Longitudinal Bone Mineral Density Changes in Female Child Artistic Gymnasts



Changes in bone mineral density (BMD), and related factors, in female child artistic gymnasts (n = 9) and their age- (±0.3 years), height- (±2.8 cm), and weight- (±1.7 kg) matched controls (n = 9) were prospectively examined. It was hypothesized that gymnasts would possess higher BMD at baseline, 6, and 12 months later and have greater gains in BMD over 1 year compared with controls. BMD (g/cm2) of the total proximal femur (TPF), Ward's triangle (WT), trochanter (Troch), femoral neck (FN), lumbar spine (LS, L1–L4), and total body (TB) were measured by dual-energy X-ray absorptiometry. Physical activity was measured by a 7-day recall; daily dietary intakes of energy and nutrients were estimated from 3-day records. Serum osteocalcin and urinary pyridinium cross-links were measured by radioimmunoassay and high performance liquid chromatography, respectively. Gymnasts versus controls possessed significantly higher BMD at all sites measured. Although not significantly different (p > 0.05), gymnasts compared with controls had moderately larger percentage changes in Troch (% Δ = 8.6 ± 3.0 vs. 3.8 ± 5.1%, d = 0.41), FN (% Δ = 6.1 ± 1.2 vs. 3.9 ± 1.6%, d = 0.55), LS (% Δ = 7.8 ± 1.1 vs. 6.8 ± 1.6%, d = 0.26), and TB BMD (% Δ = 5.6 ± 0.8 vs. 3.4 ± 0.7%, d = 0.98) as evidenced by the magnitude of the effect sizes (d). Gymnasts versus controls possessed a lower percentage body fat (p < 0.01) and engaged in more hours of very hard activity (p < 0.0001). Calcium, as a percentage of adequate intake, decreased over 12 months (p < 0.01), and urinary cross-links significantly decreased over 6 months in both groups. Female child gymnasts possess higher BMD at the TPF and related sites, LS, and TB compared with nongymnast controls, and 1 year of gymnastics training moderately increases Troch, FN, LS, and TB BMD for gymnasts compared with controls. These findings lend support to the idea that gymnastics training in childhood helps maximize peak BMD.


Cross-sectional studies indicate that college1-5 and former college6 artistic gymnasts have higher bone mineral density (BMD) than both nonathlete controls1-5 and other athletes such as swimmers3, 5 and runners.2, 3 It is uncertain why gymnasts have higher BMD, especially given reports of more restrictive eating patterns1, 7-9 and menstrual cycle irregularities1, 2 in gymnasts compared with controls. It has been hypothesized that the osteogenic effects of gymnastics may result from the unique and high mechanical loads this activity presents to the skeleton. Ground reaction forces of up to 12 times body weight have been reported during participation in gymnastics.10

One limitation of cross-sectional studies is that solid conclusions cannot be drawn about how group differences emerged. Childhood and adolescence are key stages for BMD development since nearly all of adult bone mass is achieved by the end of adolescence.11, 12 However, in our prior studies showing that college and former college gymnasts possess greater BMD than controls,1, 6 it is uncertain whether the group differences emerged during childhood, adolescence, or young adulthood. A second limitation of cross-sectional studies is that the groups being compared (e.g., gymnasts vs. controls) can differ on a variable of interest (e.g., BMD) because of some unmeasured variable rather than on the factor logically expected to be important (e.g., gymnastics training). Since it is widely accepted that multiple factors, such as genetics, physical activity, diet, and menstrual status and history,13-17 influence BMD, it is important to follow up cross-sectional findings with longitudinal investigations. The purpose of this study, therefore, was to prospectively examine changes in BMD and related factors, including body composition, physical activity, and selected nutrient intakes, in female child artistic gymnasts compared with nongymnast controls. It was hypothesized that gymnasts of this age group would have higher BMD at all time points and greater gains in BMD over a 1-year period compared with controls, despite lower body fat (BF) and nutrient intakes in gymnasts.


Study participants

Eighteen young females, aged 8–13 years, from the Athens and Atlanta, GA, U.S.A., areas were observed over a 12-month period. Nine artistic gymnasts and nine nongymnast controls from a previous, larger cross-sectional study on BMD and gymnastics activity in premenarcheal girls18 were included in the present study. Each original control (n = 16) was matched to a gymnast (n = 16) based on age (±0.35 years), height (±2.6 cm), and weight (±1.5 kg). Of the original 32 participants, only those gymnasts who completed all testing sessions and continued gymnastics training for 12 months and their matched controls were included in the present BMD, body composition, and physical activity analyses. One gymnast and one control withdrew from the longitudinal study. Three gymnasts and one control did not complete all testing sessions. Two gymnasts discontinued training. Thus, seven of the original gymnast–control pairs were excluded from the current study.

During this study period, gymnasts continued to compete in artistic gymnastics at a U.S.A. Gymnastics level 5 or higher (out of 10, with level 5 being the lowest regionally competitive level). Artistic gymnasts trained in floor, uneven bars, balance beam, and vault routines with training time devoted proportionally among all four performance areas. Only two controls currently participated in and had a history of participation in an organized sport, including soccer or martial arts. All controls did not participate in and had no history of gymnastics training. All participants were healthy, reported an absence of bone disease, and did not smoke or take medications known to affect bone metabolism (including oral contraceptives). Participants continued to be premenarcheal for the duration of the study. All subjects were Caucasian, except for one Hispanic control.

This study was approved by the Institutional Review Board for Human Subjects at The University of Georgia. All procedures and consent requirements were explained by an investigator, after which each participant and one parent gave written informed consent.


During a 3-h testing period, anthropometric, body composition, physical activity, and general health and demographic data were collected as were fasting blood and urine samples. All procedures were conducted at the Clinical and Sports Nutrition Laboratory, The University of Georgia. Baseline data (0 months) were collected between October 1, 1995, and June 15, 1996. Each participant returned to the laboratory at 6 months (5.5–8.5 months) and 12 months (11.5–13.5 months) after her initial or baseline testing date for follow-up testing; thus, 1-year data collection was completed by June 28, 1997.

Anthropometric data

Lightly clothed participants without shoes had body weights measured on a double-beam balance scale and standing heights measured against a measuring tape secured to the wall. Body weights and standing heights were recorded to the nearest 0.10 kg and 0.10 cm, respectively.

BMD and soft tissue mass

BMD (g/cm2) of the nondominant total proximal femur (TPF) (including Ward's triangle [WT], trochanter [Troch], and femoral neck [FN]), lumbar spine (LS, L1–L4), and total body (TB) was measured by dual-energy X-ray absorptiometry (QDR-1000 W; Hologic, Inc., Waltham, MA, U.S.A.) using the standard hip protocol, standard spine protocol, and version 5.71 of the Enhanced Whole Body Analysis software, respectively. Fat-free soft tissue (FFST) mass, fat mass (FM), and percentage BF were measured from the whole body scan. For consistency, all scans were analyzed by one investigator. To adjust for a possible confounding effect of bone size on BMD, bone mineral apparent density (BMAD, g/cm3) was calculated for the FN, LS, and TB as recommended by Katzman et al.19 Throughout the study, quality control measures were adhered to as previously described.1, 18 The in vivo coefficient of variation (CV) for TB, TPF, FN, WT, and LS BMD are 0.58, 0.89, 3.0, 3.11, and 0.63%, respectively (n = 10).1 Scans of the phantom LS (model DPA/QDR-1; Hologic, Inc.; x-caliber anthropometric spine phantom) during the 21 months of the study resulted in a CV of 0.28% (n = 204 scans).

Physical activity

The seven-day physical activity recall (PAR)20 was used to estimate each participant's typical hours of daily activity. In interview style, time engaged in sleep, moderate, hard, and very hard activities during the previous week were ascertained from each participant. Light activity and mean daily energy expenditure were calculated according to Blair.21 The PAR has been shown to be valid and reliable in 10- to 11-year-old children.22

Dietary intake

Three-day diet record forms along with verbal and written instructions for home completion were provided to each participant and her parent. Each participant recorded foods, beverages, and dietary supplements consumed for 2 weekdays and 1 weekend day of the participant's choosing. Preparation methods, brand names, and portion sizes were noted on the diet records. Parental assistance was provided as needed. Training for completion of diet records was provided to the participant and parent by a registered dietitian or trained research assistant. During training, food models were used to facilitate accurate estimation of portion sizes by the participant.

The dietitian contacted the participant-parent pair by telephone for diet record clarification as needed. Diet records were analyzed, and mean daily energy, macronutrient, calcium, phosphorus, and vitamin D intakes were estimated with The Food Processor® Nutrition Analysis Software for Windows (version 6.09; ESHA Research, Salem, OR, U.S.A.) by the dietitian. Despite repeated efforts to obtain diet records, several participants did not return their records; thus matching criteria for gymnast–control pairs was not employed for these data. Presented are baseline and 12-month dietary intake data from 7 gymnasts and 12 of the original 16 controls who completed diet records.

Bone modeling markers

Serum osteocalcin and urinary pyridinium cross-links were used to estimate the rates of bone formation and resorption, respectively. Between the hours of 0700 and 0900, participants provided blood and urine samples after an overnight fast. Venous blood (10 ml) was drawn and stored on ice until centrifuged (within 2 h of venipuncture). After centrifugation (20 minutes at 3000 rpm), serum was pipetted into storage tubes and frozen at −20°C. Batch determination of osteocalcin levels were conducted by radioimmunoassay.23 For young individuals, inter- and intra-assay CV for osteocalcin are <10% and 5%, respectively.23

Urine samples were placed into brown paper bags and refrigerated for 2 h. In darkness, 2 ml aliquots of urine were pipetted into storage containers which were subsequently frozen at −20°C. After submitting hydrolyzed samples to a prefractionation procedure,24 pyridinoline (Pyr) and deoxypyridinoline (Dpyr) were measured by high performance liquid chromatography. Peaks were detected by fluorescence,25 quantitated by external standards, and normalized for urinary creatinine (Pyr/Cr, Dpyr/Cr; #555; Sigma Diagnostics, St. Louis, MO, U.S.A.). For young healthy females, the interassay CV for Pyr/Cr and Dpyr/Cr are 3.8% and 5.9%, respectively.26

Blood and urine were collected at the baseline and 6-month time points only because no participants were willing to provide biological samples at 12 months. Furthermore, only a small subsample of the original 32 participants consented to blood and urine collection at baseline and 6 months. Therefore, gymnast–control matching was not employed. Baseline and 6-month bone marker data are presented for four gymnasts and eight controls.

Statistical analyses

Data were analyzed with the Statistical Analysis System (SAS for Macintosh; version 6.10; SAS Institute, Cary, NC, U.S.A.). To test the hypothesis that gymnasts would have greater BMD at all time points, paired t-tests were used to compare BMD of gymnasts and matched controls at baseline, 6, and 12 months. A 2 × 3 analysis of covariance (group × time) with repeated measures on the time factor and baseline BMD as the covariate was used to test the hypothesis that gymnasts would have greater gains in BMD than controls over time. A 2 × 3 analysis of variance (group × time) with repeated measures on the time factor was used to examine differences in FFST mass, FM,% BF, and physical activity data between and within groups. A 2 × 2 analysis of variance (group × time) with repeated measures on the time factor was used for analyses of bone modeling markers and dietary intake data. Means ± SEM (unless otherwise noted) are reported with significant differences of p < 0.05 identified.

In addition to reporting whether the p values were statistically significant (i.e., p < 0.05) or not, the magnitude of differences (i.e., the size of the effect) between or within participant groups also has been included. The effect size metric employed was Cohen's d,27 which was calculated by taking the difference between two means and dividing by the pooled SD:

equation image

Leading statisticians, such as Cohen28 and Rosnow and Rosenthal,29 have emphasized that two studies with different sample sizes yet identical results (i.e., having an equal magnitude of an observed effect such as the same correlation between two variables) can be inappropriately interpreted as having different outcomes because the study with a larger sample achieves statistical significance while the other does not. Thus, selected effect sizes are reported herein to facilitate readers' ability to compare the size of the effects from the present investigation to both past and future related research. It has been suggested that small, moderate, and large effects are represented as 0.20, 0.50, and 0.80 SD, respectively.27


Anthropometrics and soft tissue mass

Selected group characteristics are presented in Table 1. Age, height, weight, and FFST mass did not differ between gymnasts and controls at any time point. While both groups had significant increases in age, height, weight, and FFST mass over time (p < 0.0001 for all four variables), there were no differences between groups over time (group × time interaction). Gymnasts had significantly lower FM (p < 0.05) and percentage BF (p < 0.01) at all measurement points compared with controls. Although FM significantly increased in both groups (p < 0.001) over time, differences between groups over time were not observed. At the end of the study, gymnasts had trained in their sport for 7.1 ± 0.6 years (range = 5.0–9.0 years) and for 15.7 ± 1.6 h per week (range = 9.0–25.0 h/week). The soccer-trained control had participated in soccer seasonally for 2 years (between ages 9–11 years) while the control training in martial arts had participated year-round for 3.5 years (between ages 9–13 years).

Table Table 1. Selected Characteristics of Participants
original image

Bone mineral density

Gymnasts possessed significantly higher BMD at all sites at all time points compared with controls (Table 2). Both groups exhibited a significant (p < 0.05) increase in WT BMD over 12 months (Fig. 1). Between group and within group changes over time at other sites (TPF, Troch, FN, LS, and TB BMD) did not achieve statistical significance of p < 0.05. However, the increase in BMD for gymnasts at these other sites was consistently greater than for controls when expressed as a percentage change from baseline: Troch (% Δ = 8.6 ± 3.0 vs. 3.8 ± 5.1%), FN (% Δ = 6.1 ± 1.2 vs. 3.9 ± 1.6%), LS (% Δ = 7.8 ± 1.1 vs. 6.8 ± 1.6%), and TB BMD (% Δ = 5.6 ± 0.8 vs. 3.4 ± 0.7%). While group differences at these sites did not achieve statistical significance, the magnitude of the percentage change between groups represented moderate to large effects: Troch (% Δ d = 0.41), FN (% Δ d = 0.55), LS (% Δ d = 0.26), and TB BMD (% Δ d = 0.98).

Table Table 2. Bone Mineral Density Measurements at Baseline, 6-, and 12-month Intervals for Female Child Gymnasts and Controls
original image
Figure FIG. 1.

Absolute change in Ward's triangle (WT), trochanter (Troch), femoral neck (FN), lumbar spine (LS), and total body (TB) bone mineral density (BMD) over 12 months in female child gymnasts and controls. Values are means ± SEM for n = 8 matched gymnast-control pairs for WT, Troch, and FN BMD and n = 9 pairs for LS and TB BMD. *p < 0.05 for the effect over time; †d ≥ 0.50 to ≤0.80 and §d ≥ 0.81 for differences between groups over time.

Bone mineral apparent density

Gymnasts and controls were not significantly different in FN, LS, and TB BMAD when baseline BMAD was covaried. FN BMAD significantly decreased (p < 0.05) in gymnasts (0.1894 ± 0.0106 to 0.1764 ± 0.0064 g/cm3, d = 0.52) and controls (0.1615 ± 0.0058 to 0.1583 ± 0.0053 g/cm3, d = 0.20) from baseline to 12 months, but the change over time in FN BMAD did not differ between groups. TB BMAD significantly decreased (p < 0.05) in gymnasts (0.1002 ± 0.0018 to 0.0993 ± 0.0014 g/cm3, d = 0.21) and controls (0.1031 ± 0.0021 to 0.0970 ± 0.0018 g/cm3, d = 1.09) from baseline to 12 months; however, the absolute decrease in TB BMAD over time was significantly less for gymnasts than controls (p < 0.05), respectively (% change = −0.9 ± 0.9 vs. −6.4 ± 2.0%, d = 0.74). LS BMAD did not significantly change within or between groups over time.

Physical activity

Estimated energy expenditure was significantly higher in gymnasts compared with controls at baseline (6255 ± 410 vs. 4757 ± 322 kJ/day, p < 0.05), 6 months (7109 ± 628 vs. 6167 ± 582 kJ/day, p < 0.05), and 12 months (7397 ± 360 vs. 5598 ± 456 kJ/day, p < 0.05). Both groups had significant increases in estimated energy expenditure over time (p < 0.0001); however, the change over time did not differ between groups. Gymnasts engaged in significantly more hours of very hard physical activity compared with controls at the baseline (1.6 ± 0.2 vs. 0.3 ± 0.1 h/day, p < 0.0001), 6 months (1.5 ± 0.3 vs. 0.9 ± 0.2 h/day, p < 0.0001), and 12 months (2.0 ± 0.2 vs. 0.4 ± 0.1 h/day, p < 0.0001) test periods.

Dietary intake

Significant differences in mean (± SD) daily total energy and macronutrient intakes between gymnasts and controls did not exist and did not significantly change within or between groups over time. Gymnasts consumed 8761 ± 1213 and 8736 ± 3590 kJ/day while controls consumed 9100 ± 2280 and 8883 ± 2920 kJ/day at baseline and 12 months, respectively. Energy and protein intakes exceeded 97% and 173%, respectively, of the recommended dietary allowances (RDA) at both time points for both groups. For both groups and measurement points, total energy was comprised of 13–16% protein, 51–57% carbohydrate, and 30–32% dietary fat.

Gymnasts consumed significantly less protein per kilograms FFST mass (means ± SD) than controls, respectively (p < 0.05) at baseline (2.5 ± 0.4 vs. 3.9 ± 1.6 g/kg of FFST mass, d = −1.07) and 12 months (2.4 ± 0.7 vs. 3.3 ± 1.5 g/kg FFST mass, d = −0.70), but protein intake per kilogram of FFST mass did not significantly change within or between groups over time. For carbohydrate intake per kilogram of FFST mass, significant differences between groups were not found. Significant decreases in carbohydrate intake relative to FFST mass over 12 months occurred in both gymnasts (10.8 ± 2.2 to 8.9 ± 1.6 g/kg of FFST mass, d = 0.99) and controls (12.6 ± 3.1 to 10.7 ± 3.3 g/kg FFST mass, d = 0.59), although the change over time did not differ between groups.

Significant differences in mean (± SD) daily intakes of calcium, phosphorus, and vitamin D between gymnasts and controls did not exist and did not significantly change within or between groups over time. Gymnasts consumed 1104 ± 409 and 961 ± 471 mg of calcium/day while controls consumed 1119 ± 304 and 995 ± 284 mg of calcium/day at baseline and 12 months, respectively. Although the reduction in calcium intake (mg) over time was not significant, both gymnasts (85 to 74%, d = 0.87) and controls (86 to 77%, d = 0.69) had significant decreases in the percentage adequate intake (AI) of calcium consumed (p < 0.01). In both groups, phosphorus, and vitamin D intakes met or exceeded 93% of the AI at both time points.

Bone modeling markers

Both gymnasts (n = 4) and controls (n = 8) had significant decreases in Pyr/Cr (318.0 ± 52.0 to 212.2 ± 18.1 nmol/mmol, d = 1.36, and 269.7 ± 41.7 to 175.1 ± 13.3 nmol/mmol, d = 1.08, respectively, p < 0.01), and Dpyr/Cr (92.0 ± 13.3 to 58.3 ± 4.0 nmol/mmol, d = 1.69, and 75.7 ± 8.7 to 44.0 ± 3.2 nmol/mmol, d = 1.69, respectively, p < 0.001) from baseline to 6 months. The ratio of osteocalcin to Dpyr/Cr (OC:Dpyr/Cr) significantly increased (p < 0.01) in both groups over time, but a significant change in the ratio of osteocalcin to Pyr/Cr (OC:Pyr/Cr) was not observed.

Moderate effects sizes were found for the increase in osteocalcin from baseline to 6 months within gymnasts (12.3 ± 0.5 to 13.8 ± 1.7 ng/ml, d = −0.60) and decrease within controls (13.7 ± 1.7 to 11.9 ± 1.0 ng/ml, d = 0.45). Significant differences in osteocalcin, Pyr/Cr, Dpyr/Cr, OC:Pyr/Cr, or OC:Dpyr/Cr were not found between groups at either time point, and changes in bone turnover markers and ratios did not differ between groups over time.


Female child artistic gymnasts have significantly higher BMD at the TPF, WT, Troch, FN, LS, and TB compared with nongymnast controls of the same age, height, and weight, at baseline as well as 6 and 12 months later. Moreover, effect size data revealed that 1 year of gymnastics training in young competitive gymnasts results in moderately greater gains in Troch, FN, LS, and TB BMD in gymnasts compared with controls. Although differences between groups over time were not significant after covarying for the initial BMD values, this analysis lacks strong statistical power. Nonetheless, the reporting of effect sizes provides useful information that is independent of the relatively small sample size.28, 29 These BMD gains in gymnasts were observed despite higher BMD at baseline, lower BF, and a decrease in percentage AI of dietary calcium. These findings support the idea that artistic gymnastics training contributes to the development of high TB and site-specific BMD observed in child gymnasts.18, 30-32

Several investigators report that FN and LS BMD begins to increase significantly at age 12,33-35 with the greatest gains observed after the onset of puberty.35 In females, the average annual rate of gain in FN and LS BMD is ∼4.9% and ∼4.8%, respectively, until age 20 years.35 While controls in the current study had a lower percentage increase in FN BMD (3.9 ± 1.6%) but a higher percentage increase in LS BMD (6.8 ± 1.6%) than published norms,35 gymnasts had an even higher gain at both sites (6.1 ± 1.2% and 7.8 ± 1.1%, respectively) than controls and other young females. Faulkner et al.36 reported a 3.7% difference in TB BMD between 10- and 11-year-old females. Although this was a cross-sectional comparison,36 it is similar to our longitudinal findings (3.4 ± 0.7% gain/year) in controls. Thus, the available literature also supports the idea that there is an approximate 2.0% annual gain in gymnasts above that observed in less active females.

Frost's37 “mechanostat” theory suggests that bone adapts to mechanical forces placed upon it until a new steady state is achieved. In the case of gymnastics training, modeling drifts or accelerated osteoblast and decelerated osteoclast activity may occur in response to the high mechanical loads of gymnastics maneuvers, resulting in net gains in BMD beyond expected. In conjunction with Frost's theory, Rubin and Lanyon38 showed that the magnitude of the mechanical load was crucial to bone strength. Ground reaction forces generated at the hip and spine during the performance of gymnastics maneuvers have been reported to be greater than 10 times body weight in young adult females.10 Considering that gymnasts in the present study had a significant increase in body weight from baseline to 12 months and that training occurred for at least 9 h/week, the increases in Troch, FN, LS, and TB BMD are consistent with the aforementioned theories. As body weight increased, the magnitude of mechanical load may have increased, resulting in accelerated modeling drifts. These drifts may continue to occur throughout gymnasts' careers, culminating in the significantly higher BMD observed in college-age gymnasts1-2, 4, 5 and former college gymnasts6, 32 compared with nongymnast controls. Although cross-sectional in design, a study conducted by Kirchner et al.6 indicated that former college gymnasts had higher WT, FN, LS, and TB BMD than age-, height-, and weight-matched controls, but lower WT and FN BMD than current college gymnasts. While there may be a residual effect of gymnastics training, cessation of intense gymnastics training after the college years may be associated with reductions in WT and FN BMD and supports the magnitude of load theory.38

Another important contributor to the high Troch, FN, LS, and TB BMD in child female gymnasts compared with controls may be differences in the composition of soft tissue mass. At all time points, gymnasts in the present study had higher FFST mass (although not statistically significant) but significantly lower FM and percentage BF compared with controls. In children, lean mass has a positive association with BMD.39-41 Specifically it has been suggested that in girls, 80% of the variability in BMD is explained by FFST mass.36 Although genetic predisposition for high FFST mass and low FM in gymnasts cannot be discounted, the highly muscular body type appears beneficial to BMD. In fact, a recent study by Morris et al.42 showed that changes in FFST mass accounted for 10–58% of the variability in BMD accrual in 9- to 10-year-old premenarcheal girls who completed a 10-month exercise intervention program.

One-year changes in TPF and WT BMD within gymnasts and controls were similar and consistent with normative data.14, 35 It was expected that changes in gymnasts would have been higher than controls based on findings of Bass et al.32 in which 1 year of gymnastics training (for 15–36 h/week) did significantly increase BMD in elite child gymnasts compared with controls. Lack of such a finding in the present study may be explained by the age of participants, years of gymnastics training, the higher initial BMD values, and menarcheal status. Participants were 10 years old when the study began and had been training for 5–9 years. Gymnasts may have had greater percentage gains in BMD at the TPF and WT before 10 years of age. This is supported by values of TPF and WT BMD that were 17% and 35% higher, respectively, in gymnasts than the controls at the beginning of the study and remained higher after 12 months. The stimulus provided at the TPF in the first few years of gymnastics training may have produced the large cross-sectional differences in BMD observed between gymnasts and controls at the TPF and WT. All gymnasts (and controls) remained premenarcheal throughout the study, and because an association between participation in gymnastics and a later onset of menarche exists,1, 43 accelerated gains in BMD at the TPF and WT may be forthcoming in these gymnasts.

Changes in the dimension of bone, not just mineral acquisition, contribute to longitudinal differences in bone density measures19; thus, BMAD of the FN, LS, and TB were examined. Over 12 months, TB BMAD decreased by 6.4% in controls but only 0.9% in gymnasts. Because gymnasts and controls had similar increases in standing height from baseline to 12 months, this suggests that TB mineral acquisition in gymnasts paralleled changes in bone dimension while in controls it did not. Use of BMAD versus BMD measures may not be more salient for describing bone attributes, however, since either measure equally predicts bone strength and breaking point.44

Participation in gymnastics resulted in gymnasts spending more time in very hard physical activity than controls; this was reflected in a higher mean estimated energy expenditure for gymnasts. The higher energy expenditure of the gymnasts in this study is supported by findings of Davies et al.45 in which total energy expenditure in Chinese gymnasts (age 7) was 1.98 times greater than basal metabolic rate. For total energy expenditure to be nearly two times the basal metabolic rate, engagement in very hard or high intensity activities had to be eight times longer per day for gymnasts than normal girls. Because these Chinese gymnasts lived and trained in specialized schools designed to produce elite gymnasts, the additional exercise was comprised of gymnastics activities.45 Perhaps the ∼2 h of very hard physical activity per day of gymnasts of the present study was an adequate osteogenic stimulus to produce high Troch, FN, LS, and TB BMD. Moreover, energy expenditure, as determined by the 7-day PAR, was more closely linked to energy intake for gymnasts than controls in the present study. The lower energy intake relative to expenditure may relate to the significantly lower percentage BF possessed by gymnasts versus controls.

Mean total energy and macronutrient intakes were similar between and within groups at baseline and 12 months and over time. These findings are in contrast to other reports of restrictive eating patterns in young elite8, 9 and college-age1, 7 gymnasts but similar to nonelite, but competitive, young gymnasts.30, 31 When normalized for FFST mass, protein intake was lower in gymnasts versus controls. Presently, the AI for calcium in 9–13 year olds is 1300 mg/day.46 Although the study samples for BMD and dietary intake data are somewhat different, the slight decrease in calcium intake does not appear detrimental to gymnasts in the present study as moderate to large increases in Troch, FN, LS, and TB BMD occurred while the percentage AI of calcium consumed decreased. This supports other findings indicating that physical activity, specifically hours of weight-bearing activity, may be a more important determinant of BMD than calcium intake.14-15, 47

Normal serum osteocalcin levels for 10- and 11-year-old females are 26 ng/ml and 27 ng/ml, respectively.48 Mean osteocalcin levels were below published norms for both groups at each time point. While values were low, they were still consistent and within the low end (10–15 ng/ml) of the normal range for freshly drawn samples of serum from prepubertal girls studied in the laboratory in which osteocalcin samples were analyzed.49 Additionally, inconsistencies in absolute osteocalcin values have been found among laboratories based on methodologies such as examined osteocalcin fragments or subforms and use of kits or laboratory-developed assays.49 Normal urinary Pyr/Cr and Dpyr/Cr values range from 79–387 nmol/mmol and 25–104 nmol/mmol, respectively, for 10–12 year olds.50 Mean Pyr/Cr and Dpyr/Cr levels were within normal limits for both groups at both measurement times. This is the first study to report longitudinal changes in markers of bone turnover in young gymnasts and controls, but due to the limited number of participants in both groups that provided biological samples, caution is advised in the interpretation of these results. The large effect observed in both gymnasts and controls in the reduction in Pyr/Cr and Dpyr/Cr over time may suggest a trend for a decrease in bone resorption from baseline to 6 months. Over time, controls also had a moderate decrease in osteocalcin while gymnasts had a moderate increase in osteocalcin (based on effects sizes). This may indicate that bone formation was more active in gymnasts than controls and is supported by the similar increase (11%) observed in osteocalcin for gymnasts in the present study compared with college gymnasts (12% increase) during 27 weeks of intensive training.7 Such changes in osteocalcin, Pyr/Cr, and Dpyr/Cr in gymnasts may provide insight as to increases in Troch, FN, LS, and TB BMD (Table 2) in gymnasts compared with controls as well as direction for future research.

In summary, this study is the first to report longitudinal changes in BMD in nonelite, yet competitive, female child artistic gymnasts. BMD at the TPF, and related femoral sites, LS, and TB, were higher in gymnasts compared with controls at three time points: baseline and 6 and 12 months later. Thus, BMD is elevated in female child gymnasts before the age of 10 years. As evidenced by effect sizes (although not statistically significantly different [p > 0.05]), gymnasts had moderately greater percentage gains in Troch, FN, LS, and TB BMD during 12 months of gymnastics training compared with nongymnast controls, even though gymnasts and controls were matched for age, height, and weight throughout the study. These findings suggest that gymnastics training contributes to the development of BMD in girls undergoing bone modeling. Further longitudinal research with girls younger than 10 years of age and prior to initiation of gymnastics training is warranted to better identify the periods of BMD development during which gymnastics has the greatest impact on bone.


Gratitude is expressed to the study participants and parents for their devotion to this research. We thank Caren M. Gundberg for completion of osteocalcin assays, Christopher M. Modlesky for technical assistance, and Dr. Deborah D. Godwin for statistical advice. This research was supported, in part, by the Sports, Cardiovascular, and Wellness Practice Group of The American Dietetic Association (1996) and Busch Biomedical Award (6–49373). Support was also provided to Sharon M. Nickols-Richardson through The American Dietetic Association Fuschia Lucille Johnson Scholarship (1996–97) and the Hazel and Gene Franklin Scholarship (1996–97) from the College of Family and Consumer Sciences, The University of Georgia.