Effect of Long-Term Impact-Loading on Mass, Size, and Estimated Strength of Humerus and Radius of Female Racquet-Sports Players: A Peripheral Quantitative Computed Tomography Study Between Young and Old Starters and Controls†
Bone characteristics of the humeral shaft and distal radius were measured from 64 female tennis and squash players and their 27 age-, height-, and weight-matched controls with peripheral quantitative tomography (pQCT) and dual energy X-ray absorptiometry (DXA). The players were divided into two groups according to the starting age of their tennis or squash training (either before or after menarche) to examine the possible differences in the loading-induced changes in bone structure and volumetric density. The following pQCT variables were used: bone mineral content, total cross-sectional area of bone (TotA), cross-sectional area of the marrow cavity (CavA) and that of the cortical bone (CoA), cortical wall thickness (CWT), volumetric density of the cortical bone (CoD) and trabecular bone (TrD), and torsional bone strength index for the shaft (BSIt) and compressional bone strength index for the bone end (BSIc). These bone strength indices were compared with the DXA-derived areal bone mineral density (aBMD) to assess how well the latter represents the effect of mechanical loading on apparent bone strength. At the humeral shaft, the loaded arm's greater bone mineral content (an average 19% side-to-side difference in young starters and 9% in old starters), was caused by an enlarged cortex (CoA; side-to-side differences 20% and 9%, respectively). The loaded humerus seemed to have grown periosteally (the CavA did not differ between the sites), leading to 26% and 11% side-to-side BSIt differences in the young and old starters, respectively. CoD was equal between the arms (−1% difference in both player groups). The side-to-side differences in the young starters' bone mineral content, CoA, TotA, CWT, and BSIt were 8–22% higher than those of the controls and 8–14% higher than those of the old starters. Old starters' bone mineral content, CoA, and BSIt side-to-side differences were 6–7% greater than those in the controls. The DXA-derived side-to-side aBMD difference was 7% greater in young starters compared with that of the old starters and 14% compared with that in controls, whereas the difference between old starters and controls was 6%, in favor of the former. All these between-group differences were statistically significant. At the distal radius, the player groups differed significantly from controls in the side-to-side bone mineral content, TrD, and aBMD differences only: the young starters' bone mineral content difference was 9% greater, TrD and aBMD differences were 5% greater than those in the controls, and the old starters' TrD and aBMD differences were both 7% greater than those in the controls. In summary, in both of the female player groups, the structural adaptation of the humeral shaft to long-term loading seemed to be achieved through periosteal enlargement of the bone cortex, although this adaptation was clearly better in the young starters. Exercise-induced cortical enlargement was not so clear at the distal radius (a trabecular bone site), and the study suggested that at long bone ends, the trabecular density could be a modifiable factor to built a stronger bone structure. Conventional DXA-based aBMD measurement detected the intergroup differences in the exercise-induced bone gains, although, because it measured two dimensions of bone only, it seemed to underestimate the effect of exercise on the apparent bone strength, especially if the playing had been started during the growing years.
OSTEOPOROTIC FRACTURES CONSTITUTE a major health threat worldwide. A prevention strategy to solve this health epidemic should include maximization and maintenance of bone strength and prevention of elderly people's fall-related traumas.(1) In both bone research and clinical practice, the current method-of-choice to evaluate bone fragility and fracture risk is dual energy X-ray absorptiometry (DXA), particularly the DXA-derived areal bone mineral density (aBMD, g/cm2).(2–4) The benefits of DXA, that is, its precision, short examination times, low radiation dose, and the ability to predict the fracture risk at the population level, make it suitable for most clinical purposes.(2, 3) In most cases DXA is the only noninvasive method available to assess osteoporosis. However, the planar nature of DXA makes the assessment of the geometry and true composition of bone impossible, and the evaluation of bone fragility of an individual is likely to be an approximation at best.(5) In fact, the relevance of the aBMD as a surrogate of bone fragility at individual level has been recently challenged, and the need for more sophisticated noninvasive methods to characterize bone accurately has been brought up.(6–10)
The ultimate strength of bone material, that is, the maximum stress it can sustain, is bone's intrinsic material property, and thus, independent of bone size and shape, whereas the force required to break the entire bone (fracture load) will vary with bone structure (size and shape).(11, 12) Because the whole bone's mechanical integrity is determined by the interaction of both the material and structural properties of bone,(9) the structural alterations caused by mechanical loading or other treatment can be missed or underestimated with the DXA measurements unless they are complemented by other methods, such as peripheral quantitative computed tomography (pQCT), that provide mechanically more relevant information on bone structure.(13)
In our previous pQCT study of male tennis players, the playing-to-nonplaying arm comparison of bone structure showed that the extra bone mineral at the playing site was largely attributable to increased cross-sectional area, not apparent bone density.(14) The DXA-based studies, in turn, have shown that the greatest bone mineral content (BMC) increase can be obtained if bones are loaded during the growing years.(15–21) However, it is not known whether the bone's structural response to loading differs if exercise has been started during the growing years rather than in adulthood.
The aim of the study was to compare the pQCT-derived structural side-to-side differences in the upper-arm bones of players who had started training during the growing years to those who started in adulthood and to nonplaying controls. The second aim was to compare whether the loading-induced structural alterations differed at the cortical and trabecular bone sites. Finally, based on the fact noted above, the third aim was to assess whether the DXA-based exercise studies have underestimated the effects of mechanical loading on bone.
MATERIALS AND METHODS
We examined the bone characteristics of both humeral shafts and distal radii of 64 formerly nationally ranked female racquet-sports players and their 27 age-, height-, and weight-matched controls. The players were divided into two groups according to the starting age of their tennis or squash training. The mean starting age of tennis or squash was 10.5 (SD, 2.2) years among those players who had started their training before or at menarche (young starters, n = 36), whereas it was 26.4 (8.0) years among those players who had begun training a minimum 1 year after menarche (old starters, n = 28). At the time of the measurements, the mean age of the young starters was 26.5 (8.0) years and of the old starters was 44.4 (10.5) years. The average training frequency of the young starters was 1.4 (1.3) times a week and that of the old starters was 2.0 (1.4) times a week. For most players, the playing intensity level was described as recreational. Five years earlier, the time when all of them were national top-level players, the young starters played on average 4.7 (2.7) times a week and the old starters played 4.0 (1.4) times a week.(22) None of the controls had been involved in physical activity or work affecting the dominant extremity only. All subjects were clinically healthy and they did not have past upper extremity fractures. They were informed of the study procedure, purposes, and known risks, and all gave their informed consent. Characteristics of the subjects are summarized in Table 1.
Table Table 1.. Characteristics of the Subjects (Mean, SD)
Information about training and medical history of the subjects was obtained through an interview, which included questions about lifelong history of physical activity, possible special diets, vitamin or mineral supplementation, medication, menstrual cycle, consumption of alcohol, known diseases, and past injuries. Height and weight of each subject was measured with indoor clothing and no shoes.
The bone characteristics were measured with pQCT (Norland/Stratec XCT 3000; Stratec Medizintechnik GmbH, Pforzheim, Germany) and DXA (XR-26; Norland, Fort Atkinson, WI, USA) from the humeral shaft and distal radius according to our standard procedures.(13) The pQCT slice of humeral shaft was taken at 50% of the estimated humeral length (0.186 × subject height), proximal to the proximal endplate of the radius, and the distal radius slice at 4% of the estimated radial length (0.146 × subject height) proximal to the distal endplate of ulna.(13)
The pQCT outcome variables were bone mineral content (BMC, mg); cross-sectional areas of the total bone (TotA, mm2), cortical bone (CoA, mm2), and marrow cavity (CavA mm2); cortical wall thickness (CWT, mm); cortical and trabecular density (CoD, TrD, mg/mm3); and two bone strength indices (BSIt, mm3 and BSIc, g2/mm).(4) BSIt denotes density-weighted polar section modulus and reflects torsional and bending rigidity of the long bone shaft. BSIc denotes the product of a square of the total density (the apparent density of the structural materia) and the total cross-sectional area (the load bearing area) and reflects the strength of the bony structure against compression.(11) In our institute, in vivo precision of these pQCT measurements has been shown to range from 0.5% (CoD) to 5.6% (BSIt) for the humeral shaft and from 2.2% (TrD) to 4.7% (CoA) for the distal radius.(13)
aBMD measurements by DXA were performed according to our standard procedure. In vivo precision has been shown to be 0.8% for the humeral shaft and 1% for the distal radius.(23)
Relative side-to-side differences (%) were obtained using the log-transformed variables and then by the antilog-transformation of the parameter estimates. After that, the results were given in percentages. Side-to-side differences within and across the three study groups (young starters, old starters, and controls) were analyzed using the analysis of covariance (the study groups as the between-subject factor; the log-transformed value of the dominant arm as the dependent variable; and the log-transformed value of nondominant arm as the covariate). The post hoc comparisons of the group differences, with the 95% confidence intervals, were done using the Sidak adjustment. In all tests, an α level less than 5% (p < 0.05) was considered statistically significant.
Table 2 shows the absolute values of the pQCT and DXA measurements at both measured sites in three study groups. The relative side-to-side differences of the key variables (BMC, CoA, CoD, TrD, TotA, BSIt, BSIc, and aBMD) are presented in the Figures 1 and 2. The between-group comparison of the side-to-side differences is given in Table 3. Figure 3 illustrates the observed differences in the size of the humeral midshaft by displaying the cylindrical cross-sectional model of the playing arm over the nonplaying arm in the three study groups.
Table Table 2.. pQCT and DXA Values of the Dominant and Nondominant Arm (Mean, SD)
Table Table 3.. Between-Group Comparison of the Side-to-Side Differences
pQCT-derived side-to-side differences at the humeral shaft
In controls, the mean differences between dominant and nondominant arms were generally small but reached statistical significance in all measured parameters except BMC. The significant side-to-side differences, in favor of the dominant arm, were found in the cortical area (3%), total area (3%), cortical wall thickness (6%), area of marrow cavity (4%), cortical density (−1%), and torsional bone strength index (3%).
In the young starters, the average 19% greater BMC in the playing arm versus nonplaying arm was because of the 20% greater cortical area and 15% greater cortical wall thickness. These differences, together with the 13% increase in the total cross-sectional area, led to the 26% greater torsional bone strength index in the playing arm humeral shaft compared with that at the nonplaying side (Fig. 3). The post hoc analysis of the between-group comparison showed that these side-to-side differences were statistically significant (i.e., 8–22% higher in the young starters compared with controls and 8–14% higher compared with the old starters; Table 3). The side-to-side difference in the area of the marrow cavity was not statistically significant in both player groups (young starters, 1%; old starters, −2%), and in this respect, there was no difference between the three study groups. The cortical density in the young starters' humeral shaft was slightly in favor of the nonplaying arm (−0.7%). The same was seen in controls and old starters (Table 3 and Fig. 1).
Among the old starters, the average 9% greater BMC in the playing arm was also because of the 9% greater cortical area, leading to 11% higher torsional bone strength index in the playing arm (Fig. 3). Compared with the controls, these side-to-side differences were 6–7% greater, whereas the side-to-side differences in the total cross-sectional area (5%) and cortical wall thickness (6%) did not differ significantly from those in the control group (Table 3).
pQCT-derived side-to-side differences at the distal radius
In controls, there was no significant structural or density difference between the distal radii of the dominant and nondominant arms (Fig. 2).
Compared with the nonplaying arm, the young starters' playing arms had an average 12% higher BMC, caused by the 7% greater cortical area and 5% greater trabecular density (Fig. 2). However, compared with the controls, statistically significantly higher side-to-side differences were found in BMC (9%) and trabecular density (5%) only (Table 3). The young and old starters' side-to-side differences did not differ from each other in any measured variable, most likely because the interindividual variability was notable at this mainly trabecular bone site (Tables 2 and 3; Fig. 2). The total area of the young starters' playing arm distal radius was 11% larger than that in the unloaded site, but the cortical wall thickness did not differ between these sites (1%). The distal radius in the young starters' playing arm was estimated to be 13% stronger than that in the nonloaded arm.
Among the old starters, the side-to-side BMC difference was 6% in favor of the playing arm, resulting from a 7% greater cortical area and 7% greater trabecular density. Concerning the total area of the distal radius, the old starters' arms did not differ from each other (3%), whereas the cortical wall appeared to be thicker (6%) on the playing side. Compared with the controls, the old starters had 7% greater side-to-side difference in trabecular density (Table 3).
DXA-derived aBMD versus pQCT-derived bone strength indices (BSIt and BSIc)
At the humeral shaft, the DXA-derived aBMD clearly discriminated the young starters from the old starters (mean difference in the side-to-side aBMD difference was 7%) and controls (14%), and the old starters from the controls (6%; Table 3). However, these between-group differences were smaller than those seen in the BSIt (mean difference between young and old starters, 14%; between young starters and controls, 22%; between old starters and controls, 7%; Table 3; Fig. 1).
At the distal radius, the aBMD differentiated both player groups from controls (mean difference 5% between young starters and controls and 7% between old starters and controls), but there was no significant difference between young and old starters (Table 3). The side-to-side difference in the bone strength index (BSIc) did not differ statistically significantly between the groups, but the intergroup pattern seemed to be different from that in the aBMD (Table 3; Fig. 2).
The study showed that long-term tennis and squash playing was associated with an enlarged cortical area at the humeral shaft and that the loaded humerus seemed to have grown periosteally because the size of the marrow cavity was rather similar in the loaded and unloaded sides. This loading-induced geometric adaptation differed from the side-to-side difference of the controls in both groups of players despite the different starting age of the activity (either during the growing years or in adulthood), although the young starters' side-to-side differences were generally twice as large as than those in the old starters. At the distal radius, the young starters' playing arm's greater gains in the BMC and trabecular density differed from the side-to-side difference of the controls, while in the old starters, respective gain was noted in the trabecular density only.
When it comes to the mechanical competence of a whole bone, the material properties and the structure of bone (size, shape, composition, and internal architecture) are necessarily interrelated. One bone will carry the same compressive load as another bone, even if its material is half as strong, if it has twice the cross-sectional area.(9) In torsion and bending, the strength of bone is proportional to the moment of inertia. In other words, the resistance to fracture is better if material is situated well away from the central axis or the plane of bending.(9, 12) The aging-related expansion of the femoral neck and shaft(11, 24) and vertebral diameter(25) may thus reflect an effective compensatory mechanism against simultaneous loss of bone mineral. Also, greater bone size and mechanically more competent bones in the femoral and radial shafts have been shown to be related to higher physical activity and calcium intake in women, the association for physical activity being stronger with increasing age.(26)
Subperiosteal expansion of bone with age is thought to be a direct response to increased peak strains on that surface combined with the net loss of bone from the endocortical surface.(27) Beck et al.(24) gave a pragmatic example of the effects of this phenomenon on the section modulus (bone strength): expanding a 2-cm wide endocortical diameter of a 3-cm wide tubular bone by 10% (2 mm) will reduce areal BMD by about 16%. However, to maintain the section modulus unchanged, only a 0.83-mm simultaneous increase in the subperiosteal diameter would be required and a net loss of about 9% in the aBMD will remain without any real change in bending or torsional strength.(24) To illustrate the loading response at the humeral shafts of our players, we depicted representative tubular sections for loaded and unloaded bones (Fig. 3). The mean outer periosteal diameter of the cylindrical cross-section of the loaded humerus in the young starters was an average of 2 cm, including a 1-cm endocortical diameter. Compared with the unloaded site, this periosteal diameter of the loaded site was only expanded by 6.6% (i.e., no more than 1.2 mm), but it led to 19% greater BMC and 26% greater section modulus in the playing arm of the young starters. In the old starters, the playing-induced increase in this cross-sectional diameter was only 0.5 mm (2.6%), resulting however, in a 9% greater BMC and 11% greater section modulus in the loaded site.
The effect of extreme impact loading on the cortical area and cortical wall thickness has been recently demonstrated in triple jumpers; compared with the control group, over 50% greater values in these variables led to 31% higher bone section modulus at the jumpers' distal tibia.(28) Also, the apparent trabecular density at the distal and proximal tibiae was 18% and 41% higher in the jumpers than controls, respectively, whereas the cortical density of these sites did not differ between the groups.(28) Similarly, in female weightlifters, the cortical area of the distal radius was almost 40% greater compared with controls, whereas a 10% difference was found in the trabecular density.(29) One must note, however, that neither selection bias nor other confounding factors can be excluded from these cross-sectional comparisons of athletes and controls, and thus in this respect, the loaded-to-unloaded arm comparisons of racquet sports players provide a better study design. Our previous pQCT comparison between the playing and nonplaying arms of male tennis players (all of whom started their training in childhood), showed that the extra bone mineral at the loaded bone site was attributable to increased cortical area and not to apparent density of bone(14); this leaves the question of whether the loading-induced structural alterations in bone were related to the starting age of the activity.
In the current study, the side-to-side difference in the cortical area of the humeral shaft was more than twice as large in the young starters than old starters, whereas at the distal radius, the between-group difference was not so distinct. At this distal bone site, the young starters' greater cortical area at the playing arm led to the greater side-to-side difference in the total area, whereas in the old starters, no difference was observed in the total area. Instead, in the old starters, cortical wall thickness was greater at the loaded site, perhaps indicating a corticalization of the subcortical trabecular tissue. Animal studies support these observations by showing that a growing bone has better capacity to increase bone mass than a mature bone, and moreover, young and mature bones use different strategies for adaptation (i.e., a young bone seems to have a greater potential for periosteal expansion than an aging bone, whereas a mature bone can increase bone mass by increasing osteonal mean wall thickness).(30, 31)
At long bone ends, bone's structural response to loading may differ from that described above for a bone shaft due to both greater proportion of trabecular bone and the exposure to different types of stress. After cessation of growth, the joint and epiphyseal region are unlikely to enlarge their size, while increase in trabecular density, change in cortical thickness, and redistribution of bone mineral may remain as a mechanism for the structure to become stronger.(11) Our study supports these notions by showing a tendency of greater cortical wall thickness and apparent trabecular density in the old starters than young starters and controls, although the mean player-group differences, in favor of the old starters, were not statistically significant (2% [95% CI, −3%-6%] in TrD side-to-side difference and 4% [95% CI, −4%-12%] in CWT side-to-side difference).
Because of the above-mentioned difference in the type of stress at different regions of the skeleton, we used two different estimates of bone strength at the humeral shaft and distal radius. At a long bone shaft, a wide, hollow, cylindrical bone structure is optimal to resisting bending and torsional forces, and thus, the density-weighted polar section modulus (BSIt) was considered the most reasonable measure of bone strength for this site. In contrast, at the distal radius, a relatively dense trabecular bone structure with a large cross-sectional area would be optimal to cope with compressive stresses that are prevalent close to joints, and thus, the compressional bone strength index (BSIc) was considered the most reasonable measure of bone strength for this site.
The contribution of altered bone geometry to bone strength and fracture risk may not be appreciated by DXA-based assessment because DXA fails to distinguish geometry, apparent density, and mass distribution between cortical and trabecular bone compartments.(5, 31) This may lead to substantial underestimation of the effects of mechanical loading on bone strength.(32) The study of Adami et al.(33) showed, in fact, that after 6 months of site-specific moderate physical exercise, the total area of cortical bone at the ultradistal radius increased 3%, apparently caused by corticalization of the subcortical trabecular tissue, whereas hardly any effect was seen in the bone mass. Thus, reshaping of the bone cross-section by affecting both the cross-sectional area and density of the cortical component may lead to the increased bone strength. Correspondingly, in an experimental study by Järvinen et al.,(32) the cortical wall thickness and breaking load of the femoral shaft were significantly increased in the exercised versus sedentary rats, whereas no intergroup differences were observed in the DXA-derived BMC.
According to the present study, the DXA-derived aBMD at the humeral shaft clearly discriminated the young starters from the old starters (mean difference in side-to-side aBMD difference was 7%) and controls (14%), and the old starters from controls (6%). However, these between-group differences were smaller than those in the BSIt (mean difference between young and old starters, 14%; between young starters and controls, 22%; between old starters and controls, 7%). This underestimation of the intergroup difference can be explained by the planar nature of the DXA measurement, especially because of its incapability to detect bone's size and shape, and thus the greater cortical and total areas in the young starters.(5, 31)
Finally, it was interesting that the side-to-side difference in players' apparent cortical density was in favor of the unloaded site, and even if the difference was small (0.7–1%), it was systematic and statistically significant. The same phenomenon has also been seen with male tennis players.(14, 34) The reason for this is unknown, but the possibility of the greater bone turnover at the loaded site and thus a greater amount of intracortical resorption cavities, cannot be totally ruled out.
In summary, in both of the female player groups, the structural adaptation of the humeral shaft to long-term loading seemed to be achieved through periosteal enlargement of the bone cortex, although this adaptation was clearly better in the young starters. Exercise-induced cortical enlargement was not so clear at the distal radius (a trabecular bone site), and the study suggested that at long bone ends, the apparent trabecular density may alter in response to long-term loading. Conventional DXA-based aBMD measurement detected the intergroup differences in the exercise-induced bone gains, although, because it measured two dimensions of bone only, it seemed to underestimate the effect of exercise on the apparent bone strength, especially if the playing had been started during the growing years.
The authors thank Virpi Koskue and Anne Rauhio for expert bone measurements and Taru Helenius for help in the organizing the measurements. Financial support was provided by the Ministry of Education, Finland; Urheiluopistosäätiö (the Research Foundation of the Institute of Sports) Helsinki, Finland; and The Medical Research Fund of Tampere University Hospital, Tampere, Finland.