Moderate Exercise During Growth in Prepubertal Boys: Changes in Bone Mass, Size, Volumetric Density, and Bone Strength: A Controlled Prospective Study



Cross-sectional studies of elite athletes suggest that growth is an opportune time for exercise to increase areal bone mineral density (BMD). However, as the exercise undertaken by athletes is beyond the reach of most individuals, these studies provide little basis for making recommendations regarding the role of exercise in musculoskeletal health in the community. To determine whether moderate exercise increases bone mass, size, areal, and volumetric BMD, two socioeconomically equivalent schools were randomly allocated to be the source of an exercise group or controls. Twenty boys (mean age 10.4 years, range 8.4–11.8) allocated to 8 months of 30-minute sessions of weight-bearing physical education lessons three times weekly were compared with 20 controls matched for age, standing and sitting height, weight, and baseline areal BMD. Areal BMD, measured using dual-energy X-ray absorptiometry, increased in both groups at all sites, except at the head and arms. The increase in areal BMD in the exercise group was twice that in controls; lumbar spine (0.61 ± 0.11 vs. 0.26 ± 0.09%/month), legs (0.76 ± 0.07 vs. 0.34 ± 0.08%/month), and total body (0.32 ± 0.04 vs. 0.17 ± 0.06%/month) (all p < 0.05). In the exercise group, femoral midshaft cortical thickness increased by 0.97 ± 0.32%/month due to a 0.93 ± 0.33%/month decrease in endocortical (medullary) diameter (both p < 0.05). There was no periosteal expansion so that volumetric BMD increased by 1.14 ± 0.33%/month, (p < 0.05). Cortical thickness and volumetric BMD did not change in controls. Femoral midshaft section modulus increased by 2.34 ± 2.35 cm3 in the exercise group, and 3.04 ± 1.14 cm3 in controls (p < 0.05). The growing skeleton is sensitive to exercise. Moderate and readily accessible weight-bearing exercise undertaken before puberty may increase femoral volumetric BMD by increasing cortical thickness. Although endocortical apposition may be a less effective means of increasing bone strength than periosteal apposition, both mechanisms will result in higher cortical thickness that is likely to offset bone fragility conferred by menopause-related and age-related endocortical bone resorption.


Exercise may reduce fracture risk by increasing the amount of bone accrued during growth, by reducing age-related bone loss during adulthood, or by restoring bone already lost in old age. Of these periods, the years of growth appear to be a most opportune time for exercise to increase areal bone mineral density (BMD). For example, athletes involved in competitive sport from childhood may have 15–30% higher areal BMD than controls.(1–6)

These observations, although supportive of a role of exercise in increasing areal BMD during growth, are derived from cross-sectional studies.(1–6) Thus, selection bias may partly account for the higher areal BMD in competitive athletes; the larger musculoskeletal mass may be the reason exercise was undertaken, rather than the result of exercise. Even if the higher areal BMD was the result of exercise, this observation may reflect what is possible with Olympian endeavor rather than what is feasible on a day-to-day basis for most individuals. Thus, recommendations regarding the potential role of moderate exercise during growth for musculoskeletal health in adulthood cannot be made on the basis of these studies.

Little information is available regarding the effect of moderate exercise on areal BMD or the structural basis of any increase in areal BMD associated with moderate exercise during growth. Bone densitometry measures bone mineral content (BMC), unadjusted for bone size, or areal BMD, adjusted for the projected area of the region scanned but not its depth. Because both expressions are size dependent, a higher BMC or areal BMD in an exercise than control group may be due to growth in size rather than an increase in the amount of bone in the bone (volumetric BMD).(7) For example, in one prospective study, a 10-month moderate exercise program in premenarcheal girls was associated with increased proximal femur and spine BMC, areal BMD, but not femoral volumetric BMD.(8)

An increase in volumetric BMD may be achieved by increasing cortical thickness, trabecular number, thickness, or the true BMD of these structures.(7) Cross-sectional studies in tennis players suggest that exercise may increase cortical width by increasing periosteal apposition, which will increase bone width and bone strength.(9) If exercise increases endocortical apposition, then cortical width, BMC, and areal and volumetric BMD may increase with little improvement in bone strength because the bone width remains unchanged.(9)

No prospective studies have been done in prepubertal males, and no studies have been done to identify the structural basis or biomechanical consequences of any increase in areal or volumetric BMD associated with moderate exercise. The purpose of this prospective study was to address the following questions: Does moderate exercise before puberty in males result in increased bone mass, size, and areal and volumetric BMD? What is the structural basis and biomechanical effects of any changes in areal and volumetric BMD?



Two socioeconomically equivalent schools were randomly allocated to be an “exercise” or “control” school. Twenty boys (mean age 10.4 years, ranging 8.4–11.8) from the “exercise” school participated in an 8-month exercise program of 30 minutes of weight-bearing activity three times weekly for 32 weeks (basketball, weight training, aerobics, soccer, volleyball, gymnastics, folk and line dancing). Attendance was monitored by the physical education teacher in which a roll call system was adopted (compliance was 96%). The boys from the “control” school were matched for age, standing height, sitting height, weight, and baseline areal BMD, but received no exercise beyond that in their curriculum. (Both schools had 2 h of physical education classes per week as part of the curriculum.) All boys were prepubertal based on Tanner staging and undetectable serum testosterone and estradiol (E2). Subjects were excluded if they had illnesses (asthma, epilepsy, scoliosis, anorexia nervosa) or received medication affecting bone (anticonvulsants, corticosteroids). Informed consent was obtained from the children, their parents, teachers, and the school principal. The study was approved by the Austin and Repatriation Medical Center Ethics Committee.


Skeletal age was determined from radiographs of the wrist using the Greulich and Pyle method. Standing and sitting height were measured using a Holtain stadiometer (Holtain Ltd., Crosswell, Wales, U.K.). A Harpenden anthropometer was used to measure tibia, femur, humerus, and radius lengths and femoral condyle, biacromial, bi-iliac, and bitrochanteric widths. The coefficient of variation (CV) ranged from 0.2–1.6%. Femur length was the distance from the inferior border of the lateral epicondyle to the superior border of the greater trochanter. The CV ranged from 0.4–1.5%. BMC was measured using dual-energy X-ray absorptiometry (DPX-L; Lunar Corp., Madison, WI, U.S.A.) at baseline and 8 months. BMC (g) and areal BMD (g/cm2) are size dependent; an increase in either may reflect an increase in size rather than a regions' mineral content. Thus, the data were also expressed as a volumetric BMD (g/cm3). Femoral midshaft volumetric BMD was derived as BMC/shaft volume (which includes cortical and marrow volume) at a 4 cm2 projected area at the midshaft, assuming it to be cylindrical. These are “apparent” BMD measures: the amount of mineral in a region composed of mineral and bone marrow spaces.(7) True BMD of the cortex of the femoral midshaft (ignoring canaliculae and canals) was estimated as BMC/cortical volume (total shaft volume minus medullary volume). The pediatric anteroposterior spine program and the ruler function were used to obtain periosteal and endocortical widths and cortical thickness. CV was 0.2%. The vertebral volumetric BMD was calculated (BMC/volume) using the method of Carter et al. to derive the vertebral body volume of the third lumbar vertebra.(10) The CV was 0.9–1.5% for total body and regional BMD and BMC. Metacarpal morphometry was obtained at the midpoint of the third metacarpal using a Vernier calliper. The CV was 0.6%.

Serum testosterone (nmol/l), E2 (mIU/ml), leutinizing hormone (mIU/ml), and follicle-stimulating hormone (μl/mol) were assayed using an ACS:180 automated chemiluminescent system (Ciba Corning Diagnostics Corp., Medfield, MA, U.S.A.). The interassay CV for each test was 4–10%.

The following biomechanical indices of the femoral midshaft strength were derived assuming the midshaft was cylindrical: cross-sectional moment of inertia (CSMI) = π(rmath image − rmath image), where ro is the periosteal radius and ri is the endocortical radius,(11) the section modulus (Z) = CSMI/(1/2 bone width), and the strength index = section modulus/femoral length.(12)

BMC, areal BMD, and volumetric BMD were expressed in absolute terms and are presented as the group mean ± SEM and as a percentage of baseline. The data have been expressed as a per month value due to a difference in testing times (range 7–9 months in the exercise group and 7–11 months in the control group). There was no difference in the results when expressed per month or as a function of the mean duration of follow-up. One-sample t-tests were used to determine if the changes in BMC, areal BMD, volumetric BMD, anthropometric, biochemical, or biomechanical measurements differed from zero. Unpaired t-tests were used to determine whether changes observed during 8 months in the exercise group differed from changes in the controls.


There was no significant difference between the groups in the total hours of physical exercise performed outside school physical education lessons. The average hours/week of baseline physical activity performed (excluding the exercise intervention) was 2.55 ± 0.75 h/week in the exercise group versus 2.44 ± 0.69 h/week in the control group. After the completion of the exercise intervention program, the amount of outside activity still did not differ between groups (2.61 ± 0.71 vs. 3.68 ± 0.70 h/week), exercise versus control groups, respectively.

There were no trait differences between the groups at baseline (Tables 1 and 2). All participants remained prepubertal during the study; serum testosterone and E2 were undetectable at baseline and at 8 months in all subjects except for one boy from each group. Baseline follicle-stimulating hormone was higher in the exercise group than controls (1.91 ± 0.39 vs. 1.20 ± 0.12 nmol/l, p < 0.05, respectively). One boy was excluded due to epilepsy and another due to congenital heart disease, leaving 38 evaluable subjects.

Most anthropometric measurements increased during the 8 months (Table 1). The increases in biacromial and femoral intercondylar widths in the exercise group were greater than in controls. BMC and areal BMD increased in both groups at all sites except the arms and skull. The increases in the exercise group were twice those in the controls at most sites, reaching statistical significance at the lumbar spine, legs, and total body areal BMD (Table 1 and Fig. 1).

Table Table 1.. Baseline Data and Changes per Month in the Exercise and Control Groups: (A) Age, Body Composition, and Anthropometry, (B) Total Body and Regional Areal BMD (Mean ± SEM)
original image
Figure FIG. 1..

Changes in areal BMD (%/month) at the legs, lumbar spine, total body, pelvis, head, and arms in the exercise group and control group. *p < 0.05, **p < 0.01, change from baseline.p < 0.05,††p < 0.01, change in cases compared with change in controls. Values are mean ± SEM. NS, not significant.

In the exercise group, there was no significant change in femoral midshaft periosteal diameter. Cortical thickness increased by 0.97 ± 0.32%/month (p < 0.05) because endocortical (medullary) diameter decreased by 0.93 ± 0.33%/month (p < 0.05) (Fig. 2). Thus, femoral shaft volumetric BMD increased by 1.14 ± 0.33%/month, (p < 0.05). In controls, periosteal and endocortical diameters, cortical thickness, and volumetric BMD increased, but only significantly for the periosteal diameter. Cortical true BMD did not increase in either group (Table 2).

Table Table 2.. Baseline Data and Changes per Month in the Exercise and Control Groups: (A) Femoral Midshaft Dimensions, BMC, Areal, Volumetric, and True BMD and Biomechanical Measures of Bone Strength, (B) Third Metacarpal Dimensions, (C) Third Lumbar Vertebra Dimensions, BMC, and Volumetric BMD (Mean ± SEM)
original image
Figure FIG. 2..

Changes in femoral midshaft dimensions (%/month): periosteal diameter, endocortical diameter, and cortical thickness in the exercise group and control group. *p < 0.05, **p < 0.01, change from baseline.p < 0.05,††p < 0.01, change in cases compared with change in controls. Values are mean ± SEM. NS, not significant.

Metacarpal length, periosteal diameter, cortical thickness (but not endocortical diameter), increased in both groups (Table 2). The increase in metacarpal length and periosteal diameter was less in the exercise group than controls. Vertebral height (but not width) increased in the exercise group. Vertebral width (but not height) increased in controls so that the increase in vertebral area and volume were no different in the two groups. BMC of the third lumbar vertebra increased by similar amounts in both groups. There was no change in vertebral volumetric BMD in either group (Table 2).

Femoral midshaft CSMI and section modulus did not differ between the groups at baseline. Both measures of strength increased in both groups but significantly in the controls only (Table 2).


The increase in areal BMD at weight-bearing sites in the exercise group was double the increase in controls. The structural basis underlying these changes differed in the two groups. In the exercise group, femoral midshaft areal BMD increased because cortical thickness increased, itself the result of increased endocortical apposition rather than reduced endocortical resorption (because the medullary area decreased). For volumetric BMD to increase during growth, the increment in mass must be greater than the increment in size. This occurred in the exercise group; there was no significant increase in periosteal width so that mineral was accrued within the periosteal envelope of a bone that did not increase significantly in size—there was more bone in the same sized bone—volumetric BMD increased.

In controls, there was a growth-related increase in femoral midshaft areal BMD because bone size increased. However, the increase in size due to periosteal expansion was matched by a (nonsignificant) increase in endocortical (medullary) diameter so that the cortical bone thickness did not increase. Although cortical thickness was unchanged, the amount of cortical bone increased because the circumference of the bone increased. Volumetric BMD did not increase because the increments in mass and size were proportional. This constancy of volumetric BMD of long bones (such as the midshaft of the femur and radius) during advancing age in childhood is well documented. Even during the pubertal acceleration of growth and mineral accrual, volumetric BMD of long bones does not increase with age.(4,13–16) This constancy of volumetric BMD is not widely recognized because most studies of skeletal growth express BMC or areal BMD as a function of increasing age.(17–19) These are size-dependent expressions of “density”; as bone size and mass increase during growth, these measures of “density” increase.(7)

Exercise is commonly reported to increase periosteal apposition of bone because this surface is exposed to the greatest mechanical stresses.(20) However, in racket sports, the greater humeral cortical area of the playing versus nonplaying side may be the result of both greater periosteal expansion and greater endocortical contraction. The relative contributions of each surface vary according to the region of the humerus. In the study by Haapasalo et al., the respective relative contributions of periosteal expansion and endocortical contraction were 75:25 at the proximal humerus and 10:90 at both the mid- and distal humerus.(21) In the study by Jones et al., the respective relative contributions were 60:40 in the anteroposterior dimension and 80:20 in the mediolateral dimension in both men and women.(3) Margulies et al. reported increased tibial areal BMD during 14 weeks of military training in 18- to 21-year-old males. Because bone width did not change, the increased areal BMD was likely to be either reduced endocortical resorption or increased endocortical apposition.(22)

Thus, exercise may affect both the periosteal and endocortical surfaces, but the magnitude of the effects may vary according to whether the surface is anterior, posterior, medial, or lateral and according to whether the region is proximal, central, or distal. The factors determining whether exercise results in changes on periosteal, endocortical, or both surfaces are poorly defined. The intensity, direction, and frequency of exposure to exercise is likely to be important. However, during growth, the maturational stage of the bone surface exposed to stress may also be important; surfaces that are growing rapidly, or that are growing at the same rate but are on the accelerating rather than decelerating phase of the growth velocity curve, may respond either more or less favorably to exercise. Ruff et al. in an analysis of the work by Jones et al.(3) and of archaeological samples,(9,20) suggested that exercise may affect the periosteal surface in early adolescence, when the periosteal surface is growing rapidly, while exercise may have greater effects on the endocortical surface later in puberty, when endocortical apposition is increasing cortical thickness and reducing the medullary diameter.(20,23)

This hypothesis is not supported by the observations in this study of prepubertal males or by the study of elite prepubertal female gymnasts.(4) However, the differing exercise in racket sports and weight bearing may partly account for the discrepant observations. Hitting a tennis ball produces high-impact bending and torsional stresses with force magnitudes and directions that may not be adequately opposed by muscle contraction. This may cause greater bending and torsional stresses on the periosteum, stimulating apposition. The activities performed by the exercise group involved essentially normal loading but increased in magnitude. The bones may be already well adapted to these stresses. Muscle contraction may oppose the generation of periosteal stresses, converting these to axial compressive stress which, unlike bending and torsional stresses, are uniformly distributed through the cortex rather than being concentrated on the periosteal surface. The net increase in axial stress may be offset by an increase in cortical thickness by endocortical apposition.(24)

Increased periosteal apposition may not have occurred in the studies in elite prepubertal gymnasts,(4) because other factors may reduce longitudinal growth.(4,25) For example, reduced energy intake (relative to expenditure), may independently reduce the longitudinal bone growth, offsetting any direct anabolic effects of exercise on the periosteum.(26) However, energy imbalance is unlikely to explain the lack of periosteal expansion at weight-bearing sites during the very modest exercise undertaken in this study. An increase in periosteal diameter may have been undetected because of measurement error. The ruler function may lack the precision that is required to detect small increases in periosteal diameter. In addition, periosteal and endocortical widths were measured between the medial and lateral surfaces. Loading may result in periosteal expansion in the anteroposterior direction without change in the mediolateral diameter, producing a change in bone shape. During immobilization in growing animals, there is a failure to attain the triangular shape of the tibia, consistent with the notion that biomechanical stress influences bone shape.(24) Femoral condylar width did increase more in the exercise group than controls, while midshaft diameter did not increase. This differing regional growth in the diaphysis and metaphysis may result in a change in bone shape.

The pattern of growth of the vertebral body also differed in the two groups. Vertebral height, but not width, increased in the exercise group, and vertebral width, but not height, increased in the controls so that the increase in vertebral area and volume were no different in the two groups. Vertebral volumetric BMD did not increase in either group. However, these were prepubertal boys; vertebral volumetric BMD remains constant before puberty (like volumetric BMD of long bones) then increases late in puberty (unlike volumetric BMD in long bones), when sex hormones increase vertebral size and increase cancellous density, probably by increasing trabecular thickness rather than trabecular numbers.(27–29) The exercise group also had reduced periosteal expansion and longitudinal growth of the metacarpal. No group differences occurred at the skull.

Endocortical apposition may be less biomechanically advantageous than periosteal apposition,(9) because the endocortical apposition did not increase section modulus or strength index, despite the increased bone mass. Similarly, further analyses of the data published by Bass et al.(4) showed that the higher bone mass achieved by greater endocortical bone modeling or remodeling and reduced periosteal expansion produced no biomechanical advantage. In that study for gymnasts versus controls, respectively, neither section modulus (445.2 ± 29.19 versus 449.00 ± 33.85 cm3, NS) nor strength index (1.3 ± 0.8 versus 1.3 ± 0.8, NS) differed.(4)

Whether the higher cortical bone mass is maintained after cessation of moderate exercise is not known. Gymnastics, weight lifting, and racket sports produce similar changes in cortical thickness due to periosteal expansion and/or endocortical contraction that appear to be maintained following retirement, even into old age.(4,6,30,31) Even though the biomechanical advantage conferred by endocortical apposition may be less than that of periosteal expansion, increased endocortical apposition may be an advantage during aging given that endocortical resorption contributes to cortical thinning. Individuals with a higher peak “reserve” of cortical thickness resulting from endocortical apposition during growth may be protected against bone fragility that would result from menopause-related and age-related endocortical resorption.

The results of this, and other studies, suggest that the effects of exercise are surface specific. Thus, the evaluation of the effects of exercise during growth requires attention to periosteal and endocortical surfaces of bone in the anteroposterior and mediolateral directions of the proximal, central, and distal segments of a limb. Because growth is rapid, matching of exercise and control groups by growth velocity of the surface being studied, by bone age, and by pubertal stage will be needed; mismatching in any one of these variables will make it difficult to establish whether any group differences are due to differences in the exercise or in the stage of growth of the region. For example, in the study by Morris et al., BMC and areal BMD, but not femoral volumetric BMD, increased more in the exercise group than controls. Mismatching by bone age and growth velocity of the region appears to be a more likely explanation of the 2- to 8-fold more rapid increases in bone size than the exercise.(8)

Randomized trials of exercise in childhood using anti-fracture efficacy in childhood as an endpoint are feasible. However, given the many years of follow-up needed, it is unlikely that the effects of exercise during childhood can be evaluated using either bone mass or fracture rates in adulthood as endpoints. Thus, inferences will have to be made based on short-term studies in children using surrogate biomechanical measures of bone strength rather than either bone mass alone or fracture rates. However, exercise may alter skeletal dimensions differently in the anteroposterior and mediolateral directions. Changes in bone shape cannot be defined using the two-dimensional dual-energy X-ray absorptiometry images. Since the CSMI varies as the fourth power of the outer radius of the bone, small errors in measurement of bone size may result in large errors in the estimate of bone strength. Thus, accurate methods of measuring bone size and shape in three dimensions are needed.

In summary and conclusion, the growing skeleton is responsive to moderate exercise. The exercise is readily accessible to children and was associated with site-specific effects; femoral midshaft volumetric BMD increased while no effects were observed at non–weight-bearing sites, such as the skull, in this study or other studies.(4–6) The effects were surface specific, with endocortical apposition at the midshaft of the femur and increased femoral condylar width producing a change in bone shape. Whether these changes produce stronger bones is uncertain. A better understanding is needed of the patterns of growth (age of onset, age at acceleration, peak, and deceleration of growth velocity) of the periosteal surface that forms external bone size, and the endocortical surface that defines cortical thickness by its proximity to the periosteum. This information can then be used to design exercise programs that produce biomechanically advantageous structural changes in bone size, shape, and architecture.


The study was supported by the Dairy Research and Development Corporation of Australia. The authors thank the staff and students of St. Frances De Sales Primary School, Ivanhoe Boys Grammar School for making this work possible, and Vanessa De Luca, Senior Technologist, Jane Wilmot, Research Nurse, and Dr. Duan Yunbo for their assistance in the execution of the study.