Effects of High-Impact Exercise on Ultrasonic and Biochemical Indices of Skeletal Status: A Prospective Study in Young Male Gymnasts

Abstract

Physical activity has been proposed as one strategy to enhance bone mineral acquisition during growth. The aim of this study was to determine whether frequent impact loading associated with gymnastics training confers a skeletal benefit on pre- and peripubertal male gymnasts. We measured broadband ultrasonic attenuation (BUA, dB/MHz) at the calcaneus (CBUA); ultrasound velocity (m/s) at the calcaneus (CVOS), distal radius (RVOS) and phalanx (PVOS); serum osteocalcin (OC); total alkaline phosphatase (ALP) and insulin-like growth factor-I (IGF-I) every 3–4 months over an 18-month period in elite male gymnasts and matched normoactive controls (pubertal stage ≤2). Ground reaction forces of common gymnastics maneuvers were determined using a force platform and loading histories of the upper and lower extremities approximated from video recordings. Ultrasound results were expressed as a standardized score (Z score) adjusted for age, height, and weight. At baseline, no differences were detected between the gymnasts (n = 31) and controls (n = 50) for CBUA, although ultrasound velocity at each site was higher in the gymnasts (0.6–1.5 SD) than the predicted mean in controls (p ≤ 0.001). Over 18 months, CBUA Z scores increased significantly in the gymnasts from baseline (0.3 vs. 1.0, p < 0.05, n = 18). In contrast, ultrasound velocity did not increase in either group, although CVOS and RVOS remained significantly higher in gymnasts compared with controls (range p < 0.01 and < 0.001). No differences between groups were found for OC, ALP, or IGF-I at any time. Gymnastics training was associated with on average 102 and 217 impacts per session on the upper and lower extremities, respectively, with peak magnitudes of 3.6 and 10.4 times body weight. These results suggest that frequent high-impact, weight-bearing exercise during the pre and peripubertal period may enhance the mechanical competence of the skeleton, perhaps offering an important strategy for osteoporosis prevention if the benefits are maintained.

INTRODUCTION

Since osteoporosis is a skeletal disease characterized by low bone mass and microarchitectural deterioration,¹ and up to 90% of adult peak bone mass may be attained by the end of the second decade,² maximizing both bone mass and bone quality during the growing years may offer the greatest protection against osteoporosis and fractures in later life. Physical activity has been recommended as one strategy for improving the mass, density, and/or the structural competence of bone. The results from both cross-sectional studies^{3, 4} and one intervention⁵ support the notion that exercise during growth augments bone mass, although the type, frequency, duration, and intensity necessary to elicit an osteogenic response remains to be established. Experimental studies in animals have shown that magnitude, rate, and pattern of distribution are more important osteogenic features of dynamic loading than the number of cycles.^6-8 Since immature bone appears to have a greater capacity to adapt to mechanical loading than mature bone,⁹ high-impact exercise prior to skeletal maturity may represent an appropriate strategy for maximizing bone mass.

At present, the clinical assessment of osteoporosis relies mainly on bone mineral density (BMD) measurements using dual-energy X-ray absorptiometry, which does not provide any information on bone structure and nonmass factors. Ultrasound is recognized as an alternative noninvasive technique believed to measure elements of bone quality, such as bone microarchitecture, bone elasticity, and bone density.¹⁰ Studies have shown that ultrasound can predict future fracture risk in older women, independently of BMD,¹¹ discriminate between normal and osteoporotic women,¹² and monitor skeletal responses to exercise in adults, with a sensitivity similar to densitometry techniques.¹³ In addition, changes in ultrasonic variables have been shown to correlate with changes in BMD (g/cm²) in children and adolescents, and discriminate between osteopenic and normal children.¹⁴ However, no studies have examined the long-term effects of high-impact exercise on ultrasonic properties during growth. In addition, there has been little research into exercise-related bone turnover in children, which may increase our understanding of the mechanisms underlying bone accretion during growth.

The aim of the present study was to examine the effects of high-impact, weight-bearing exercise on ultrasonic bone properties, biochemical markers of bone turnover, insulin-like growth factor-I (IGF-I) and longitudinal growth in pre- and peripubertal male gymnasts over an 18 month period. In addition, the features of loading on the skeleton associated with gymnastics training were estimated by measuring the average number and magnitude of peak ground reaction forces sustained by the musculoskeletal system during training.

MATERIALS AND METHODS

Study design

An 18-month longitudinal cohort design was used, with repeated measurements performed at 3- to 4-monthly intervals over the course of the study.

Subjects

Thirty-one pre- and peripubertal male gymnasts and 50 untrained normoactive children were initially recruited into the study. Gymnasts trained on average 16.4 h/week (range 10–29 h/week) and were recruited from the Victorian Institute of Gymnastics (V.I.G.) and two of its feeder programs where competitive gymnastics were offered. All followed a similar training schedule in preparation for the same domestic, state, and national competitions. Length of participation in formal gymnastics training ranged from 1.5 to 6 years (mean ± SE, 3.5 ± 0.2 years). Control group children were recruited from three primary schools within the northern metropolitan region of Melbourne. All schools were of similar socioeconomic background. Informed consent was obtained from all parents, children, and school principals. The study was approved by the Royal Melbourne Institute of Technology Ethics Committee and the V.I.G.

All parents and children were interviewed and completed a health questionnaire, which included items relating to previous and current medical status, use of medications, and past injuries. Control group participants also completed an activity questionnaire, modified from Grimston et al.,¹⁵ which was used to determine current and past activity patterns. All subjects were clinically healthy and were not receiving medication known to affect bone metabolism, and none had had a recent fracture. All control group children engaged in school physical education classes, although none were involved in any systematic training programs.

Sexual maturity was determined using the Tanner stage for pubic hair (self assessment), as described by Marshall and Tanner.¹⁶ This method has been shown to correlate significantly with physician assessment of pubertal maturation.¹⁷ Since testosterone levels during growth tend to parallel changes in bone age, Tanner stage, and testicular volume,¹⁸ testosterone was also used to assess sexual maturity in those subjects who provided fasted morning blood samples. Subjects were excluded if either serum total testosterone levels exceeded 1.8 nmol/l at any time^{18, 19} and/or if pubic hair development exceeded Tanner stage 2.

During the 18 months, the attrition rate was 29% for the gymnasts and 14% for the controls. Nine gymnasts were unable to complete the study due to retirement (n = 8) or injury (n = 1), with a further four excluded due to advanced maturation (pubertal stage > 2). Seven control group children left the project due to a change of address (n = 3) or personal reasons unrelated to the study (n = 4). A further eight children were excluded at different times due to advanced maturation, leaving 18 gymnasts and 35 controls, all of whom were pre- or peripubertal (pubertal stage ≤2).

Due to time demands associated with the dietary analysis and the invasive nature of the blood sampling, complete data were not available for all eligible children for all variables. However, no significant differences in age, height, weight, or ultrasonic variables were detected between participants and nonparticipants in any of the elements examined.

Ultrasound bone measurements

Measurements of ultrasound velocity (VOS in m/s) and broadband ultrasonic attenuation (BUA in dB/MHz) were made using a Contact Ultrasound Bone Analyzer (McCue Ultrasonics, CUBA Research, Winchester, U.K.), which utilizes two 17 mm 1 MHz unfocused transducers mounted coaxially. VOS was calculated from the transit time of the ultrasound pulse through bone and soft tissue and the distance between the transducers. A digital caliper was used to calculate transducer separation (accuracy 0.01 mm). BUA was calculated over 0.2–0.6 MHz. For both measurements, transducers were coupled to the skin (dominant limb) using either an aquasonic gel (calcaneus) or castor oil (distal radius and phalanx).

Measurements of BUA and VOS at the calcaneus (CBUA and CVOS, respectively) were made by placing the foot in a specially designed footplate in which the transducers remained fixed, but the footplate could be moved horizontally and vertically (accuracy 1 mm). The transducers were initially positioned midway between the midportion of the posterior calcaneum and the sustentaculum tali. Two measurements at three different sites (midpoint of the calcaneus, 4 mm anterior, and 4 mm posterior to the midpoint), with repositioning after each, were made through the calcaneus (approximate area 25 mm²), with the mean reported. VOS at the distal radius (RVOS) and proximal phalanx of the index finger (PVOS) were made with transducers mounted on a hand-held caliper (accuracy 0.01 mm). At least two consecutive measurements were made through the distal portion of the radius immediately proximal to the dorsal tubercle and through the middle of the proximal phalanx of the index finger (medial-lateral), with the mean reported. These three sites were chosen because they are subjected to pronounced loading as a result of gymnastics training, and ultrasound techniques are limited to peripheral sites surrounded by minimal soft tissue.

Short-term in vivo precision for all ultrasound measurements was assessed by duplicate measurements on 40 normoactive healthy children aged between 7.4 and 12.2 years. Coefficients of variation (CVs) were as follows: 0.39% CVOS, 0.65% RVOS, 0.68% PVOS, and 3.8% CBUA. Long-term in vitro reproducibility was determined from repeated measurements of the manufacturer's phantoms over the course of the study. For VOS and BUA, the CVs were 0.71% and 3.7%, respectively.

Biochemical and hormonal measurements

Fasted, resting morning blood samples (10 ml) were collected on Monday, Wednesday, and Friday (0730–0900 h) at each visit from a subset of gymnasts and controls (n = 20 in each group at baseline). To minimize discomfort, all children were given an anesthetic cream (EMLA; Astra Pharmaceuticals, Sydney, Australia) which was applied to the cubital region approximately 1 h prior to venipuncture. Samples were allowed to clot, serum separated by centrifugation, and stored at −80°C until assayed.

Serum total testosterone was determined by radioimmunoassay (Department of Clinical Biochemistry, Monash Medical Center, Melbourne, Australia). The sensitivity of the assay was 0.30 nmol/l. Interassay CV was 12.4%, and intra-assay CV was 6.3%. The results from the three samples (Monday, Wednesday, Friday) taken during each collection period were averaged. For all remaining assays, only Monday samples were analyzed. Serum concentrations of IGF-I were analyzed using a double antibody radioimmunoassay (Somatomedin-C Kit; Bioclone, New South Wales, Sydney, Australia), following acid ethanol extraction to release IGF-I from it binding proteins. The inter- and intra-assay CVs were 10.9% and 5.4%, respectively. Assay sensitivity was 8 ng/ml. Circulating osteocalcin (OC) levels were measured by radioimmunoassay in accordance with the general kit protocol of Phoenix Pharmaceutical's (Mountain View, CA, U.S.A.). The inter- and intra-assay CVs were 11.3% and 4.1%, respectively. Serum total alkaline phosphatase (ALP) was measured using a colorimetric technique based upon the hydrolysis of p-nitrophenylphosphate (COBAS Ready; Roche Diagnostic Systems, Basel, Switzerland). All samples were assessed together, and the intra-assay CV was 1.5%.

Dietary analysis and anthropometry

Dietary intakes were estimated from an average of five, 7-day weighed food diaries completed by a subset of gymnasts (n = 15) and controls (n = 19). Children and parents received a detailed verbal explanation and written instructions in maintaining accurate dietary records. Subjects were instructed to complete diaries in as much detail as possible, including specification of brand names when known. Data were analyzed using the DIET/3 software package (Xyris Software, Australia Pty. Ltd., ) which utilizes the NUTTAB Australian database (National Food Authority, Canberra, Australia). Mean nutrient values from all completed diaries were used in subsequent analyses.

The height of each subject was measured using a Harpenden anthropometer (British Indicators, West Sussex, U.K.) to the nearest 0.1 cm, while body mass was recorded to the nearest 0.1 kg using a portable digital scale with subjects wearing light clothing without footwear.

Video analysis and ground reaction forces

Eight gymnastics training sessions (range 3–4 h) encompassing three different phases of annual training were videotaped over a 10-month period to examine the loading histories of the upper and lower extremities. The three training phases as described by the coaches included: Routine development (RD), development of routines and new movements for competition; precompetition (PC), refinement of routines for local, state, and/or national competition; strength/conditioning (SC), general and apparatus specific strength and conditioning work. All videotaping was conducted at the V.I.G., since the gymnasts at each venue followed a similar training schedule in preparation for the same competitions. On each occasion, one member of the training group was followed for the entire session since all gymnasts followed a similar pattern. Loading was classified according to the following scheme: static hand support (time of contact >1 s), swinging handsupport (high bar), and impact loading of the upper and lower extremities (time of contact <1 s). Nonspecific activities such as walking and running were not included.

The video recordings were also used to select common training elements for an examination of the musculoskeletal loads encountered during training. Forces developed by selected activities (Fig. 1) were measured using a Kistler Multi-Component force platform (Type Z4852/c), incorporating four, three-dimensional piezoelectric load measuring transducers. For each trial, the data were sampled at 500 Hz. To simulate standard competition and training conditions, a section of sprung floor was fitted to the force platform. Peak vertical (F_z) and horizontal (F_x) ground reaction forces (in N) and time to peak vertical force (in ms) were calculated from digitized signals fed to an IBM compatible computer. For comparative purposes, all ground reaction force data were expressed in terms of individual body weight (BW). Time to peak vertical force (t_zmax) was defined as the time elapsing between initial contact and the occurrence of peak vertical force. Nine gymnasts belonging to the same training squad used for videotaping were included in this component of the study. After a normal warm-up, each subject performed between three and six practice trials of any given movement, followed by at least two trials (bare feet) for each of the movements selected, with the average result reported. To analyze forces associated with the pommel horse, the training device used by the gymnasts was placed over the section of sprung floor.

Figure FIG. 1.

Abstract of gymnastics movements analyzed on the force platform.

Statistical analysis

All statistical analyses were conducted using the Statistical Package for the Social Sciences for Windows (Release 6.1, SPSS; Norusis/SPSS, Inc., Chicago, IL, U.S.A.). Comparisons between gymnasts and controls were performed in a number of ways. First, Student's t-tests or analysis of covariance (ANCOVA), adjusted for age, height, and/or weight, were used to test for differences between the groups at baseline. Because serum total testosterone and total ALP values were not normally distributed, logarithmic transformations preceded all statistical analyses on these data. Second, all ultrasound measurements for the gymnasts were expressed as Z scores (derived from the control group children). The influence of age, height, and weight on ultrasonic properties were estimated by linear regression analysis of the measurements in the control subjects. Ultrasound measurements for each subject were standardized by calculating the difference between the observed and predicted score (based upon the fitted equation and the observed characteristics of that subject), divided by the square root of the estimated variance.²⁰ The resultant standardized value (Z score), is a measure of the deviation of an individual's score above or below the expected value for a reference subject, adjusted for the covariance on a scale with zero mean and unit SD, so that 95% of the normal population will have a Z score between ±1.96.²⁰ A one-sample t-test was used to determine whether the Z scores of the gymnasts differed significantly from the predicted mean (zero) in controls. One-way repeated measures analysis of variance was used to determine whether the ultrasound Z scores for the gymnasts changed during the study. Third, linear regression analysis was used to examine the relationships between percentage change in CBUA and potential predictors including anthropometric, hormonal, dietary, and baseline ultrasound measurements. Fourth, longitudinal hormonal changes were analyzed using separate 2 × 6 (group × time) mixed design ANCOVA with age and/or height as covariates. All absolute and Z scores were expressed as mean ± SEM, unless otherwise stated.

RESULTS

Anthropometry and maturation

Table 1 shows that the gymnasts were on average 0.7 years older (p < 0.05) than controls at baseline, although no differences were detected between the groups for either Tanner stage or serum total testosterone. When anthropometric results for gymnasts were expressed as Z scores (age adjusted), they were found to be 0.5 ± 0.2 SD shorter and 0.4 ± 0.2 SD lighter than controls (both p < 0.05). During 18 months, there were no significant differences in the rate of growth between the gymnasts and controls for height (5.1 ± 0.2 vs. 5.4 ± 0.1 cm/year, respectively) or weight (3.1 ± 0.2 vs. 3.4 ± 0.2 kg/year, respectively), although the magnitude of the initial differences persisted throughout the study.

Table Table 1.. Chronological Age, Anthropometry, Ultrasound Bone Measurements, Insulin-like Growth Factor-I, Testosterone, and Biochemical Markers of Bone Turnover in Male Gymnasts and Controls at Baseline

Ultrasound bone measurements

At baseline, CBUA did not differ between the gymnasts and controls (Table 1), even when expressed as a Z score, adjusted for age, height, and weight (0.18 ± 0.16). In contrast, the age-, height-, and weight-adjusted Z scores for CVOS (0.83 ± 0.17), RVOS (1.51 ± 0.25), and PVOS (0.65 ± 0.17) were significantly greater (all p ≤ 0.001) than the predicted mean (zero) in controls.

The longitudinal changes (absolute and Z scores for gymnasts) for all ultrasound measurements in the 18 gymnasts and 35 controls, who remained at pubertal stage ≤2, are shown in Fig. 2. During the 18 months of follow-up, CBUA increased significantly in gymnasts (12.8 ± 3.1%, p < 0.001), but not in controls (7.2 ± 3.0%, p > 0.05). CBUA Z scores for the gymnasts increased significantly from baseline (0.3 vs. 1.0, p < 0.05). In contrast, VOS at each site did not increase in either gymnasts or controls, and no change was detected relative to baseline for VOS Z scores, even after the higher initial values in the gymnasts were included as covariates. However, CVOS and RVOS remained significantly greater than the predicted mean in controls throughout the study (p ranging from < 0.01 to < 0.001). When transducer separation was compared at each site, no significant intergroup differences were detected at any time throughout the study.

Figure FIG. 2.

Longitudinal data showing changes (absolute and Z scores for gymnasts) for CBUA (A,B), CVOS (C,D), RVOS (E,F), and PVOS (G,H), in male gymnasts and controls across 18 months. *p < 0.05,^#p < 0.01, **p < 0.001 compared with the predicted mean (zero) in controls. Like letters indicate significant difference in the Z scores of the gymnasts relative to baseline (^A,Bp < 0.05). Values are mean ± SE.

Since VOS at the calcaneus, distal radius, and phalanx did not change over the course of the study, predictors of change were not sought for these sites. For the percentage change in CBUA in the gymnasts, a positive correlation was found with baseline IGF-I (r = 0.67, p < 0.05). In addition, baseline CBUA values were found to be inversely related to the percentage change in CBUA (r = 0.57, p < 0.05), indicating that those gymnasts with the lowest initial values had the greater increase over time. No other significant relationships were detected.

IGF-I and biochemical markers

Serum concentrations of IGF-I, OC, and total ALP did not differ significantly between gymnasts and controls at the commencement of the study (Table 1). IGF-I correlated with testosterone in gymnasts and controls (r = 0.55–0.58, p < 0.01) and CVOS in the gymnasts (r = 0.69, p < 0.001). During 18 months of follow-up, there was no significant increase nor difference in IGF-I between the groups. Although serum OC increased in both gymnasts and controls over the course of the study (p < 0.001), no intergroup differences were detected (Fig. 3). In addition, total ALP did not increase nor differ between the groups at any time, and neither biochemical marker correlated with any of the ultrasonic bone measurements in either group.

Figure FIG. 3.

Longitudinal changes in serum concentrations of OC in male gymnasts and controls across 18 months. Absolute means (± SE) are presented, but significance tests are from ANCOVA (adjusted for age). Like letters indicate significant within-group difference^A,Bp < 0.001.

Diet

In comparison with controls, gymnasts had higher dietary intake of protein (79 ± 4 vs. 66 ± 2 g/day, p < 0.01), calcium (998 ± 84 vs. 761 ± 36 mg/day, p < 0.05) and phosphorus (1325 ± 77 vs. 1114 ± 34 mg/day, p < 0.05). No differences were detected for total kilojoule (8408 ± 222 vs. 7853 ± 205 kJ/day), carbohydrate (267 ± 11 vs. 252 ± 8 g/day) or fat intake (75 ± 2 vs. 74 ± 2 g/day). After adjusting for differences in diet, the magnitude of the differences between groups for all ultrasound measurements persisted.

Video analysis and ground reaction forces

Video analysis of gymnastics training revealed that impact loads on the upper and lower extremities from all forms of activity averaged 102 and 217 repetitions per session, respectively. Static support loading by BW on the wrists and hands ranged between 11 and 16 minutes per session, while tensile loading produced during swinging activities on the high bar averaged 4.5 minutes per session. A summary of the peak vertical (F_z) and horizontal (F_x) ground reaction forces and time to peak vertical force of common gymnastics manoeuvres are presented in Table 2. Mean peak vertical ground reaction forces sustained by the upper and lower extremities ranged between 1.5 and 3.6 times BW and 3.7 and 10.4 times BW, respectively. Horizontal ground reaction forces ranged between 1.4 and 4.8 times BW for the lower extremities and 0.7 and 1.0 times BW for the upper extremities. While these forces were associated with a rapid rise to peak F_z (17–70 ms), there was a trend for time to reach peak F_z to decrease as the magnitude of peak F_z and F_x increased.

Table Table 2.. Mean Peak Vertical (F_z) and Horizontal (F_x) Ground Reaction Forces (Expressed Relative to Body Weight; BW) and Time to Peak Vertical Force (t_zmax) for Common Gymnastics Maneuvers (Mean ± SD)

DISCUSSION

The main finding of this 18-month prospective study was that participation in high-level gymnastics during the pre- and peripubertal period was associated with a 12.8% increase in BUA at the calcaneus in male gymnasts. Previous studies in young female gymnasts have suggested that the loading characteristics associated with gymnastics have a positive effect on BMD at both axial and appendicular sites,^{3, 4, 21} although information on the number and magnitude of loads sustained by the musculoskeletal system during training has not been reported. Based upon Frost's²² theory of mechanical usage set points, mechanical deformation (strain) must exceed a certain threshold to elicit a positive skeletal response. In the present study, we found that the gymnasts sustained on average 102 and 217 impacts per session on the upper and lower extremities, respectively. The magnitudes of such impacts peaked at 3.6 and 10.4 times BW and were associated with a high rate of loading (rapid rise to peak vertical force). Since the growing skeleton appears to have a greater capacity to adapt to loading than the mature skeleton,⁹ and the osteogenic response to loading is maximal with high magnitude strains applied at high rates engendering unusual strain distributions,^6-8 the significant increase in BUA at the calcaneus in the gymnasts may be attributed directly to the impact loading activities associated with daily training. In support of the potential for gymnastics to elicit a positive skeletal response, a recent study in prepubertal female gymnasts found that the increase in BMD (g/cm²) of load bearing sites, such as the arms and legs, over a 12-month period was 30–85% greater than in bone-aged matched controls.²¹ Given that the increase in BUA for both gymnasts (12.8%) and controls (7.2%) in the present study were greater than the precision error of the CUBA Research instrument (3.7%), it is unlikely that changes were due to measurement error.

Although the precise skeletal properties reflected in the measurement of BUA remain to be established, evidence suggests that it is related to bone structure (connectivity, trabecular orientation, spacing), in addition to bone density.^23-25 Hence, it is plausible that the greater increase reported in the gymnasts was associated with exercise-induced alterations in the pattern, spacing, connectivity, and/or thickness of trabecular elements. Although we cannot exclude the possibility that part of the change may have been associated with an increase in density per se, the lack of a significant change for CVOS, which is believed to primarily reflect bone density and elasticity, adds support to the argument that architectural alterations were the most likely explanation. These findings are consistent with a recent study which found that CBUA and trabecular bone mass in the tibia (quantitative computed tomography), decreased in two cosmonauts after 1–6 months in space, whereas little change was detected for CVOS and cortical bone mass of the tibia.²⁶

When the change in CBUA in the gymnasts was related to the initial baseline values in each subject, a significant negative correlation was detected (r = −0.57), indicating that gymnasts with the lowest initial values experienced the greatest gains over the course of the study. This finding adds support to Frost's mechanostat theory, which predicts that mechanically induced strains will increase the structure, mass, and/or density of bone, until it adapts to the load with a subthreshold strain.²² Therefore, the osteogenic response to loading for a given force is likely to be less in individuals with denser or stronger bone. In support of this notion, a study of male military recruits found that those recruits with the lowest baseline BMD had the greatest increases after 9 weeks of intense training.²⁷

An interesting finding of the present study was that IGF-I correlated positively with the percentage change in CBUA in the gymnasts. Previous research in healthy elderly women has shown IGF-I to be independently associated with both VOS and BUA, indicating that the activity of the GH–IGF-I axis may play an important role in skeletal integrity during aging.²⁸ Furthermore, the results from animal studies have shown that IGF-I stimulates bone formation on existing trabecular surfaces, increasing trabecular width and area, but not trabecular number.²⁹ While no studies appear to have examined the relationship between IGF-I and ultrasound measurements in physically active children, there is evidence of greater BMD in high-level female gymnasts compared with less active controls, despite significantly reduced IGF-I levels in the gymnasts.^{4, 30} Although no detectable differences were observed between gymnasts and controls for IGF-I in the present study, Lean et al.³¹ reported an increase in the expression of IGF-I mRNA in mechanically stimulated osteocytes, suggesting a role for local IGF-I in adaptive changes to loading. The possibility that part of the change in the present study may have also been mediated by IGF-binding proteins, which modulate IGF-I action, cannot be ruled out.³² These concerns notwithstanding, the finding that neither changes in height, weight, diet, nor any of the hormonal factors analyzed, could explain the greater increase in BUA in the gymnasts, adds further support to the idea that the loading regime was the major factor in any adaptive skeletal response.

In contrast to the significant increase in CBUA in the gymnasts, VOS did not increase in either group, although CVOS and RVOS remained significantly higher in gymnasts compared with controls throughout the duration of the study. Unlike most other sporting activities, gymnastics requires use of the upper extremities in weight bearing. The finding that RVOS and PVOS were higher in gymnasts at baseline provides support for a local effect of mechanical loading on bone formation. Similar results were apparent for BMD in the arms of prepubertal female gymnasts compared with nonathletic controls.^{3, 21} Although the upper extremities of the gymnasts in the present study were subjected to a combination of compressive, torsional, and traction forces, RVOS and PVOS did not change relative to baseline. While there is evidence to suggest that bone should respond in a positive manner to increased mechanical loading, Nichols et al.³⁰ proposed that an equilibrium point may be established such that no further skeletal gains are achieved unless the intensity of the training program (loading regime) is increased to provide bone with the necessary stimulus to reach a new threshold for adaptation. Since the average weekly training time for the gymnasts did not change significantly (16.4 ± 0.9 vs. 18.6 ± 1.5 h/week), it is conceivable that the bones of the gymnasts had already adapted optimally to this form and pattern of loading, such that the greatest changes had occurred prior to the commencement of the study.

A number of other factors may also help to explain the absence of VOS changes in the present investigation. A recent study in prepubertal children found that VOS through the patella was negatively related to physical activity.³³ The authors speculated that a delay in the development of bone microarchitecture may have resulted in excessive microdamage in more active children during this period, resulting in lower velocity values. While further studies are needed to test this hypothesis, the lack of change in RVOS and PVOS in the present study may also be attributed to the fact that both these sites contain a high proportion of cortical bone, which is metabolically less active than trabecular bone.³⁴ Furthermore, previous research has shown that much of the age-related increase in bone mineral (BMC and areal BMD) during growth results from an increase in bone size, rather than to greater mineralization per se.³⁵ If VOS is related to volumetric BMD rather than to bone size, one may speculate that the lack of change in velocity may have been due to a proportional increase of bone in the bone over the 18-month period in both gymnasts and controls. In addition, the failure to detect changes in VOS may have also been partly due to the large baseline velocities associated with bone and soft tissue (>1540 m/s).³⁶

Previous research has shown that a large proportion of the age-related increase in bone mineral content (BMC) and areal BMD measured by absorptiometric techniques, is associated with increases in height and weight,³⁷ which are to a large extent genetically determined. Since the gymnasts in the present study were characterized by small stature and low BW, it is unlikely that they had a genetic predisposition for higher than normal ultrasound values when compared with controls. Although selection bias cannot be ruled out as an explanation for the initial differences in VOS between the groups, we have previously reported a positive relationship between VOS and training time in high-level male gymnasts.³⁸

Although it was anticipated that gymnastics training would provide a sufficient stimulus to elicit an increase in osteoblastic activity, no significant differences were found between the groups for either serum OC nor total ALP. While this result contrasts with several recent intervention studies examining the bone turnover response to exercise in adolescent males,³⁹ others have found no differences,⁴⁰ or even a decrease²¹ in bone formation markers in athletes compared with controls. In light of the proposal that the loaded sites of the gymnasts in the present study appear to have already adapted optimally to the high functional loads associated with training, perhaps the maximal effect of exercise on bone formation also occurred early in the course of the adaptation to training. The lack of significant intergroup differences may also be attributed to the fact that biochemical markers cannot provide specific information about different skeletal envelopes, which may have different rates of bone turnover. In addition, since markers of bone turnover increase in parallel with height and the accumulation of bone mineral,⁴¹ perhaps the normal changes associated with growth concealed any training effects. It is likely that a combination of more specific and sensitive bone formation and resorption markers may have given a more accurate evaluation of the effects of exercise on bone metabolism.

In summary, we found that BUA at the calcaneus increased significantly in pre- and peripubertal male gymnasts over the 18-month study period. Although VOS at the calcaneus, distal radius, and phalanx were higher in gymnasts compared with controls at baseline, no significant change was observed in either group. Since ultrasound is believed to measure both qualitative and quantitative aspects of bone, it would appear that the frequent impact, weight-bearing activities associated with gymnastics were sufficient to enhance the mechanical competence of bone. The finding that CBUA and CVOS did not exhibit a similar pattern of change may be explained by alterations in bone microarchitecture that may or may not be associated with an increase in bone density. While further research is needed to determine the precise skeletal properties measured by ultrasonic parameters, it would appear that ultrasound (BUA) provides a safe, noninvasive means of comparing exercise-related skeletal changes in physically active children. High-impact, weight-bearing exercise during the pre- and peripubertal years may represent an ideal approach to enhance the mechanical competence of bone, although further research is needed to determine whether the skeletal benefits gained during this period are maintained in adulthood.

Acknowledgements

The authors gratefully acknowledge the Faculty of Biomedical and Health Sciences and the Department of Human Biology and Movement Science, RMIT University, Bundoora, Australia for funding provided.