The authors state that they have no conflicts of interest.
This 2-year prospective controlled exercise intervention trial in 99 girls at Tanner stage 1, evaluating a school curriculum–based training program on a population-based level, showed that the annual gain in BMC, aBMD, and bone size was greater in the intervention group than in the controls.
Introduction: Most exercise intervention studies in children, evaluating the accrual of BMD, include volunteers and use specifically designed osteogenic exercise programs. The aim of this study was to evaluate a 2-year general school-based exercise intervention program in a population-based cohort of girls at Tanner stage 1.
Materials and Methods: Forty-nine girls 7–9 years of age in grades 1 and 2 in one school were included in a school curriculum–based exercise intervention program of general physical activity for 40 minutes per school day (200 minutes/week). Fifty healthy age-matched girls in three neighboring schools, assigned to the general Swedish school curriculum of physical activity (60 minutes/week), served as controls. All girls were premenarchal, remaining in Tanner stage 1 during the study. BMC (g) and areal BMD (aBMD; g/cm2) were measured with DXA of the total body (TB), the lumbar spine (L2–L4 vertebrae), the third lumbar vertebra (L3), the femoral neck (FN), and the leg. Volumetric BMD (vBMD; g/cm3) and bone size were calculated at L3 and FN. Total lean body mass and total fat mass were estimated from the total body scan. Height and weight were also registered. Baseline measurements were performed before the intervention was initiated. Follow-up was done after 2 years.
Results: No differences between the groups were found at baseline in age, anthropometrics, or bone parameters. The annual gain in BMC was greater in the intervention group than in the controls: L2–L4, mean 3.8 percentage points (p = 0.007); L3 vertebra, mean 7.2 percentage points (p < 0.001); legs, mean 3.0 percentage points (p = 0.07). The intervention group had a greater annual gain in aBMD: total body, mean 0.6 percentage points (p = 0.006), L2–L4, mean 1.2 percentage points (p = 0.02), L3 vertebra, mean 1.6 percentage points (p = 0.006); legs, mean 1.2 percentage points (p = 0.007). There was also a greater mean annual gain in bone size in the L3 vertebra (mean 1.8 percentage points; p < 0.001) and in the FN (mean 0.3 percentage points; p = 0.02).
Conclusions: A general school-based exercise program for 2 years for 7- to 9-year-old girls (baseline) enhances the accrual of BMC and BMD and increases bone size.
High physical activity is associated with a high peak bone mass,(1–3) and the pre- and early peripubertal years have been suggested as the ideal period for training, because during this period, there is a higher gain in BMC than the amount that is lost during the rest of the lifespan.(4,5) A variety of cross-sectional and prospective observational studies suggest that physical activity in growing children is associated with a high areal BMD (aBMD).(5–7) In contrast, only eight elegantly performed prospective controlled exercise intervention studies have been published, all but one following the children for a maximum of 1 year, to evaluate the accrual of BMD in children.(8–17) In addition, all but two of the studies include volunteers, children with such an interest in sports that they volunteered to participate in extra exercise, a fact that increases the risk of selection bias. Second, most studies use specifically designed osteogenic intervention programs, such as jumping down a small height, programs that could be difficult to motivate the children to continue for a long period, a hypothesis supported in previous reports.(18) Third, the best timing to initiate the intervention is debated, because there are data inferring that exercise does not enhance the accrual of BMD until the adolescent growth spurt, at Tanner stages 2-3.(5,19) Fourth, it must be clarified whether bone size in early years could also be increased by moderately intense exercise, which would further increase bone strength.(20–22)
Against this background, we designed this study, with the longest follow-up period published, to evaluate whether a general moderately intense exercise intervention program, possible for all children to perform, within the school curriculum in girls at Tanner stage 1, could increase the accrual of BMC and BMD and increase bone size.
MATERIALS AND METHODS
This study, the Malmö Pediatric Osteoporosis Prevention (POP) Study, is a prospective controlled exercise intervention study, following skeletal development in girls from school start. Baseline measurements in the intervention group were performed in August and September, just after school start and before the exercise intervention started. To avoid seasonal variations in a BMD, the follow-up evaluations were carried out during the same months 2 years later. The controls, selected from three neighboring schools, were evaluated in November and December, with the follow-up measurements performed during the same months but 2 years later. From these data, we calculated the annual changes in the evaluated traits (changes per 365 days). During the study period, there were two 9-week summer breaks without any school classes in both the intervention and the control group.
Four neighbor elementary schools in a middle-class area with a similar socioeconomic background in the south part of the city, all government founded with the compulsory Swedish standard curriculum, all with the same physical education curriculum, and all with the children allocated to the school according to their residential address, were invited to the study. None of the invited schools denied participation; thus, the cohort could be regarded as a cluster of convenience, the schools being the clusters and the convenience being that they are all from the same neighborhood. We invited one of the schools to participate as the intervention school; that is, no randomization was done. This school accepted, even if they had to modify their curriculum by increasing the amount of physical educational classes. We did not choose a school with an already high level of physical activity as the intervention school.
In the intervention group, all girls in grades 1 and 2 were invited to attend. Of 61 girls, 55 agreed to participate, an attendance rate of 90%. One girl was excluded at baseline, because she was 11 months younger than the second-youngest girl. At follow-up, 5 girls declined participation, leaving 49 girls with a mean age of 7.6 ± 0.6 years (range, 6.5–8.7 years) at baseline to be included in this survey. The age- and sex-matched controls were collected from three neighboring schools. Sixty-four volunteers participated at baseline, whereas at follow-up, 13 of the girls had moved out of the region or declined further participation and 1 was excluded because she was being treated with growth hormone, leaving 50 controls with a mean age of 7.9 ± 0.6 years (range, 6.8–8.9 years) at baseline to be included in this survey. All were healthy white girls without any medications known to influence bone metabolism. To further evaluate the representatives of the study sample, height and weight of all invited children were retrieved from the first acquired general school health examination, registered by the school nurses at the first medical school check in grade 1, to evaluate if selection bias had occurred.
The intervention, which started at school start just after the baseline measurement was performed, consisted of the ordinary physical activity used within the Swedish school curriculum, now increased to 40 minutes/day (200 minutes/week). The physical education classes were supervised by ordinary teachers, so no extra resources increasing the costs were needed to conduct the intervention. The physical education classes did not consist of any programs specifically designed as being osteogenetic. Instead, the classes included both indoor and outdoor general physical activities used in the Swedish school curriculum, such as a variety of ball games, running, jumping, and climbing. The teachers aimed to conduct a variety of physical activities, so as not to bore the children with repeated standardized activities. This was done with the aim of minimizing the dropouts in the long term, as is reported to occur frequently in other exercise intervention studies.(18) In the control schools, the same type of physical activities were used as in the intervention school, but at a level within the compulsory Swedish school curriculum of physical education, consisting of one to two sessions per week (60 minutes/week). Thus, the only direct modification of the school curriculum was the increased duration of physical activity. No other regular health-related modifications were performed. However, during the study period, we also provided a few irregular, other health-related activities, such as health education and health information, for the pupils, the parents, and the teachers. However, the same education was given in all four schools.
The children were evaluated dressed in light clothes with no shoes. Bone mass was measured by DXA (DPX-L version 1.3z; Lunar, Madison, WI, USA). Pediatric software was used for children with a weight <35 kg. BMC (g) and aBMD (g/cm2) were evaluated for the total body, the lumbar spine (L2–L4 vertebrae), the third lumbar vertebra (L3), the femoral neck (FN), and the legs. BMC and aBMD in the total body and the leg (both legs included) were estimated from a total body scan, in the L2–L4 and L3 regions from a lumbar spine scan, and in the FN from a hip scan. The legs were included because this is a region highly affected by most types of physical activity. The width of the L3 vertebra, estimated as the distance from one edge of the vertebra to the other, and the width of the FN, calculated as the FN area divided by the scan length of the measured area, was evaluated by the DXA scan, as previously described in the literature.(20,21) Volumetric BMD (vBMD, g/cm3) was calculated for L3 using the algorithm introduced by Carter,(23) and for the FN using the formula vBMD = BMC/estimated FN volume (π × r2 × FN length), where r = FN mid-diameter/2, assuming the FN to be cylindrical.(24) During the measurement of the lumbar spine, the child was supine, and the physiological lumbar lordosis was flattened by elevation of the knees. The precision (CV) in the measured regions, evaluated by duplicate measurements in 14 healthy young adult men and women, was 1.3–3.2% for BMC, 0.4–1.6% for aBMD, 2.8–3.1% for vBMD, 1.5–1.7% for bone width, 4.1% for total fat mass, and 0.6% for total lean body mass. Daily calibration of the machine was executed with the Lunar phantom. The technicians in our research group performed all the measurements and the software analyses. Total lean mass and total fat mass were estimated from the DXA total body scan, body weight was measured with an electric scale to the nearest 0.1 kg, and body height was measured by a wall-tapered height meter to the nearest 0.5 cm.
A questionnaire, previously used in several studies but modified to suit prepubertal children,(25–27) evaluated lifestyle factors such as socioeconomic and ethnic background, diseases, medications, fractures, dairy products, exclusion of anything in the food, coffee consumption, and physical activity in school and during leisure time. The total time spent in physical activity was calculated as the duration of activity in the school and at leisure time at baseline (after the intervention was started) and at follow-up divided by two. The questionnaire was answered together with the parents to minimize errors, with the knowledge that there are difficulties estimating an accurate level of physical activity in young children. The maturity of the children was assessed by Tanner staging,(28) conducted by our research nurses.
Informed consent was obtained from parents or guardians of participants before the study start. The study was approved by the Ethics Committee of Lund University and the Radiographic Committee at Malmo University Hospital, Malmo, Sweden. The Swedish Data Inspection Board approved the data collection and the setup of the database.
Statistical calculations were performed with Statistica, version 6.1 (StatWin). Data are presented as means ± SD. Student's t-test between means and Fisher's exact test were used for group comparisons. Analyses of variance (ANCOVAs) were used to adjust for chronological age and increment in height and weight in the follow-up evaluations to adjust for any difference in growth. The annual change for the traits, expressed in SD, was calculated as the annual absolute changes divided by the SD at baseline. Pearson's correlation test was used to correlate the total mean physical activity, calculated as the mean of the total physical activity at baseline and at follow-up, with changes in the bone parameters during the study period. p < 0.05 was considered as a statistically significant difference.
There were no differences at baseline in lifestyle factors when the intervention group and the control group were compared, except that the control group exercised more during leisure time (Table 1). All girls remained premenarcheal and at Tanner 1 during the study. After the intervention was initiated, this group spent more time on physical activity both in school and in total compared with the controls (Table 1). There was no difference at baseline in anthropometrics or bone parameters between the intervention and control groups (Table 2).
Table Table 1.. Background Factors at Baseline and Organized Physical Activity at Baseline and at Follow-up in the Girls in the Exercise Intervention Group (n = 49) and in the Control Group (n = 50)
Table Table 2.. Baseline Data, 2-Year Follow-up Data, and Annual Changes Evaluating the Effect of 2 Years of Daily Physical Activity in the School Curriculum Regarding Anthropometry and Bone Mineral Parameters in the Intervention Group (n = 49) and the Control Group (n = 50)
The annual gain in BMC was higher in the intervention group than in the control group: L2–L4, a mean difference of 0.21 SD (p = 0.007); L3 vertebra, a mean difference of 0.29 SD (p < 0.001; Fig. 1); legs, a mean difference of 0.13 SD (p = 0.07; Table 2). In addition, the annual gain in aBMD was higher in the intervention group: total body, a mean difference of 0.14 (p = 0.006); L2–L4, a mean difference of 0.10 SD (p = 0.02); L3 vertebra, a mean difference of 0.14 SD (p = 0.006; Fig. 1); legs, a mean difference of 0.15 SD (p = 0.007). The annual increase in bone width was also higher in the intervention group than in the control group: L3 vertebrae, a mean difference of 0.19 SD (p < 0.001; Fig. 1); FN, a mean difference of 0.03 SD (p = 0.02). There was also a discrepancy in the changes in FN vBMD gain, where there was a loss in the intervention group and there was a gain in the control group, with a mean difference of 0.24 SD (p = 0.002; Table 2). The annual gain in weight (mean difference, 0.12 SD; p = 0.02), total lean mass (mean difference, 0.14 SD; p = 0.01), and total fat mass (mean difference, 0.24 SD; p < 0.001) was higher in the intervention group than in the control group (Table 2). Finally, most discrepancies in the annual changes in BMC, aBMD, vBMD, and bone width remained after adjustment for age at baseline and annual changes in weight and height (Table 2). The only group differences that changed to a borderline significance were the total body aBMD and the FN width comparisons.
When all girls were included, the total duration of physical activity correlated with the annual changes in the third lumbar vertebra in BMC (r = 0.33, p = 0.001), aBMD (r = 0.37, p = 0.002), and width (r = 0.22, p = 0.03). No such correlations were found for FN. In addition, no correlation was found between the duration of activity and the annual changes in fat mass (r = 0.10, p = 0.34) or total lean body mass (r = 0.19, p = 0.07).
The drop-out analyses revealed that, based on the school record data, there were no significant differences at baseline in height, weight, or body mass index (BMI) between children who participated in the study and those who did not. The girls that participated in the study had mean values for height (cm), body mass (kg), and BMI (kg/m2) of 126.5 cm, 26.3 kg, and 16.3 kg/m2, respectively, whereas the corresponding values for girls that did not participate were 127.0 cm, 26.3 kg, and 16.2 kg/m2 (p values 0.61, 0.95, and 0.65, respectively).
This study suggests that through merely increasing the amount of normal moderate physical education in the school curriculum, there is a possibility to increase the accrual of BMC and aBMD and the gain in bone size in prepubertal girls at Tanner stage 1. It is also not unexpected to find the most obvious effects in weight-loaded regions with predominantly trabecular bone, because these regions are known to respond fastest to changes in the mechanical load.(8–17) In addition, the finding of a dose–response relationship, even if weak, between the total duration of physical activity and the gain in BMC, aBMD, and bone size, support the view that there exist an association between the duration of physical activity and the accrual of the bone parameters. When discussing physical activity in prepubertal girls as a possible prevention strategy for future fractures, our results are of great importance because both aBMD(29) and bone size(20,21) independently influence bone strength and bone resistance to fracture.
This 24-month follow-up is the longest prospective controlled exercise intervention study published in children, a study that support previous trials with a shorter follow-up.(8,10–17) However, it is still a matter of speculation whether the exercise-induced benefits are temporary or if they also remain with years of intervention, although cross-sectional and prospective uncontrolled and controlled trials do suggest that high physical activity is associated with a high peak bone mass.(2,5,7,30)
The problem with seasonal differences in growth and accrual of aBMD was solved by measuring the children in the same month both at baseline and at follow-up in both cases and controls. Another strength of this study is the high attendance rate. This indicates that our sample could be regarded as representative for Swedish school girls, which is also supported by data that showed that there were no differences in anthropometrics or bone parameters between the participants ant the nonparticipants at baseline. Furthermore, the only registered difference in lifestyle factors between the intervention group and the control group was a difference in leisure time physical activity. By using virtually all children in the school, we also included children with minimal interest and minimal participation in other physical activities, perhaps those who will achieve the most benefits from an intervention program. This is also why we chose a school as the intervention arena, because this is the only place where we could reach all children. However, the inclusion of less exercise-interested girls in the intervention group and the higher leisure time activity in the control group could obscure any beneficial effects of our intervention program. Despite this flaw, our study supports that an intervention program in school could enhance the accrual of BMC, aBMD, and bone size in girls at Tanner stage I.
Furthermore, a variety of studies evaluating the effect of exercise in children have used specifically designed exercise programs with a high osteogenic effect. One study reported, for example, that exercise including jumping down a small step up to 30 minutes, three times per week, increased the accrual of aBMD in the greater trochanter by 1.4% over 8 months.(12) Similar programs seem to be effective in the short-term perspective,(8,10–14) but also involve difficulties in motivating the children to proceed with this type of exercise for a longer period.(15,17,18) This is extremely important because continued physical activity seems to be essential for increasing peak bone mass.(2,3) However, this study implies that we could achieve similar benefits in the growing skeleton in girls using a variety of activities generally used in the physical educational classes.
We also deliberately let ordinary teachers lead the physical education, without including any activities defined by us. By doing so, the ordinary lessons were kept intact, only increased in duration compared with the control schools. Furthermore, we did not support the schools with extra resources that increased costs, to show that a similar intervention could be organized in all schools. This further increased the value of this report and enhances our inferences that physical activity could be a cost-effective strategy to increase bone strength.
Another strength of this study is the evaluation of bone size. There are previous cross-sectional studies indicating that highly intense exercise increases bone size.(31,32) Bone strength is not only dependent on aBMD but also on skeletal geometry, architecture, and bone size.(20–22,33) For example, women with spine fractures have smaller lumbar spine vertebrae but normal FN size, whereas women with hip fractures have a smaller FN size but normal vertebral body size compared with controls.(33) Also, straight mechanical calculations reveal the importance of bone size for resistance to fractures as bone strength increases by the fourth power of a tubular structure.(20–22,33) This study further improves our understanding of the exercise-induced effects by suggesting that even moderate physical activity in girls at Tanner stage 1 enhances periosteal apposition.
There are also limitations of the study. The trial is not a randomized study, that is, we can not exclude the risk of introducing selection bias. However, randomization was not possible, because the principals, the teachers, the parents, and the pupils made it clear that the school could not accept that some children were sent to physical activity during compulsory school hours while others were not. A randomization of the school, even if not being a true randomization, would perhaps have been preferable to further reduce the risk of selection bias, but because we did not include an intervention school with an already high duration of exercise at baseline, because all schools had a similar amount of physical education before the study start, the only direct modification of the curriculum was to increase the physical educational classes; this reduced the risk of introducing selection bias based on the choice of the intervention school. In addition, randomization within the classes would, through time, have led to enormous problems with children crossing over between the groups. Therefore, we accepted that one specific school was chosen as the intervention school, without individual randomization. Second, the attendance rate at baseline was lower and the proportional drop-out frequency was higher in the control schools, facts that also indicate a risk of introducing selection bias. However, because the children in the control cohort come from a similar socioeconomic area as the intervention group, the same geographic region of the city, and the same lifestyle (except the leisure time activity), ethnicity, anthropometrics, and bone parameters, and as the absolute drop-out frequency was low, it seems probable that the risk of selection bias is negligible. Third, it could be debated whether the duration of organized physical activity is the best method to quantify the level of physical activity. Even with the same duration of activity, different types of activities may have a different osteogenetic response, as could different intensities of the same type of exercise performed during the same period. In this survey, we used a self-evaluating questionnaire to assess the duration of physical activity within and outside the school curriculum, used in several previous studies.(25–27) However, the validity of using the questionnaire in children could also be discussed, because we know that there are difficulties in estimating an accurate level of physical activity in young children. However, the main purpose of this study was not to exactly define the duration of physical activity in each child or define the osteogenic activity for each type of activity. The main aim was to evaluate whether an exercise intervention program could enhance the accrual of aBMD and bone size on a group level. It is also possible that there were other changes in lifestyle factors associated with the increased exercise that were not captured by the questionnaire, changes that could affect the skeletal development. One such confounder could be that increased duration of exercise could be accompanied by an increased appetite and an increased food intake, suggested by the higher gain in weight, muscle mass, and fat content in the intervention group. The higher food intake could hypothetically affect the bone accrual. However, it is also possible that the increased weight and muscle gain could be the direct result of the exercise program, benefits that could influence the accrual of bone mass and bone size. Fourth, the estimation of bone size and vBMD in this study was done by the DXA technique. Even if this method is regarded as the gold standard when measuring aBMD, we know that the estimation of vBMD and bone size from a 2D imaging technique is associated with errors, especially in growing children.(34) This could possibly, at least partly, explain the discrepancy found when comparing the annual changes in FN vBMD in the intervention group and the controls. Fifth, during the summer breaks, we do not know the activity level in the children. Finally, there was a greater annual gain in fat mass in the intervention group compared with the controls, a discrepancy we could not explain, but a discrepancy that could influence the inferences. However, because there was no dose–response relationship between the duration of activity and gain in fat mass, this suggests that the discrepancy in fat gain was the result of other influences than the activity. There is also a possibility that exercise influences the skeleton through changes in soft tissue composition, because in the intervention group there was a higher gain in total lean mass. This is also the reason why we only adjusted for chronological age and changes in weight and height at baseline, because we only wanted to adjust for differences in growth.
Therefore, despite the discussed flaws, this study was able to conclude that a school-based exercise intervention program for 2 years during the first school years in prepubertal girls at Tanner stage 1 seems to influence the accrual of BMC and aBMD and the gain in bone size in a positive way. This study supports the view that general moderate intense physical activity could be recommended as a strategy to increase BMC, aBMD, and bone size in prepubertal children.
This study was supported by Swedish Research Council Grant K2004-73X-14080-04A, Center for Athletic Research 136/05 and 137/06, the Lund University, the Österlund, and the Region Skane Foundations.