The crucial role of physical activity in the acquisition and maintenance of bone mass is becoming widely accepted (11,14,27). The beneficial effects of exercise on bone mass and mechanical competence can be attributed to bone tissue’s sensitivity to physical forces created in the skeleton during exercise. Bone cells respond to tissue deformation, or its consequences (e.g., fluid flow), by adapting the structure to more adequately withstand future deformations (6). This adaptive process entails adding bone (either with or without prior resorption) to appropriate skeletal surfaces (7,8). However, the type of exercise modulates the anabolic response; high-impact exercises (e.g., volleyball, gymnastics) are more effective than low-impact exercises (e.g., cycling, swimming) in promoting bone gain (2,23,24). Data from animal experiments and mathematical modeling suggest that the osteogenic effects of high-impact exercise in humans are probably related to the greater strain rates associated with those activities (15,16,26). Another important factor governing the anabolic response to exercise is skeletal age. Although exercise during the adult years can retard the natural bone loss associated with aging, a much greater improvement in bone mass (and fracture resistance later in life) is achieved if vigorous exercise is engaged in during the childhood and adolescent years, when peak bone mass can still be affected (10,12,13,17).
Although the effects of exercise on bone health are becoming clear, little is known about how exercise protocols can be optimized to further promote bone mass accumulation and maintenance. A major factor to consider in designing exercise programs aimed at maintaining or improving bone mass is the sensitivity of the resident bone cell populations to mechanical stimuli. Animal loading experiments have demonstrated that bone cells desensitize soon after a loading session is initiated. The osteogenic effects of exercise could therefore reach a saturation point (where the cells are essentially unresponsive to further stimulation), perhaps after 50–100 repetitions of loading (22,26,28). These data indicate that extended exercise sessions are no more beneficial to bone mass than shorter sessions because the maximal osteogenic response is probably achieved within the first few minutes.
In a previous communication, we showed that partitioning a daily mechanical stimulus (360 repetitions of load per day) into smaller loading bouts, separated by recovery periods, enhanced bone formation over that elicited from the same stimulus applied in a single, uninterrupted loading bout (20). Presumably, the bone cells in the single, longer bout had lost sensitivity early in the session, so that load repetitions occurring toward the end of the bout were ignored. By making the loading sessions shorter and providing recovery periods between sessions (during which mechanosensitivity could be restored), bone formation improved by as much as 90%. A follow-up experiment showed that approximately 8 h of load-free recovery restores full mechanosensitivity to previously stimulated bone cells (21). Mechanical loading sessions initiated before full recovery was achieved resulted in bone formation rates that were proportional to the recovery time, indicating that even modest recovery periods can restore some degree of sensitivity.
Although our previous studies have shed light on the dynamics of bone cell mechanosensitivity, they were limited in that the duration of the loading protocol lasted from 1 to 2 wk. In light of strain-feedback models of bone adaptation (6), it was unclear whether the improvement in bone formation using short, separated bouts would be maintained if the loading protocols were implemented over a longer time period. Using a noninvasive loading model that applies well-controlled forces to the rat ulna, we sought to determine whether 4 months of loading using short loading bouts, interspersed with recovery periods, is more beneficial than single, longer daily sessions. We hypothesized that after 16 wk of loading, bone mass and structural properties of ulnae loaded for 90 cycles, 4 times·d−1 would be greater than in ulnae loaded for 360 uninterrupted cycles·d−1.
MATERIALS AND METHODS
Forty-four virgin female Sprague-Dawley (12 wk old) rats were purchased for the experiment from Harlan Sprague-Dawley (Indianapolis, IN). The rats were housed two per cage at Indiana University’s Laboratory Animal Resource Center for 15 wk before the experiment began (acclimation period) and were provided standard rat chow and water ad libitum during the acclimation and experimental periods. Body mass measurements were collected periodically during the acclimation period and 3 times/wk during the experimental (loading) period. Under ether-induced anesthesia, the right ulna of rats in the loading groups was subjected to axially applied compressive loads, using a nonsurgical loading preparation that transmits mechanical force to the ulna through the olecranon and flexed carpus (Fig. 1A) (25). The natural curvature of the ulnar diaphysis translates the axial load into a bending moment in the middiaphysis that produces tension in the lateral cortex and compression in the medial cortex. Force was applied to the ulnae by an open loop, stepper motor-driven spring linkage with an in-line load cell. All procedures performed in this experiment were conducted in AAALAC-approved facilities and were in accordance with ACSM, NIH, and Indiana University Animal Care and Use Committee guidelines.
Ten days before the start of the loading period, the rats were divided randomly into two loaded groups (N = 13/group) and two control groups (N = 9/group). The right ulnae of animals in the two loaded groups were subjected to 360 load cycles·d−1, 3 d·wk−1 for 16 consecutive weeks. Load was applied as a haversine waveform at a frequency of 2 Hz and peak load magnitude of 17 N, which elicits a compressive strain of approximately 3600 με on the medial surface of the ulnar midshaft (9). The two loaded groups differed from each another only in the timed delivery of the 360 load cycles received throughout each load day. One group was administered all 360 cycles in a single, uninterrupted session (360×1), which lasted 3 min. The other loaded group was administered the 360 cycles in four discrete bouts of 90 cycles/bout (90×4), with 3 h of recovery inserted between each of the brief (45-s long) loading bouts. We have shown previously, using the rat tibia bending model, that the 90×4 schedule enhances bone formation markedly over that produced by the 360×1 schedule after 1 wk of loading (20). All rats were allowed normal cage activity between bouts. The two control groups comprised a baseline control (BLC) group, which was sacrificed on the first loading day, and an age-matched control (AMC) group, which was sacrificed on the same day that the loaded groups were sacrificed (16 wk after baseline sacrifice). Neither control group was subjected to loading or anesthesia. In the loaded groups, the left ulnae were not loaded and served as internal controls for the loaded limb. After sacrifice, the right and left ulnae were dissected free of the articulating bones, cleaned of soft tissues, fixed in 10% neutral buffered formalin for 48 h, then transferred to 70% ethanol for storage.
Each right-left ulna pair was scanned side-by-side on the bed of a Hologic QDR-1000 x-ray densitometer (Hologic Inc., Waltham, MA) equipped with Hologic version 6.20C software. The bones were positioned with the lateral surface of the diaphysis facing down and were scanned at 0.127-mm resolution. Upon completion of each scan, mutually exclusive region of interest (ROI) boxes were drawn around the right and left bone, from which bone area (BA; cm2), bone mineral content (BMC; mg), and areal bone mineral density (aBMD; g·cm−2) measurements were collected. Two bones from the study, chosen at random, were scanned 10 times to assess reproducibility. The bone was removed from the scanner bed and repositioned between each repeat scan. From the 10 repeated measures, the coefficient of variation (CV) was calculated. The CVs for BMC and aBMD were 1.5% and 1.0%, respectively.
Each ulna was placed in a plastic tube filled with 70% ethanol and centered in the gantry of a Norland Stratec XCT Research SA+ pQCT (Stratec Electronics, Pforzheim, Germany). Two cross-sectional levels were scanned on each ulna—one at the midshaft and one through the olecranon process, using 0.46 mm collimation (4×105 counts·s−1) and 0.08 mm voxel size (Fig. 1B). The slice through the olecranon was taken 3.5 mm distal to the proximal tip of the bone and included the cortical shell and secondary spongiosa of the proximal metaphysis. For each section, the x-ray source was rotated through 180° of projection (1 block). The scans were imported into BonAlyse version 1.3 software (BonAlyse Ltd., Jyväskyla, Finland) for post hoc pQCT analyses. From the midshaft slices, total cross sectional area (CSA—area within periosteum; mm2), cortical volumetric BMD (vBMD; mg·cm−3), and maximum and minimum second moments of area (IMAX and IMIN; mm4) were calculated in BonAlyse. The second moment of area (I) reflects a structure’s resistance to bending by considering both cross-sectional area and material distribution (geometry). Beams (or long bone shafts) with the material distributed farther from the plane of bending will exhibit greater resistance to bending (I) than beams with material distributed closer to the plane of bending. This geometry-related increase in bending rigidity can occur even with reduced cross sectional area if the material is appropriately distributed. I is calculated by dividing the section into a series of small areas (pixels) and multiplying each area (dA) by its squared distance (y) from the neutral plane. This procedure is integrated over the entire cross section:MATH
This function is repeated about all possible neutral planes; the largest value is returned as IMAX, and the smallest value is returned as IMIN.
Analysis of bone envelopes at the midshaft slice was restricted to cortical bone, because trabecular bone is not normally present in the central diaphysis. From the olecranon slices, CSA and total vBMD were calculated in BonAlyse. Two additional measurements were conducted on the proximal slices to investigate trabecular bone adaptation in the proximal metaphysis. The peel mode in the Stratec software was used to separate cortical from trabecular bone, and vBMD was calculated for each bone envelope separately. Cortical bone was separated from trabecular bone using a density threshold of 900 mg·cm−3. Two bones from the study, chosen at random, were scanned 10 times to assess reproducibility. CVs for the pQCT measurements were as follows: CSA midshaft = 4.4%, cortical vBMD midshaft = 1.8%, IMAX midshaft = 2.0%, IMIN midshaft = 3.8%, CSA proximal = 4.0%, and total vBMD proximal = 1.9%.
Differences between right (loaded limb) versus left (control limb) values for all DXA and pQCT measurements were tested for significance using paired t-tests. Percent differences between right (R) and left (L) limbs were calculated as follows: (R − L)/L·100. One-way analysis of variance (ANOVA) was performed on control limb and percent difference values to detect differences among the four experimental groups. Significant ANOVAs were followed by Fisher’s PLSD post hoc tests to determine significant differences between individual experimental groups. For the t-tests, ANOVA, and post hoc tests, α = 0.05.
Two of the rats in the 360×1 group died toward the end of the experimental period, one from anesthesia-related complications and the other from unknown causes. One of the baseline control animals died during the acclimation period from unknown causes, and one animal from the age-matched control group was excluded from the DXA analysis and from the pQCT analysis of the olecranon because of tissue damage to the proximal ulna during processing. Over the 16-wk loading period, mean body mass increased slightly in each group (6–12%) but exhibited substantial fluctuation, even in the age-matched control group (Fig. 2). The age-matched control group gained significantly more weight than the 90×4 group (Fig. 2, inset), suggesting an effect of multiple ether exposures on body weight.
BMC in the left (control) ulnae was significantly different among groups (P = 0.016). Post hoc tests revealed significant differences between the baseline control group and the remaining three groups only; none of the 16-wk animals were significantly different from one another. Thus, there appears to be an aging effect but no systemic loading effect on BMC in the control limb. Paired t-tests revealed no significant differences between right and left limbs for BMC in either control group, but both loading groups exhibited significantly (P < 0.001) greater BMC in the loaded limb compared with the control limb (Table 1). Although both loaded groups showed significant loading effects, the percent difference between right and left ulnar BMC in the 90×4 group (11.7%) was 70% greater (P = 0.001) than the right versus left difference in the 360×1 group (6.9%;Fig. 3).
Areal BMD (aBMD) in the control ulna was not significantly different among groups (P = 0.68). Paired t-tests revealed no significant differences between right and left limbs for aBMD in either control group, but both loading groups exhibited significantly (P < 0.001) greater aBMD in the loaded limb when compared with the control limb (Table 1). The percent difference between right and left ulnar aBMD in the 90×4 group (8.6%) was approximately 60% greater (P = 0.012) than the right versus left difference in the 360×1 group (5.4%;Fig. 3).
Peripheral QCT measurements collected from the ulnar midshaft of control (left) limbs yielded significant differences among groups for vBMD and IMAX. Post hoc pairwise comparisons among groups performed on control limb vBMD and IMAX revealed significant differences (P < 0.05) between the baseline control group and each of the remaining three groups. The three groups sacrificed at 16 wk were not significantly different from one another. Paired t-tests showed significantly (P < 0.001) greater CSA, vBMD, IMAX, and IMIN in the loaded arm compared with the control arm in both loaded groups (Table 2). With the exception of IMIN in the baseline control group and IMAX in the age-matched control group, paired t-test P-values were not significant for any of the variables in the two control groups.
The percent difference between right and left midshaft ulna cross-sectional area was significantly greater in the loaded groups compared with the control groups (Fig. 4). Differences were also detected between loading groups—the 90×4 group exhibited 37% greater (P = 0.012) right versus left difference in CSA than the 360×1 group (Fig. 4). Percent difference (right vs left) for vBMD was significantly greater in the loaded groups compared with the control groups, but no significant differences between the two loaded groups were detected (Fig. 5). Percent difference in IMAX and IMIN was also significantly greater among the loaded groups compared with the control groups, with the exception of the 360×1 versus baseline control post hoc comparison for IMAX (Fig. 6). Between loaded groups, IMIN (but not IMAX) was significantly greater in the 90×4 group (46% greater;P < 0.001) compared with the 360×1 group (Fig. 6).
Measurements at the olecranon of the left ulnae yielded significant differences among groups in total vBMD and cortical vBMD; however, no differences in left limb CSA or trabecular bone vBMD were detected. Post hoc pairwise comparisons among groups were performed on total and cortical vBMD, and revealed significant differences (P < 0.05) between the baseline control group and each of the remaining three groups. The three groups sacrificed at 16 wk were not significantly different from one another. Paired t-tests detected no significant differences between right and left values for any of the olecranon measurements, with the exception of total vBMD in the 90×4 group (Table 2).
Our objective was to determine whether the beneficial osteogenic effects of a daily high-impact exercise protocol employing multiple short bouts would be preserved after 4 months of training. We found that ulnae from the 90×4 group had significantly greater BMC, aBMD, and minimum second moment of area at midshaft, compared with ulnae from the 360×1 group. These findings suggest that long-term (several months in duration) exercise protocols targeting bone health might result in greater returns in bone mass and structural properties if the daily exercise is partitioned into shorter, discrete bouts, rather than a single longer bout.
Both IMAX and IMIN were enhanced significantly by loading, but the load-induced increase in IMIN over controls was 5–6 times greater than the load-induced increase in IMAX over controls. The principal axes at the rat ulnar midshaft correspond roughly to the anatomical axes, with the major axis (plane along which IMAX exists) oriented in the cranial–caudal direction and the minor axis (plane along which IMIN exists) oriented in the medial–lateral direction. Previously characterized strain patterns at the midshaft during external loading (and during normal ambulation in vivo) show that bending occurs in the medial–lateral direction (25). Thus, the greatest change in strains during loading was produced on the medial and lateral surfaces of the midshaft, which is where the majority of the new bone formation occurred as a result of loading. The preferential localization of new bone to the medial and lateral surfaces explains why such large increases in IMIN were found in the loaded groups, particularly in the 90×4 group.
Cortical vBMD at midshaft was significantly increased in the two loaded groups, indicating that loading increased mineralization of the tissue. We did not, however, detect a scheduling difference in vBMD between the two loaded groups, which suggests that mineralization was not affected by the timed delivery of load cycles.
The group differences observed in final body mass corresponded to the number of ether exposures during the loading period, a result we have reported previously (20). Despite the lower final body mass in the 90×4 group, bone mass and structural properties of the right ulna were greatest in this group. The differences in final body mass only strengthen our conclusions regarding the osteogenic response to different loading schedules; if whole bone (BMC, aBMD) and midshaft bending moment (IMAX, IMIN) values are standardized by final body mass, the differences between the 90×4 and 360×1 groups become even greater (from 12 to 50% greater than in unstandardized comparisons). Thus, differences in body mass do not appear to confound the results of this experiment.
Loading failed to produce any significant increase in bone density or mass at the olecranon. The 90×4 group exhibited significantly lower total vBMD in the loaded ulna when compared with the contralateral control ulna, but inspection of Table 2 reveals that this result was produced by an abnormally high control limb value rather than a suppressed value in the loaded bone. The lack of a positive-loading effect at the olecranon might be attributable to the small strains produced at that location during loading. Using a mean cortical bone cross sectional area of 4.2 mm2 collected from the proximal pQCT scans (data not shown) and a previously calculated elastic modulus for rat cortical bone of 29.4 GPa (1), we estimate that strains at the olecranon during peak (17 N) loads reached only 140 με. Considering the much larger strains occurring at the midshaft during peak loads (∼3600 με) (9), it appears possible that strain values at the olecranon were not great enough to exceed the threshold necessary to elicit an osteogenic response.
The Centers for Disease Control and the American College of Sports Medicine have made recommendations for all adults to “accumulate 30 minutes or more of moderate-intensity physical activity on most, preferably all, days of the week”(18). These guidelines were developed to promote exercise among the sedentary population, with the goal of improving general health and lowering risks for many diseases, including heart disease, osteoporosis, cancer, hypertension, and diabetes mellitus (among others). The temporal manner in which the 30 min·d−1 of exercise should be “accumulated” is not yet clear, but it may depend on the physiologic system being targeted for improvement. DeBusk et al. (4) showed that maximal oxygen uptake (V̇O2max) increased significantly more in men who were put on a daily exercise program involving a single, 30-min session each day, when compared with men who were put on a program of similar intensity involving three 10-min sessions each day. Conversely, Ebisu (5) found that high-density lipoprotein cholesterol levels increased significantly in men who ran three times per day, compared with men who ran the same total distance each day but did so in a single bout. Our data suggest that short periods of physical activity, conducted several times each day, might improve bone mass over that achieved from a single, sustained period of daily physical activity.
These data should be considered in light of several limitations of the experiment. We only tested one multiple-bout exercise schedule (90×4). Although we have shown previously in short-term experiments (1 wk of loading) that the osteogenic response to loading varies according to the number of bouts per day (20), it is unclear whether a different multi-bout schedule (e.g., 180×2 or 60×6) would be more or less beneficial than the 90×4 schedule after 4 months. Second, the loads used in this experiment elicited strains that were in excess of those measured in humans during vigorous exercises (3). It is unclear whether proportional benefits would occur at lower strains. Finally, we did find significant right versus left differences in IMAX in the age-matched control group and in IMIN for the baseline control group. In both of these comparisons, right ulna values were significantly lower than left ulna values despite the fact that these animals were not loaded or handled beyond measuring body mass. These two results are difficult to explain. However, in light of the fact that the remaining midshaft and whole-bone measurements showed no significant side-to-side differences in the control groups, it is unlikely that these two observations suggest lateral dominance (left-handedness) in the rats used.
In conclusion, when 360 load repetitions are administered to the rat ulna 3 times·wk−1 for 16 wk, the anabolic response is much greater if the repetitions are divided into four smaller bouts of 90 repetitions/bout, separated by 3-h recovery periods, than if they are applied in a single, uninterrupted bout. These findings support other experimental evidence showing that bone cell mechanosensitivity declines quickly after initiation of a loading bout and suggest that by scheduling bone loading (exercise) sessions during times when the cells are more sensitive to mechanical stimuli, the osteogenic response can be improved. These concepts, applied to human exercise programs, hold potential for improving peak bone mass in the growing skeleton and/or preventing excessive bone loss in the aging skeleton.
We thank J. Ratliff and F. Peacock for technical assistance. This work was supported by NIH grants AR43730 and T32, AR07581.
Address for correspondence: Alexander G. Robling, Ph.D., Department of Anatomy & Cell Biology, Indiana University School of Medicine, 635 Barnhill Dr., MS 5045, Indianapolis, IN 46202; E-mail: email@example.com.
1. Akhter, M. P., D. M. Raab, C. H. Turner, D. B. Kimmel, and R. R. Recker. Characterization of in vivo strain in the rat tibia during external application of a four-point bending load. J. Biomech. 25: 1241–1246, 1992.
2. Bassey, E. J., and S. J. Ramsdale. Increase in femoral bone density in young women following high-impact exercise. Osteoporos. Int. 4: 72–75, 1994.
3. Burr, D. B., C. Milgrom, D. Fyhrie, et al. In vivo measurement of human tibial strains during vigorous activity. Bone 18: 405–410, 1996.
4. Debusk, R. F., U. Stenestrand, M. Sheehan, and W. L. Haskell. Training effects of long versus short bouts of exercise in healthy subjects. Am. J. Cardiol 65: 1010–1013, 1990.
5. Ebisu, T. Splitting the distance of endurance running: on cardiovascular endurance and blood lipids. Jpn. J. Phys. Educ. 30: 37–43, 1985.
6. Frost, H. M. Bone “mass” and the “mechanostat”: a proposal. Anat. Rec. 219: 1–9, 1987.
7. Frost, H. M. Skeletal structural adaptations to mechanical usage (SATMU): 1. Redefining Wolff’s law: the bone modeling problem. Anat. Rec. 226: 403–413, 1990.
8. Frost, H. M. Skeletal structural adaptations to mechanical usage (SATMU): 2. Redefining Wolff’s law: the remodeling problem. Anat. Rec. 226: 414–422, 1990.
9. Hsieh, Y. F., T. Wang, and C. H. Turner. Viscoelastic response of the rat loading model: implications for studies of strain-adaptive bone formation. Bone 25: 379–382, 1999.
10. Kannus, P., H. Haapasalo, M. Sankelo, et al. Effect of starting age of physical activity on bone mass in the dominant arm of tennis and squash players. Ann. Intern. Med. 123: 27–31, 1995.
11. Kannus, P., H. Sievanen, and I. Vuori. Physical loading, exercise, and bone. Bone 18: 1S–3S, 1996.
12. Khan, K., H. A. McKay, H. Haapasalo, et al. Does childhood and adolescence provide a unique opportunity for exercise to strengthen the skeleton? J. Sci. Med. Sport 3: 150–164, 2000.
13. Kontulainen, S., P. Kannus, H. Haapasalo, et al. Good maintenance of exercise-induced bone gain with decreased training of female tennis and squash players: a prospective 5-year follow-up study of young and old starters and controls. J. Bone Miner. Res. 16: 195–201, 2001.
14. Marcus, R. Exercise: moving in the right direction. J. Bone Miner. Res. 13: 1793–1796, 1998.
15. Milgrom, C., A. Finestone, A. Simkin, et al. In-vivo strain measurements to evaluate the strengthening potential of exercises on the tibial bone. J. Bone Joint Surg. Br. 82: 591–594, 2000.
16. Mosley, J. R., and L. E. Lanyon. Strain rate as a controlling influence on adaptive modeling in response to dynamic loading of the ulna in growing male rats. Bone 23: 313–318, 1998.
17. Nelson, D. A., and M. L. Bouxsein. Exercise maintains bone mass, but do people maintain exercise? J. Bone Miner. Res. 16: 202–205, 2001.
18. Pate, R. R., M. Pratt, S. N. Blair, et al. Physical activity and public health: a recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine. JAMA 273: 402–407, 1995.
19. Robling, A. G., K. M. Duijvelaar, J. V. Geevers, N. Ohashi, and C. H. Turner. Modulation of longitudinal and appositional bone growth in the rat ulna by applied static and dynamic force. Bone 29: 105–113, 2001.
20. Robling, A. G., D. B. Burr, and C. H. Turner. Partitioning a daily mechanical stimulus into discrete loading bouts improves the osteogenic response to loading. J. Bone Miner. Res. 15: 1596–1602, 2000.
21. Robling, A. G., D. B. Burr, and C. H. Turner. Recovery periods restore mechanosensitivity to dynamically loaded bone. J. Exp. Biol. 204: 3389–3399, 2001.
22. Rubin, C. T., and L. E. Lanyon. Regulation of bone formation by applied dynamic loads. J. Bone Joint Surg. [Am.] 66: 397–402, 1984.
23. Taaffe, D. R., T. L. Robinson, C. M. Snow, and R. Marcus. High-impact exercise promotes bone gain in well-trained female athletes. J. Bone Miner. Res. 12: 255–260, 1997.
24. Taaffe, D. R., C. Snow-Harter, D. A. Connolly, T. L. Robinson, M. D. Brown, and R. Marcus. Differential effects of swimming versus weight-bearing activity on bone mineral status of eumenorrheic athletes. J. Bone Miner. Res. 10: 586–593, 1995.
25. Torrance, A. G., J. R. Mosley, R. F. Suswillo, and L. E. Lanyon. Noninvasive loading of the rat ulna in vivo
induces a strain-related modeling response uncomplicated by trauma or periosteal pressure. Calcif. Tissue Int. 54: 241–247, 1994.
26. Turner, C. H. Three rules for bone adaptation to mechanical stimuli. Bone 23: 399–407, 1998.
27. Turner, C. H. Exercising the skeleton: beneficial effects of mechanical loading on bone structure. The Endocrinologist 10: 164–169, 2000.
28. Umemura, Y., T. Ishiko, T. Yamauchi, M. Kurono, and S. Mashiko. Five jumps per day increase bone mass and breaking force in rats. J. Bone Miner. Res. 12: 1480–1485, 1997.
Keywords:© 2002 Lippincott Williams & Wilkins, Inc.
MECHANICAL LOADING; BONE ADAPTATION; RECOVERY; EXERCISE; OSTEOPOROSIS; BMD