudinal study of young women (n = 50) using both dual x-ray absorptiometry (DXA) and computed tomography (CT), Wren et al. (11) reported that in the axial skeleton, CT-assessed vertebral bone mineral content (BMC) and bone density reach their peak around the time of sexual and skeletal maturity. Vertebral BMC and bone mineral density (BMD) DXA values continued to increase beyond sexual and skeletal maturity, increases that are likely due to the influence of soft tissues and/or the posterior elements, Wren et al. concluded. Large longitudinal studies investigating the age of attainment of site-specific PBM, elucidating bone accumulation divided according to the appendicular and axial skeletal sites and the trabecular and cortical bone compartments, are lacking. Previous studies using peripheral quantitative CT (pQCT) include a cross-sectional study that reported greater trabecular density of the tibia in younger men (18–30 yr, n = 31) than older men (50–64 yr, n = 37), whereas no significant age group differences in cortical bone measures were seen (12). Khosla et al. (13) reported, in a cross-sectional study using high-resolution three-dimensional pQCT imaging, that cortical volumetric BMD (vBMD), cortical thickness, and trabecular thickness in the wrist decreased 16, 38, and 24%, respectively, between the ages of 20 and 90 yr in men. For the same age span, Riggs et al. (14) reported decreases in trabecular vBMD of the distal radius and tibia of 27 and 25%, respectively, whereas cortical vBMD of the distal radius and tibia decreased by 21%. In a large age-stratified population followed up longitudinally using quantitative CT, Riggs et al. (15) investigated bone development in men and women but found no significant alterations in cortical or trabecular bone departments in a small (n = 8) subsample of men 20–29 yr of age. We have earlier reported, in the cross-sectional Gothenburg Osteoporosis and Obesity Determinants (GOOD) study of men 18–20 yr old, that increasing age was associated with higher vBMD and decreasing endosteal circumference (EC) in the radius and tibia, indicating ongoing skeletal mineralization with an endosteal contraction taking place in these long bones. In contrast, age was not associated with areal BMD (aBMD) at the lumbar spine and femoral neck, suggesting attained peak bone mass at these sites (16). In the present study, we examined the changes in bone geometry, aBMD, and vBMD, over a 5-yr period, in young adult males originally enrolled in the GOOD study. Using both DXA and pQCT, we aimed to determine the site-specific bone development in young men.

Subjects and Methods

Subjects

The population-based Gothenburg osteoporosis and obesity determinants (GOOD study), aiming to determine both environmental and genetic factors involved in the regulation of bone and fat mass, was initiated in 2003. Five years later, the study subjects in the original GOOD study were contacted by letter and telephone and invited to participate in the 5-yr follow-up study. Of the original 1068 subjects, 833 men, 24.1 ± 0.6 yr of age, were included in the follow-up study (78% of the original population). Despite the use of previously known addresses and searching of national address registers, 107 subjects could not be reached (10%). A total of 128 subjects declined to participate (12%). The original GOOD cohort was found representative of the general young male population in Gothenburg (16). To determine whether the cohort of the 5-yr follow-up was also representative of the initial population, we compared the age, height, weight, and amount of present physical activity (all variables measured at the time of inclusion in the original GOOD study) of the included subjects (n = 833) with the subjects that were not included (n = 235). There were no significant differences between the subjects included and not included in age (years) (included 18.9 ± 0.6, not included 18.8 ± 0.6, P = 0.071), height (centimeters) (181.6 ± 6.7 vs. 180.9 ± 6.9, P = 0.20), weight (kilograms) (73.8 ± 11.8 vs. 73.9 ± 12.1, P = 0.90), or amount of present physical activity (hours per week) (4.3 ± 5.1 vs. 4.7 ± 6.0, P = 0.33), using an independent samples t test (mean ± sd). A standardized self-administered questionnaire was used to collect information about present smoking (yes/no), present physical activity (hours per week, weeks per year, and duration in years), and nutritional intake. Calcium intake was estimated from dairy product intake. The follow-up period was 61.2 ± 2.3 months (mean ± sd) (range 55–70 months). The study was approved by the regional ethical review board at the University of Gothenburg. Written and oral consent was obtained from all study participants.

Anthropometrical measurements

Height and weight were measured using standardized equipment. The coefficient of variation was below 1% for these measurements.

Dual x-ray absorptiometry

aBMD (grams per square centimeter) of the whole body, femoral neck, total hip (of the left leg), lumbar spine, and the left and right radius were assessed using the Lunar Prodigy DXA (GE Lunar Corp., Madison, WI). The coefficients of variation for the aBMD measurements ranged from 0.5 to 3%, depending on application. Measurements were obtained using the same procedure as in the original GOOD study. One person performed all the measurements of the baseline study, and another person performed all the measurements of the follow-up study. The Lunar Prodigy DXA used in the follow-up visit was not the same specimen as the one used in the baseline visit. Cross-calibration between the two Lunar Prodigy DXA machines was performed at the time of the follow-up. Phantom measurements (phantom no. 9397) were made regularly with the original Lunar Prodigy DXA and were stable over time [1.270 (grams per square centimeter) at the time of the original study and 1.268 at the time of the follow-up study]. In the cross-calibration, 24 young men, 24.7 ± 1.7 yr of age, were measured with both Lunar Prodigy DXA devices within a period of 4 d.

Regression equations using aBMD derived from the baseline instrument as a dependent variable and aBMD derived from the follow-up instrument as an independent variable were calculated from aBMD data from these 24 men. The regression coefficient and constant was 1.046 and −0.040 for the total hip, 1.021 and −0.015 for the femoral neck, 1.038 and −0.012 for the lumbar spine, 1.017 and −0.007 for the total body, 0.961 and 0.030 for the left radius, and 0.938 and 0.044 for the right radius, respectively. All DXA measurements of the follow-up visit were adjusted according to these regression equations. Due to weight limits of the Lunar Prodigy DXA (GE Lunar Corp.), five subjects in the follow-up study could not undergo the total body scan, lumbar spine scan, or hip/femur scan. One subject had an erroneous DXA measurement of the radius in the follow-up (not being positioned correctly to allow for trending). This measurement was excluded.

Peripheral quantitative CT

A pQCT device (XCT-2000; Stratec Medizintechnik, GmbH, Pforzheim, Germany) was used to scan the distal leg (tibia) and the distal arm (radius) of the nondominant leg and arm, respectively. The pQCT was calibrated every week using a standard phantom and once every 30 d using a cone phantom provided by the manufacturer. A 2-mm-thick single tomographic slice was scanned with a voxel size of 0.50 mm. The cortical vBMD (not including the bone marrow) (milligrams per cubic centimeter), cortical BMC (milligrams per millimeter), cortical cross-sectional area (CSA; square millimeters), EC and periosteal circumference (PC), and cortical thickness (millimeters) were measured using a scan through the diaphysis (at 25% of the bone length in the proximal direction of the distal end of the bone) of the radius and tibia. The threshold for cortical bone was 711. Trabecular vBMD (milligrams per cubic centimeter) was measured using a scan through the metaphysis (at 4% of the bone length in the proximal direction of the distal end of the bone) of these bones. Trabecular vBMD was assessed using the inner 45% of these bones. Tibial bone length was measured from the medial malleolus to the medial condyle of the tibia, and length of the forearm was defined as the distance from the olecranon to the ulna styloid process. Measurements in the follow-up visit were made using the same procedure and the same equipment as in the original GOOD study. One person performed all the measurements of the baseline study, and another person performed all the measurements of the follow-up study. The coefficients of variation were less than 1% for all pQCT measurements. Due to movement artifacts, two radius scans and one tibia scan were excluded. Two scans were excluded (one tibia scan and one radius scan) because of bone metal being present in the measuring field, and one radius scan was excluded due to an incorrectly positioned measuring field.

Statistical analysis

Differences in bone parameters between the study subjects at baseline and follow-up, as well as differences in the rate of yearly change for different skeletal sites, were investigated using a paired-samples t test. To assess the correlation between bone variables and age, bivariate correlations were used. The associations between age and bone variables were investigated using linear regression analysis, with height, weight, smoking, calcium intake, and physical activity as covariates. These covariates were also used in linear regression equations to calculate adjusted baseline and 5-yr bone variables and the 5-yr changes (also adjusted for follow-up time) observed (Fig. 1). Weight was not normally distributed and was therefore log transformed before being entered into the regression analysis. A stepwise multiple linear regression analysis, including changes over 5 yr in trabecular and cortical vBMD, cross-sectional area, periosteal circumference, EC, and cortical thickness as predictors, was used to determine whether changes in vBMD or parameters of cortical bone geometry predicted the increase in radius aBMD. The stepwise selection process criterion for entry into the model was a P ≤ 0.05, and the criterion for removal from the model was a P ≥ 0.10. A P < 0.05 was considered significant. Changes over 5 yr in effect size were calculated by subtracting the baseline value from the follow-up value, and then dividing by the average of the sd of the baseline value and the follow-up value, for each variable. Changes in effect size were also adjusted for follow-up time. The data were analyzed using SPSS software, version 15.0 and 19.0 (SPSS Inc., Chicago, IL).

Fig. 1

Five-year changes expressed in percent in cortical bone variables (A and B) and aBMD (C) after adjustment for covariates including calcium intake, smoking, physical activity, height, log weight, and follow-up time, presented as mean ± sem. THICKNESS, Cortical thickness, PC, periosteal circumference.

Results

Anthropometric characteristics

The 833 men included in the follow-up study were substantially heavier and slightly taller than at the baseline visit. Their reported calcium intake and weekly physical activity were reduced since the baseline visit (Table 1).

Table 1

Characteristics of the total cohort at the baseline and follow-up visits

	Men at baseline (n = 833)	Men at follow-up (n = 833)
Age (yr)	18.9 ± 0.6	24.1 ± 0.6a
Smoking (%)	7.2%	7.3%
Physical activity (h/wk)	4.3 ± 5.1	2.7 ± 3.6a
Calcium intake (mg/d)	1105 ± 697	789 ± 510a
Height (cm)	181.6 ± 6.7	182.1 ± 6.7a
Weight (kg)	73.8 ± 11.8	78.5 ± 12.6a

Values are given as mean ± sd. Differences between baseline and follow-up were investigated using paired-samples t test, except for smoking in which χ² test was used.

^a

P < 0.00001.

Changes in bone parameters from 18–20 to 23–25 yr of age

aBMD of the total body and lumbar spine (L2-L4) increased 3.4 and 4.2%, respectively, between the baseline and follow-up visits. At the radius, aBMD increased by 7.8%, whereas decreases in aBMD at the femoral neck and total hip of 3.6 and 1.9%, respectively, were observed over the 5-yr follow-up period (Table 2).

Table 2

Longitudinal changes in bone parameters measured with DXA and pQCT over 5 yr in young men (n = 833)

	Baseline (18–20 yr) mean ± sd	Follow-up (23–25 yr) mean ± sd	5-yr change (%)	5-yr change (effect size)
DXA
Total body aBMD (g/cm2)a	1.25 ± 0.10	1.29 ± 0.10b	3.4 ± 3.3	0.41 ± 0.40
Lumbar spine L2-L4 aBMD (g/cm2)a	1.23 ± 0.15	1.29 ± 0.16b	4.2 ± 5.3	0.33 ± 0.42
Total hip aBMD (g/cm2)a	1.17 ± 0.16	1.14 ± 0.16b	−1.9 ± 5.2	−0.14 ± 0.38
Femoral neck aBMD (g/cm2)a	1.17 ± 0.16	1.12 ± 0.16b	−3.6 ± 5.7	−0.26 ± 0.42
Radius nondominant aBMD (g/cm2)c	0.58 ± 0.06	0.63 ± 0.05b	7.8 ± 4.3	0.82 ± 0.41
pQCT
Radius cortical vBMD (mg/cm3)d	1166 ± 23	1191 ± 17b	2.1 ± 1.4	1.25 ± 0.77
Radius cortical CSA (mm2)d	96 ± 12	99 ± 12b	3.1 ± 3.8	0.24 ± 0.29
Radius cortical thickness (mm)d	2.93 ± 0.27	3.04 ± 0.27b	3.8 ± 4.8	0.40 ± 0.48
Radius periostal circumference (mm)d	42.1 ± 3.0	42.2 ± 2.9b	0.3 ± 1.6	0.03 ± 0.23
Radius EC (mm)d	23.7 ± 3.1	23.1 ± 2.9b	−2.3 ± 4.4	−0.19 ± 0.38
Radius polar SSI (mm3)d	306 ± 61	320 ± 64b	4.9 ± 5.5	0.23 ± 0.26
Radius trabecular vBMD (mg/cm3)d	220 ± 41	226 ± 41b	2.9 ± 6.7	0.14 ± 0.35
Tibia cortical vBMD (mg/cm3)c	1156 ± 20	1164 ± 19b	0.7 ± 1.2	0.38 ± 0.68
Tibia cortical CSA (mm2)c	269 ± 35	281 ± 36b	4.2 ± 3.3	0.32 ± 0.25
Tibia cortical thickness (mm)c	4.43 ± 0.51	4.72 ± 0.55b	6.5 ± 4.4	0.54 ± 0.36
Tibia periostal circumference (mm)c	74.9 ± 4.9	74.5 ± 4.8b	−0.5 ± 1.3	−0.08 ± 0.21
Tibia EC (mm)c	47.1 ± 5.4	44.9 ± 5.4b	−4.6 ± 3.6	−0.40 ± 0.32
Tibia polar SSI (mm3)c	1648 ± 307	1761 ± 326b	6.8 ± 3.9	0.35 ± 0.20
Tibia trabecular vBMD (mg/cm3)e	266 ± 34	261 ± 35b	−1.7 ± 5.3	−0.14 ± 0.41

Values are given as mean ± sd. Changes over 5 yr (mean ± sd) are adjusted for follow-up time and presented in percent and effect size ((follow-up baseline value)/sd_average). Differences in bone parameters at baseline are compared with follow-up after 5 yr were investigated using paired-samples t test.

^a

n = 828.

^b

P < 0.00001.

^c

n = 832.

^d

n = 829.

^e

n = 831.

Cortical vBMD of the radius and tibia, measured with pQCT, demonstrated an increase of 2.1 and 0.7%, respectively, over the 5-yr follow-up period. The polar stress strain index (SSI) increased 4.9 and 6.8% at the radius and tibia, respectively, between the examinations. Cortical thickness increased by 3.8% for the radius and 6.5% for the tibia between examinations due to diminished endosteal circumference, which decreased at the radius and tibia by 2.3 and 4.6%, respectively. Trabecular vBMD increased by 2.9% at the radius and decreased at the tibia by 1.7% between the baseline visit and the follow-up visit (Table 2). The largest changes measured in effect size were seen for aBMD and cortical vBMD of the radius (Table 2). All changes remained highly significant after adjustment for covariates (calcium intake, smoking, physical activity, height, weight) (Figs. 1 and 2).

Fig. 2

Five-year changes expressed in effect size in cortical bone variables (A and B) and aBMD (C) after adjustment for covariates including calcium intake, smoking, physical activity, height, log weight, and follow-up time, presented as mean ± sem. THICKNESS, Cortical thickness, PC, periosteal circumference.

A stepwise linear regression model was used to determine whether the changes in adjusted (as above) vBMD (cortical and trabecular) of the radius or adjusted parameters of bone size of the radius (CSA, periosteal circumference, endosteal circumference, or cortical thickness) most strongly predicted the adjusted increase in radius aBMD. In this model, the increase in cortical CSA most strongly predicted the increase in radius aBMD (standardized β = 0.36, R²= 29%, P < 0.00001), followed by changes in trabecular (β = 0.32, R² = 12%, P < 0.00001) and cortical (β = 0.29, R² = 6.9%, P < 0.00001) vBMD. In addition, we used a similar regression analysis to determine which longitudinal changes (adjusted for covariates) in pQCT variables of the tibia best predicted adjusted changes in total hip aBMD. In this model, tibia cortical CSA was the strongest predictor of change in total hip aBMD (standardized β = 0.43, R² = 39%, P < 0.00001), followed by trabecular vBMD of the tibia (β = 0.42, R² = 14%, P < 0.00001).

Association between age and bone parameters at 23–25 yr of age

Areal BMD of the radius was weakly correlated to age at the follow-up visit (Table 3). Age was not correlated to aBMD of the total body, lumbar spine, femoral neck or total hip (Table 3).

Table 3

Association between age and bone parameters in men at the follow-up visit (23–25 yr of age)

	Follow-up (23–25 yr) r value	Follow-up (23–25 yr) β−value
DXA
Total body aBMD (g/cm2)a	0.01	0.01
Lumbar spine L2-L4 aBMD (g/cm2)a	−0.02	−0.02
Total hip aBMD (g/cm2)a	−0.01	−0.01
Femoral neck aBMD (g/cm2)a	−0.01	−0.01
Radius nondominant aBMD (g/cm2)b	0.09c	0.08c
pQCT
Radius cortical vBMD (mg/cm3)d	0.18e	0.18e
Radius cortical CSA (mm2)d	0.02	0.03
Radius cortical thickness (mm)d	0.08c	0.09c
Radius periostal circumference (mm)d	−0.03	−0.02
Radius EC (mm)d	−0.08c	−0.07c
Radius polar SSI (mm3)d	−0.02	−0.005
Radius trabecular vBMD (mg/cm3)d	0.07c	0.07c
Tibia cortical vBMD (mg/cm3)b	0.13f	0.13e
Tibia cortical CSA (mm2)b	0.03	0.04
Tibia cortical thickness (mm)b	0.07	0.07c
Tibia periostal circumference (mm)b	−0.03	−0.02
Tibia EC (mm)b	−0.06	−0.06
Tibia polar SSI (mm3)b	−0.01	−0.001
Tibia trabecular vBMD (mg/cm3)g	0.02	0.02

Bivariate correlation and linear regression analyses (including age and calcium intake, smoking, physical activity, height, and log weight as covariates) were performed. R and standardized β-values are shown.

^a

n = 828.

^b

n = 832.

^c

P < 0.05.

^d

n = 829.

^e

P < 0.00001.

^f

P < 0.001.

^g

n = 831.

Age was correlated to cortical vBMD at the radius and tibia at the follow-up visit (Table 3). At the radius, weak correlations between age and cortical thickness, EC, and trabecular vBMD were seen (Table 3).

Factors known to affect PBM, such as physical activity, smoking, calcium intake, weight, and height, were used, together with age, in multiple linear regression models to determine the independent predictors of the different bone parameters measured. Using this model, age was an independent predictor of aBMD of the radius, cortical vBMD, and thickness of the radius and tibia as well as endosteal circumference and trabecular vBMD of the radius at the follow-up visit (Table 3).

Differences in changes over 5 yr, depending on bone site

To investigate whether the 5-yr changes differed between weight-bearing and non-weight-bearing bones, we compared the rate of 5-yr change (percentage) between different skeletal sites, using a paired-samples t test. The increase of radius cortical vBMD was significantly higher than that of tibia cortical vBMD (2.1 and 0.7, respectively, P < 0.0001). On the other hand, the increase in CSA and cortical thickness was significantly higher in the tibia than the radius (4.2 and 3.1 for CSA and 6.5 and 3.8 for cortical thickness, respectively, P < 0.0001). The decrease of the EC of the radius was significantly lower than of the tibia (−2.3 and −4.6, respectively, P < 0.0001). The change in trabecular vBMD was also significantly different between the radius and the tibia (2.9 and −1.7, respectively, P < 0.0001). The change in lumbar spine aBMD was significantly different from the change in femoral neck aBMD (4.2 and −3.6, respectively, P < 0.0001).

Discussion

To our knowledge, our study is the first longitudinal study with a large population-based cohort, examining changes in bone parameters, assessed with DXA and pQCT, in young male adulthood. In the present study, we observed an increase in aBMD, measured with DXA, of especially the radius but also smaller gains in aBMD were seen for the lumbar spine and total body over the 5-yr study period. In contrast, a decrease in aBMD at the femoral neck and total hip was seen.

Using pQCT measurements, we found that the increase in radius aBMD was mainly due to an increase in cortical vBMD together with an increased cortical thickness dependent on endosteal contraction, suggesting that bone strength is augmented in early adulthood by means of cortical consolidation (Fig. 3). During the study period, the polar SSI, an estimate of bone strength, increased 4.9 and 6.8% of the radius and tibia, respectively. Our regression analysis demonstrated that altered CSA in particular, but also changed cortical and trabecular mineralization, could statistically explain the gain in radius aBMD, indicating that mainly an increase in bone size, as well as continued mineralization, contributes to the increased aBMD here observed in young males.

Fig. 3

Schematic illustration of cortical alterations occurring in men between ages 18–20 and 23–25 yr.

In a cross-sectional study using high-resolution pQCT, Kirmani et al. (17) investigated bone development during adolescence in boys and girls (aged 6–21 yr). Their results indicated that the EC increased during childhood and early teenage years, whereas a tendency toward a decrease in EC was observed between the age group 15–17 and 18–21 yr. In our study, the EC continued to decrease between 18–20 and 23–25 yr of age. Xu et al. (10) reported in a 7-yr longitudinal study of girls (mean age at baseline 11.2 yr) that at the age of 18 yr, bone values were still significantly lower than in an adult reference population, suggesting that bone mass accrual is still ongoing at the end of the second decade in life, results that are consistent with our present findings. Another longitudinal study including men and women 21–102 yr of age reported a slight increase in cortical bone area in younger men as well as periosteal apposition and endocortical resorption over the aging process (18). That study, as well as cross-sectional studies by Khosla et al. (13)and Riggs et al. (14), primarily focus on changes in bone structure and microstructure over a lifetime, with study subjects spanning a broad age range, and are therefore not easily comparable with the present study. Unexpectedly, tibial periosteal circumference decreased marginally between baseline and follow-up in our study, in contrast to previous findings (18, 19). Although we used the same procedure and equipment in the follow-up as in the baseline study, we cannot rule out that this very small but significant difference was related to measurement errors. However, periosteal circumference of the radius increased as expected, arguing against a general measuring error being the cause of the periosteal circumference reduction at the tibia. In a longitudinal pQCT study (15), trabecular bone loss at the distal radius, distal tibia, and lumbar spine was reported in men 21–49 yr of age. When divided into subgroups according to age, no significant changes were seen in men 20–29 yr of age, but these results were limited by the small sample of young men (15). In our present study, trabecular vBMD decreased at the distal tibia while at the distal radius, an increase was detected.

To examine whether bone developed in a site-specific manner, we investigated whether there were significant differences in bone development between different sites. We found that the increase in cortical and trabecular vBMD in the radius was significantly higher than in the tibia, whereas the increase in CSA and cortical thickness, as well as the decrease in endosteal circumference, was significantly greater in the tibia than in the radius, suggesting that weight-bearing and non-weight-bearing bones develop differently in young men between 19 and 24 yr of age.

In a previous study using DXA, Henry et al. (6) found that at the femoral neck, peak vBMD had already been achieved by 12 yr of age in both sexes. By late adolescence, BMD had already begun to decline marginally. This is consistent with our findings, showing that between the ages of 18–20 and 23–25 yr, aBMD of the femoral neck and total femur decreases by 3.6 and 1.9%, respectively. A recent longitudinal study by Berger et al. (20) reported that lumbar spine PBM occurred at ages 19–33 yr, and total hip PBM occurred at ages 19–21 yr in men. However, a longitudinal study using DXA reported increases in aBMD at the total hip, as well as the radius and ulna, for young men (mean age 30.5 yr, range 27.5–35.8 yr) during a 4-yr follow-up period (21). In contrast, the midlateral spine BMD decreased. In that study, authors surprisingly detected increases in hip BMD in both middle-aged and elderly men, although the increase in the latter group was not significant. In an analysis limited by comparing two different bones, we found that the changes in cortical CSA of the tibia was the strongest predictor and could explain 39% of the variation of the change in total hip aBMD, indicating that changes in cortical CSA are most strongly linked to changes in aBMD seen at the total hip in addition to the above-mentioned changes in radius aBMD.

Previous studies using DXA have reported that bone accrual at the lumbar spine continues beyond the end of longitudinal growth (6, 22). Wren et al. (11) reported that bone acquisition in the lumbar spine as measured with CT reaches its peak by sexual and skeletal maturity. In contrast, bone values by DXA continued to increase after puberty and cessation of longitudinal growth (11). It was suggested by Wren et al. (11) that their findings indicated that increases in DXA measurements were likely to reflect changes in the soft tissues surrounding the spine and/or the posterior elements of the vertebrae (posterior arch and spinal processes), rather than changes within the vertebral body itself. DXA values are assessed using the entire vertebrae, whereas CT measurements include exclusively the vertebral body, and consequently BMD results assessed by DXA reflect changes also outside the vertebrae itself. In addition, studies of DXA measurements have reported that changes in fat content and distribution around the lumbar spine can also affect BMD results (23, 24). In our study, the lumbar spine was only measured with DXA, and there was an increase in aBMD of 4.2% between baseline and follow-up. The men in our cohort increased in weight by almost 5 kg between baseline and follow-up. Nordström et al. (25) reported in a longitudinal study of healthy young men that there was a significant increase in abdominal fat between the age of 17 and 25 yr. It is possible that the increase in aBMD of the lumbar spine observed in our study could partly be due to soft tissue changes, particularly increased abdominal fat influencing DXA measurements.

Although our longitudinal data revealed increments in aBMD of the radius, lumbar spine, and total body, our cross-sectional correlation analysis, at the follow-up visit, demonstrated no association between age and aBMD at all sites excepting the radius. These analyses indicate that for aBMD at the lumbar spine and total body, the increments are attenuated or stopped between 23 and 25 yr. In contrast, the cross-sectional analysis at the follow-up demonstrated a significant correlation between age and radius aBMD, cortical vBMD, and EC, indicating that the process of cortical consolidation of the long bones is not attenuated or stopped at 23–25 yr. However, the cross-sectional data are less reliable than our longitudinal data and should be interpreted with caution. Considering that the age range of the study population was relatively narrow, the age term had relatively little variability, and consequently the cross-sectional analyses with respect to age effects may have limited power.

The present study has several notable strengths. It is a large longitudinal study involving a well-characterized cohort, measured with both DXA and pQCT. There are also some limitations with the present study. One limitation was our inability to include all men in the follow-up examination (a dropout rate of 22%). Furthermore, the study population was primarily white. Thus, the results cannot automatically be transferred to other ethnicities. Although we performed a very thorough cross-calibration between the two different DXA machines used at the baseline and follow-up visits, the use of two devices could have skewed our results concerning aBMD, and cross-calibration at follow-up cannot correct for time differences in the machine. However, results on longitudinal changes in cortical bone parameters obtained with a single pQCT device at both examinations support our findings from the DXA measurements concerning the results for the radius. Another limitation is the relatively poor resolution of the pQCT used, which is especially relevant to the findings regarding trabecular bone traits. Furthermore, the lack of CT to assess the axial skeleton constitutes a limitation with the present study. It is also important to consider that the study sample was large, which increases the possibility of detecting small changes that are statistically significant but not clinically relevant, such as the changes in periosteal circumference in the radius and the tibia.

Ideally, large population-based studies, following up participants over several decades, could be used to learn more about site-specific PBM and its effect on BMD later in life. It is well known that BMD development differs in males and females (26, 27). Consequently, studies aiming to establish the timing of PBM in men and women separately are needed to make adequate recommendations concerning nutrition and physical activity.

In conclusion, aBMD decreased at the hip but increased at the spine and radius, and the increment in the long bones was explained by continued mineralization and augmented cortical thickness due to endosteal contraction in men between ages 19 and 24 yr. This cortical consolidation could considerably increase the cortical bone strength and could diminish the susceptibility to fractures during this phase in life.

Abbreviations:

 
• aBMD

  Areal BMD

 
• BMC

  bone mineral content

 
• BMD

  bone mineral density

 
• CT

  computed tomography

 
• DXA

  dual x-ray absorptiometry

 
• EC

  endosteal circumference

 
• GOOD

  Gothenburg Osteoporosis and Obesity Determinants

 
• PBM

  peak bone mass

 
• pQCT

  peripheral quantitative CT

 
• SSI

  stress strain index

 
• vBMD

  volumetric BMD.

Acknowledgments

This work was supported by the Swedish Research Council, the Swedish Foundation for Strategic Research, the European Commission, the Lundberg Foundation, the Torsten and Ragnar Söderberg's Foundation, Petrus and Augusta Hedlund's Foundation, an Avtal för Läkarutbildning och Forskning grant from the Sahlgrenska University Hospital, the Novo Nordisk Foundation, and the European Society for Clinical and Economic Aspects of Osteoporosis and Osteoarthritis-Amgen Fellowship Award (to M.L.).

Disclosure Summary: All authors have no conflict of interest.

Author notes

^*

C.O. and A.D. contributed equally to this work.