A Six-Year Longitudinal Study of the Relationship of Physical Activity to Bone Mineral Accrual in Growing Children: The University of Saskatchewan Bone Mineral Accrual Study



To investigate the influence of physical activity on bone mineral accrual during the adolescent years, we analyzed 6 years of data from 53 girls and 60 boys. Physical activity, dietary intakes, and anthropometry were measured every 6 months and dual-energy X-ray absorptiometry scans of the total body (TB), lumbar spine (LS), and proximal femur (Hologic 2000, array mode) were collected annually. Distance and velocity curves for height and bone mineral content (BMC) were fitted for each child at several skeletal sites using a cubic spline procedure, from which ages at peak height velocity (PHV) and peak BMC velocity (PBMCV) were identified. A mean age- and gender-specific standardized activity (Z) score was calculated for each subject based on multiple yearly activity assessments collected up until age of PHV. This score was used to identify active (top quartile), average (middle 2 quartiles), or inactive (bottom quartile) groups. Two-way analysis of covariance, with height and weight at PHV controlled for, demonstrated significant physical activity and gender main effects (but no interaction) for PBMCV, for BMC accrued for 2 years around peak velocity, and for BMC at 1 year post-PBMCV for the TB and femoral neck and for physical activity but not gender at the LS (all p < 0.05). Controlling for maturational and size differences between groups, we noted a 9% and 17% greater TB BMC for active boys and girls, respectively, over their inactive peers 1 year after the age of PBMCV. We also estimated that, on average, 26% of adult TB bone mineral was accrued during the 2 years around PBMCV.


The amount of bone accrued during the growing years is a major determinant of the risk of fractures in later life.(1) Weight bearing physical activity is a modifiable determinant of peak bone mineral accrual(2) and has therefore been increasingly targeted as an important area of research. However, compared with the large number of studies addressing loading activities and bone loss in older adults, there are relatively few studies of childhood and adolescent physical activity and its association with bone mineral accretion. Most studies of childhood activity and bone mineral have been undertaken in elite sporting populations and are not therefore generalizable to normally active children. Consequently, the long term effects of physical activity on bone mineral accretion in healthy pediatric populations are not fully understood.

Because longitudinal studies that span the entire pubertal period are time consuming and, subsequently, costly, most of what we know of skeletal growth and bone mineral accrual has, for the most part, been determined from cross-sectional(3–8) and short-term longitudinal(9,10) designs. Our group has previously reported cross-sectional findings on bone mineral accrual as it relates to dietary calcium intake(11) and timing of peak bone mineral accrual relative to peak linear growth.(12) Cross-sectional studies generally report a positive association between mechanical loading from a variety of physical activities and bone mineral in preadult age groups.(13,14) However, discrepant findings in the prospective literature emphasizes the need for longitudinal studies to provide definitive answers to questions related to bone mineral accretion and mechanical loading in children.

A substantial limitation of studies of growing children is the difficulty in controlling for the considerable maturational differences in children of the same chronological age. To evaluate effectively the role of childhood and adolescent physical activity on bone mineral accrual, there is a need for longitudinal studies that not only span the entire pubertal period, but in which maturity is controlled for.(13,14) Recent prospective trials lend strong support to the contention that increased weight bearing physical activity can augment bone mineral in the prepubertal skeleton.(15,16)

The present 6-year longitudinal study evaluates the relationship between everyday physical activity and peak bone mineral accrual in a group of healthy Canadian children passing through adolescence. This study addresses two questions. First, is there a difference in magnitude of bone mineral accrual at the age of peak bone mineral content (BMC) velocity, in bone mineral accumulated 2 years around PBMCV, and in total accumulation of bone mineral at 1 year postpeak, between inactive, average, and high active children within a range of normal physical activity? Second, is there a gender difference in these bone mineral accrual variables after controlling for maturity and size differences between boys and girls? An important aspect of this study was its longitudinal design which allowed us to identify and, thus, compare the children at a common maturational landmark, the age of peak bone mineral content velocity (PBMCV). This effectively controlled for the wide range of maturational differences in boys and girls of the same chronological age(17) which was also apparent in our sample.



The subjects for this study were drawn from a longitudinal study of bone mineral accretion in growing children initiated in 1991 which has been described elsewhere.(12) Of 375 eligible students (ages 8–14) attending two elementary schools in the city of Saskatoon (population 200,000), the parents of 228 students (113 boys and 115 girls) provided written consent for their children to be involved in this study.

From the 68 boys and 72 girls remaining in the study after 6 years, we had sufficient longitudinal data over the adolescent years to fit growth curves and determine the age and value of peak height velocity (PHV) and PBMCV for 60 boys and 53 girls. These subjects represent the study population for the present investigation. All subjects were of Caucasian descent.

Bone densitometry and anthropometry

BMC of the total body (TB), posteroanterior lumbar spine (LS, L1–L4), and the femoral neck (FN) was measured annually in October or November by dual-energy X-ray absorptiometry using the Hologic 2000 QDR (Hologic, Inc., Waltham, MA, U.S.A.). The array mode was used for all scans employing enhanced global software version 7.10. TB scans were analyzed using software version 5.67A, and scans of the proximal femur and LS were analyzed using software version 4.66A. The coefficient of variation in vitro for LS was 0.50 and the coefficients of variation in vivo were 0.61, 0.91, and 0.60 for LS, FN, and TB, respectively. These values are consistent with other studies utilizing the QDR 2000 in the array mode.(18–20) To minimize operator related variability over the years, the same qualified individual analyzed all scans over the 6-year period.

Height and weight were measured semiannually to monitor growth and maturation. Height was recorded without shoes as stretch stature to the nearest 0.1 cm using a wall stadiometer, and weight was measured to the 0.01 kg on a calibrated electronic scale. Age at menarche was determined by questionnaire.

Dietary calcium intake

We performed 24-h dietary recalls semiannually. On the basis of all the recalls obtained for each child up to the age of PHV, an average daily intake of calcium was determined. The number of assessments per child until the age of peak ranged from 6 to 13.

Physical activity assessment and activity classifications

A physical activity questionnaire was administered a minimum of three times per year for the first 3 years of the study and two times per year thereafter. The physical activity questionnaire for children (PAC-Q) consists of nine items designed to provide a measure of a child's general physical activity level during the school year. Physical activity is described as “sports, games, gym, dance, or other activities that make you breathe harder, make your legs feel tired, and make you sweat.” Each item is scored on a five-point scale, with higher scores indicating higher levels of activity. The mean of these items forms a composite activity score. In diverse samples of children, the scale has consistently demonstrated acceptable internal consistency and validity.(21,22) External validity has been examined by comparing results with teacher evaluation of activity, Caltrac motion sensors, 7-day activity recalls, step tests of fitness, and leisure time activity scales. Results have been generally favorable with moderate relationships reported.(23)

To establish activity groups based on the PAC-Q inventory, an age–gender-specific Z score, based on the mean and SD for the entire large sample of a similar age, was determined for each individual on each test administration. An average Z score for all assessments up to a similar maturational benchmark (the age of PHV), was then calculated for each subject. The number of activity assessments per child ranged from a maximum of 13 for boys and 11 for girls, with a mean of 7 assessments for the boys and 6 for the girls. A subject whose average Z score fell in the lowest quartile was classified as physically inactive, while a subject whose average Z score fell in the highest quartile was classified as physically active. Those subjects who fell between the upper and lower quartiles were judged to be of average activity. In this way, 15 boys and 13 girls were classified as active, 30 boys and 27 girls as average, and 15 boys and 13 girls as inactive.

Maturational control

To control for the well documented maturational differences between adolescent boys or girls of the same chronological age, we determined the age of peak linear growth (PHV) and the age of PBMCV for each subject. This allowed us to control for size (height and weight) at a common maturational landmark (PHV). Whole-year velocity values were calculated for each subject by dividing the difference between the annual distance measurements by the age increment (the mean age increment was 0.998 ± 0.048 years). A cubic spline fit was then applied to the whole year velocity values for each child. A spline is interpolating polynomials, which uses information from neighboring points to obtain a degree of global smoothness. The cubic spline procedure was chosen over other curve-fitting protocols because it maintains the integrity of the data without transforming or modifying the underlying growth characteristics, such as the age of peak velocity or the peak velocity value. To evaluate differences in size at this common developmental landmark, the height and weight of each child at the age of PHV was determined and subsequently used in the covariance statistical analyses. BMC values were similarly determined at points on the individual velocity curves for each site representing developmental ages 1 year on either side of PBMCV as well as at the age of peak.

Statistical analysis

Pearson product moment correlations were used to assess the relationship between the independent and dependent variables. A general linear model (analysis of covariance [ANCOVA]) was used to compare BMC measurements at PBMCV and for the 2 years around PBMCV as a function of three groups (active, average, and inactive) and two genders, with height and weight at the age of PHV entered as covariates to control for differences in size. A similar statistical model was used to compare absolute values for TB, LS, and FN BMC 1 year after PBMCV. A level of significance of p < 0.05 was used and all statistical tests were two-tailed. SPSS-X software version 8.0 (SPSS, Inc., Chicago, IL, U.S.A.) was utilized for these analyses. Values are reported as means (SD) unless otherwise noted.



Sample means (SD) for age, height, and weight at the age of PBMCV and for dietary calcium intake during the year preceding PBMCV are presented (Table 1). There were no statistically significant differences in height, weight, or dietary calcium intake between groups with the exception of height in the active girls who were taller than the inactive girls (p = 0.006; Table 1). Figure 1 illustrates the mean age (SD) at PBMCV for the TB which was ∼2 years earlier for girls (12.5 (0.90) years) compared with boys (14.1 (0.95) years). This is consistent with previous reports from our group.(12,24) The 0.7 year interval between the timing of PHV and TB PBMCV for both boys and girls is less than the ∼1 year interval previously reported from our cross-sectional analyses of the larger dataset using a different curve-fitting procedure.(12) The close timing of mean age at menarche (12.7 (0.98) years) with PBMCV (12.5 (0.86) years) in girls, and as determined in the longitudinal data analyses, has also recently been reported and discussed by our group.(24)

Table Table 1. Descriptives at Peak Bone Mineral Accrual for Age, Height, Weight, and Dietary Calcium Intake by Activity Group (Inactive, Average, Active), and Gender for the 60 Boys and 53 Girls Followed Longitudinally over 6 Years
original image
Figure FIG. 1.

TB PBMCV curve illustrating velocity at peak and ages at peak BMC and peak height velocities by chronological age for boys and girls.

Physical activity

The correlation between standardized physical activity score and peak TB bone mineral accrual for boys was r = 0.39. The correlation between boys' physical activity and the total amount of bone mineral accumulated during the 2 years around the age of PBMCV was r = 0.40. The corresponding values for girls were 0.41 and 0.38, respectively. Correlations at the FN and LS were within the same range. The highest correlation (r = 0.47) between physical activity and PBMCV was observed at the LS in girls. All correlations were significant at p < 0.05.

To investigate further the relationship between physical activity and bone mineral accrual during adolescence, values for PBMCV and 2-year accrual (Table 2) were analyzed by ANCOVA (general linear model) with a three group (inactive/average/active) by two (male/female) design. To control for size differences, height and weight at the age of PHV were entered as covariates. For TB and FN, significant (p < 0.01) physical activity and gender main effects for both peak accrual and accrual over 2 years were observed with a nonsignificant interaction effect. At the LS there was a significant physical activity main effect for both accrual variables but no gender difference and no interaction effect. Pair-wise comparisons revealed significant differences between the active boys and girls combined and the inactive group for both the magnitude of TB bone mineral accrual at peak (PBMCV = 409 g/year vs. 331 g/year) and for the total amount of bone mineral accumulated during the 2 years around PBMCV (2 year accrual = 699 g vs. 582 g). These findings in favor of the active group over the inactive group were consistent and statistically significant across all sites for both groups (TB, LS [both p < 0.001] and FN [p < 0.005]) (Table 2). A comparison between the activity groups for PBMCV at all skeletal sites is provided (Fig. 2).

Table Table 2. Unadjusted Means (SD) for Peak Bone Mineral Accrual (PBMCV, G/Year) 2 Year Accrual (PBMCV ± 1 Year) and Absolute Values for BMC at 1 Year Post-PBMCV
original image
Figure FIG. 2.

TB, LS, and FN peak bone mineral accrual velocity (g/year) by inactive, average, and active physical activity groups for girls and boys. Means (SD bars) *significantly greater than inactive p < 0.005 **significantly greater than inactive p < 0.001.

Finally, a similar (3 × 2 ANCOVA) design was used to compare absolute values for BMC at the TB, LS, and FN at 1 year after PBMCV (Table 2). Significant main effects for physical activity (p < 0.01) and gender (p < 0.05) were noted for TB and FN BMC. At the LS there was a significant physical activity main effect (p < 0.05) but no gender difference. No interaction effects were observed at any site. Controlling for maturational and size differences between groups, the active boys exhibited a 9% greater BMC 1 year after peak (our closest to adult value) compared with the inactive boys. The corresponding figure for girls was 16%. For the same time period at the FN the adjusted BMC values were 7% and 11% higher for the active boys and girls, respectively, as compared with inactive groups.

Gender differences

When we compared overall accrual rates for boys and girls while controlling for height and weight, significant gender differences for PBMCV at the TB (p < 0.005) and the FN (p < 0.01), and for 2 years accrual values at these sites (TB [p < 0.001], FN [p < 0.005]) were noted. TB PBMCV was, on average, 15% greater in males than females (394 g/year vs. 342 g/year) (Fig. 2). Bone mineral accrual over 2 years followed the same pattern (696 g/year vs. 592 g/year). In contrast to the other two sites, there were no gender differences in PBMCV or 2-year bone mineral accrual at the LS.

Absolute values for TB BMC at 1 year post-PBMCV were significantly (p < 0.05) greater for boys than girls. Similarly, boys had significantly (p < 0.001) greater FN BMC compared with girls, whereas there was no gender difference for LS BMC (Table 2).


Physical activity

Childhood and adolescence have been identified as “the most critical periods of skeletal mineralization.”(25) This study demonstrates a greater peak bone mineral accrual rate and a greater bone mineral accumulation for the 2 years around peak for children in the highest activity quartile for physical activity over those in the lowest quartile within a group of normally active children. This represents a 9% and 17% greater TB BMC for active boys and girls, respectively, over their inactive peers 1 year after the age of PBMCV. Slemenda et al. have previously reported a 4–7% greater increase in BMD (as compared with our size-adjusted BMC values) for prepubertal children in the uppermost quartile of physical activity compared with those in the lowest quartile.(25)

A site-specific difference in the bone mineral accrual advantage was noted in the present study. While percentage differences between active and inactive groups were similar for boys and girls at the TB, greater differences existed between the least and most active girls at the FN as compared with the boys. It seems plausible that a lower level of physical activity for the least active girls as compared with the least active boys might offer some explanation. Differences in the timing of pubertal events and associated hormonal differences between the genders at puberty may also be involved.

To investigate whether the increased accrual values noted at puberty translated into greater absolute values for bone mineral, we compared BMC at all sites across activity groups, 1 year after PBMCV. Although this represents our cohort at its most mature, children had not yet achieved adult status. After controlling for height and weight, the high active children demonstrated ∼18% greater BMC at the LS compared with the least active group 1 year after peak. If, as reported in retrospective recall studies of retired female ballet dancers,(26) gymnasts,(27–29) and male weightlifters,(30) the benefits of childhood physical activity persist into adult life, the present findings have significant implications for adult bone health and for the prevention of osteopenia and osteoporosis in later life. It has been proposed that a 10% increase in adult bone mineral density at the FN reduces the risk of fracture at that site by one half.(31) One year after PBMCV (the most mature age for our cohort) active boys had 7% more bone mineral at the FN than inactive boys, for girls the difference was 11%. Although the results of the present study strongly suggest that recreational childhood activity promotes adult bone health, until we follow our cohort into their adult years definitive claims as to the specific benefits can not be made.

Gender differences

After controlling for maturation and size differences between boys and girls there was a gender difference in bone mineral accrued at peak, and for the 2 year interval around peak, at both the TB and the FN, but not at the spine. These findings are in agreement with Slemenda et al. who similarly reported no difference in LS bone mineral accrual for 38 boys and girls across 3 peripubertal years.(25) Quantitative computerized tomography studies have shown that reported gender differences in spine BMC postpuberty may be a function of the larger vertebral bodies in boys.(32)

Adolescent bone mineral accrual

Along with others, our longitudinal findings suggest that the amount of bone mineral accumulated during the adolescent years is substantial. If we assume a TB BMC value of 2200 g for a mature female and 2800 g for a male, then in our cohort ∼26% of final adult bone mineral status is accrued in the 2 adolescent years surrounding PBMCV. From a previous cross-sectional analysis of our data, we have reported a 36% accrual over 4 adolescent years surrounding PHV.(12) For the LS, Slemenda et al.(25) report a bone mineral accumulation of 30% of adult BMC over 3 peripubertal years as determined by Tanner staging. In the present study, using a value of 60 g as an average spinal BMC for a young adult female,(33) 32% of adult BMC was accrued in the 2 years around PBMCV. The clinical significance of these percentages can be appreciated by considering the fact that as much bone mineral is being laid down during the adolescent years as most people will lose during their entire adult lives.(12)

Dissociation between linear growth and bone mineral accrual

Our data also confirm that at adolescence there is a dissociation between linear growth and bone mineral accrual. For the TB and LS in both boys and girls, peak velocity in BMC occurred ∼0.7 years after PHV. For the FN the difference was 0.5 years. This dissociation has been reported by others(34) and suggests a transient period of relative weakness during the adolescent growth spurt. This may provide partial explanation for the increase in fractures seen in children around the time of peak linear growth.(35–37) It also suggests that activities that impose a high mechanical load on the skeleton may be contraindicated during the short time period accompanying the adolescent growth spurt.

Maturational control

In any study of growing children, the highly variable increases in size, shape, and mass that accompany normal growth are a major confound. The contribution of other factors or interventions to observed changes, independent of normal growth, is often difficult to elucidate. A strength of the present study was its longitudinal nature which allowed us to clearly identify a key maturational event (PHV) and to subsequently control for size differences at this common landmark. Peak BMC velocity was also determined for each child from longitudinal data which allowed us to compare rates of bone mineral accrual at and around peak within the same maturational time frame. Failure to account for baseline differences in maturity, or for differences in the timing or magnitude of these events during the course of a prospective study, with an initially similar cohort, may confuse or confound the outcomes of intervention studies of growing children.

Previous studies have directed researchers to the prepubertal and early pubertal years as the, seemingly, optimum time for exercise intervention.(25,38,39) Subsequent exercise intervention studies in prepubertal groups have utilized a mixed program of additional hours of physical activity (30 minutes, three times per week). These programs were successful in stimulating increased bone mineral accrual in the intervention groups at the proximal femur and LS in both girls(15) and boys.(16)


Intense exercise as undertaken by young, elite level athletes such as gymnasts(27,40,41) and weightlifters(42) is commonly associated with absolutely greater values for bone mineral at various sites. However, the athletic capabilities of most children do not necessarily allow them to participate actively in sport at this level. Further, self-selection is a well recognized potential confound in studies of elite athlete groups. The present study is unique in demonstrating that bone mineral accrual rates varied significantly among children partaking in a wide range of everyday physical activities across 6 years.

The outcomes of the present study also have implications for fracture risk in later life. It has been proposed that “patients with fractures attain a lower peak bone mass at maturity and lose bone at the same rate and for the same length of time during aging … as the rest of the population” and this, and not rapid bone loss, may explain the lower bone density found in patients with fractures relative to controls.(43) Although the findings used to support this claim presuppose a genetic basis for peak bone mass, a similar argument can be made regarding any factor that is likely to increase peak bone mass. The highly active children in the present study accrued more bone than the inactive children around the time of maximum bone mineral accrual and had significantly greater bone mineral at all sites 1 year after PBMCV compared with the least active children. It has been previously reported that as much as 30% of total adult bone mineral is accrued in the 3 years around this pubertal time period.(14,44) Our longitudinal results suggest that this may in fact be an underestimate of the magnitude of bone accrued during the critical peripubertal years.

This study is one of the first to demonstrate that the growing skeleton responds to increased everyday physical activity by increased bone mineral accrual. Whether these changes persist into adult life remains to be seen and can only be determined by long and therefore costly studies. Until such time that research provides definitive answers to these questions, lifestyle choices and public health care policies that promote physical activity during the growing years seem well advised.


This work was supported by National Health and Research Development Program (NHRDP) grant no. 6608–1261, Canada.