Publication in this issue of JBMR of three papers(1–3) concerning the beneficial effects of exercise on the skeleton highlights the remarkable manner in which the importance of mechanical loading on skeletal integrity, long understood by bioengineers, has been recognized by the osteoporosis community in a relatively few years. More than indicating the present high level of interest in exercise, these papers illustrate the degree to which this field has matured. It has been 10 years since most studies in this area did little more than compare bone mineral density (BMD) of athletes to that of sedentary controls. Not until the early 1990s did intervention trials begin to appear in which allocation of participants was randomized and in which exercise protocols themselves were described in precise and quantifiable terms. Results of those studies confirmed the general optimism that an imposed exercise program can modestly increase BMD. During this same era, insights were drawn from animal and epidemiological studies that now permit the design of clinical trials that do not simply ask whether exercise increases bone mass, but actually probe specific hypotheses about the nature of the skeletal response. The current papers are excellent examples of this approach.
Exercise intervention trials frequently demonstrate increases in lumbar spine BMD of about 1.5%, but only a few have been able to show improvement at the proximal femur,(4,5) despite the fact that loading conditions appeared suitable for achieving such a response.(6,7) One plausible explanation for these disappointing results is that the incremental loads imposed by training are low compared with habitual loads experienced at the hip during the course of daily activities. During a relaxed walk, each step imposes a load on the axial skeleton of 1 body weight. Load magnitudes increase to 3–4 body weights from jogging and about 5 body weights from jumping hurdles. Since an average person ambulates from 4–8 h each day, adding a few thousand walking or jogging steps, typical of most exercise trials, constitutes only a slight change in the daily mechanical history and might fail to initiate an adaptive response.
Bassey and colleagues(1) report the effects of a jumping program on BMD in three groups of healthy women: premenopausal, postmenopausal estrogen deficient, and postmenopausal women on hormone replacement therapy. The exercise program consisted of 50 vertical jumps each day, with an average height of 8.5 cm and loads equivalent to 3–4 body weights per jump, as estimated by measurement of ground reaction forces. Five months after beginning the trial, premenopausal women showed an increase in femoral neck and trochanteric BMD of 2–3%. By contrast, postmenopausal women, regardless of hormone replacement therapy status, showed no change in BMD, even among those women who completed a full year of training. Thus, jumping exercise improved BMD in young but not in older women, and the failure of the latter to respond to impact loading appears independent of estrogen deficiency.
The results of this study pose two important questions: Should we consider proven the concept of deficient skeletal adaptation in older people? Do the BMD increases observed in the younger women confirm the view that high-impact activity is optimal for increasing BMD? The answer to both questions is no.
Although evidence for attenuation of skeletal responsiveness with age does exist, it is incorrect to conclude that older women fail to undergo skeletal adaptation at the hip. Using a different strategy to train older postmenopausal women, Kerr et al.(5) conducted a year-long trial of bicycle exercise in which one leg served as the control and the other was subjected to a progressive increase in resistance. Results showed a gain of ∼2% in the trained hip as opposed to no change on the control side. To understand the disparate results of these two protocols some appreciation of the complex nature of mechanical loading is required. With jumping, two types of loads are transmitted to bone: those due to impact absorption and those due to muscle-generated forces. A dismount from parallel bars clearly exposes a gymnast to enormous impact force, about 11 body weights, which may partially explain why the hip BMD of gymnasts significantly exceeds that of other athletes.(8) However, in the jumping protocol of Bassey et al.,(1) the measured ground reaction forces did not exceed 4 body weights, a load readily achieved by jogging, which has generally proven to be unsuccessful in increasing proximal femur BMD. Therefore, the positive response shown by Bassey's younger cohort is unlikely due to impact but to strains induced by muscle tension on bone.
Since the results of Kerr et al.(5) clearly illustrate that older women can improve hip BMD with muscular loading, it remains unknown why the older women in the Bassey study failed to respond. The most striking difference between these studies is that Kerr's subjects increased their training regimen over time, whereas Bassey's regimen is static, never varying from a jump from standing height. It is possible that the younger women in the Bassey study had sufficiently high muscle strength at entry that a constant stimulus sufficed to initiate a skeletal response, whereas relatively poor muscle strength in older women was inadequate to do so. Because the training was not progressive, skeletal loading in the older women likely remained subthreshold throughout the study. Thus, one cogent lesson from these publications is that to achieve a skeletal response in weaker individuals requires a progressive increase in loading over time.
Turning to another topic of current interest, it is considered likely that the optimal time to capitalize on the potential benefits of exercise is childhood, when an increased loading environment would not only promote bone acquisition but also stimulate changes in overall bone architecture. Recent support for this notion comes from studies in which Haapsalo et al.(9) showed that skeletal dimensions of tennis players who began training in childhood significantly exceed those of athletes who started to play tennis later in life. Since bone strength is directly related to bone size, increased bony dimensions would theoretically reduce the long-term fracture risk of those starting training at an early age.
To date, cross-sectional studies provide the bulk of the primary evidence favoring this concept. The current paper by Bradney et al.(2) represents a direct test of the hypothesis in two groups of prepubertal schoolboys, mean age 10.4 years. Two schools were randomly assigned to control or intervention status, and boys in the intervention cohort participated in an 8-month protocol of thrice weekly half-hour physical education lessons. The program consisted of diverse sports activities and enjoyed very high compliance. Except for the activities imposed by the protocol, both groups undertook the same total hours of physical exercise and this did not change over time.
The results of this trial are fascinating. As one might predict, both groups of growing boys increased BMD over time, but the increase in the intervention group was twice that of the controls. In addition, the exercise group had differences in bony dimensions over time that differed from those of the control group. Although these findings are simple to declare, any attempt to understand them beyond the most superficial level quickly becomes nontrivial. For example, the standard dual-energy X-ray absorptiometry BMD measurement is not a volumetric term but an area measurement (reported in g/cm2) obtained by dividing the measured bone mineral content by the bone area. As such, BMD is greatly influenced by bone size. For most clinical purposes, this is of modest significance, becoming relevant only for patients of extraordinary size, either small or large. However, this issue assumes great importance when assessing changes over time in children, in whom bones gain both size and mineral, since an increase in either bone mineral content (g) or areal BMD (g/cm2) may reflect an increase in bone size, volumetric BMD (g/cm3), or both. The only technique that provides a truly volumetric measurement is quantitative computed tomography. Despite attempts to deal with this issue by mathematical or allometric models,(10) the optimal approach to assess longitudinally the parameters of bone acquisition in children has not been firmly settled. In the study of Bradney et al.,(2) volumetric estimates required the use of several reasonable assumptions regarding the shape and symmetry of bone. Although entry of some children into the pubertal growth spurt might have rendered these assumptions inoperative, one is reassured that fastidious attention was given to pubertal status at entry and throughout the study, with no evidence observed for pubertal progression in any child.
In any case, the work of Bradney et al.(2) confirms that exercise can alter overall skeletal geometry. However, the nature of these alterations was wholly unanticipated and in apparent conflict with the changes predicted by the earlier cross-sectional reports in tennis players. Although the exercisers showed an increase in midfemoral cortical thickness, this was not due to an increase in overall femoral shaft diameter but to a decrease in the medullary diameter. In other words, new bone had been applied to the interior (endocortical), not the exterior (periosteal) surface! By contrast, the control group showed a significant increase in periosteal diameter. This resulted in a significantly increased femoral shaft cross-sectional moment of inertia (CSMI) and section modulus in the controls, but not in the exercisers. These two variables describe the distribution of a material around a bending axis and they are directly related to the strength of an object.
The results of this trial pose more questions than they answer. Absent an increase in CSMI or in section modulus, would an increased bone mass confer any important mechanical advantage? In the short term, the answer is probably not. However, since the primary site of cortical bone loss in later life is from endocortical surfaces, a smaller medullary diameter at the time of skeletal maturity may provide a buffer against this process. For that to be the case, the added endocortical bone apposition observed prepubertally would have to persist into adult life. Since medullary narrowing occurs as a normal feature of pubertal development,(11) it is not clear that differences observed between these groups of prepubertal boys will persevere once they begin pubertal growth.
A second question is whether this study negates the observations in tennis players associating increased forearm bone diameter with early training. Again, the answer is probably not. The pattern of skeletal loading with racket sports may be unique, the forces of ball impact and muscle tension creating predominantly bending loads in the absence of axial compressive loading, such as occurs during weightlifting. These forces may induce fundamentally different skeletal responses than those induced by the loads arising from the activities employed by Bradney et al.,(2) which were similar in direction to usual weight-bearing stresses, but increased in magnitude. This interpretation emphasizes the specificity of exercise training. If one wants to achieve an increase in a given periosteal diameter, one needs to apply a stimulus directly to that site.
Although the ultimate significance of these results remains uncertain, the authors are to be congratulated for conducting a study that has advanced the science of clinical trials in children. They have shown an effective method to assign children to intervention and control status. They have shown the importance and feasibility of accounting for pubertal status and for activity levels outside of the trial itself, and they have established the need to account for multiple aspects of skeletal geometry, rather than relying solely on BMD.
The third paper represents an attempt to employ contemporary markers of bone turnover to distinguish the effects of different categories of exercise on skeletal dynamics.(3) It was anticipated that differential responses of resorption and formation markers to training might provide better insight into the mechanisms of skeletal response than changes in BMD alone.
Twenty healthy young men were randomly assigned to one of two 8-week exercise programs. Both programs involved thrice-weekly training sessions, each session 1 h in duration. The first program (“aerobic”) consisted of moderately intense endurance running at 60–85% of predetermined maximal aerobic capacity. The second program (“anaerobic”) consisted of running at >90% of maximal capacity in a series of short bursts separated by recovery intervals as well as a once per week leg-strengthening weight-lifting program. Achieved blood lactate concentrations following each session in the first program (up to 2.9 mM) were considerably lower than following the second (up to 8 mM). In the aerobic training group, markers of both bone resorption and formation decreased below baseline values by 4 weeks; by 8 weeks the resorption markers were still suppressed, but formation markers had returned to baseline. In the anaerobic training group, no changes in any of the markers had been observed after 4 weeks, but evidence for increased turnover was observed by 8 weeks. The authors suggest that aerobic training reduces bone resorption, whereas anaerobic training accelerates bone turnover. They also considered that subsequent normalization of bone formation activity in the context of continued suppression of resorption provided evidence for a long-term favorable effect on bone. By contrast, the sustained increase in bone turnover with anaerobic training was viewed as potentially deleterious, perhaps responsible for alleged negative skeletal effects of overzealous exercise.
Although interesting and provocative, these results must be viewed with considerable reserve. The number of subjects was small, as was the number of samples collected for each (one at baseline, one at 4 weeks, and one at 8 weeks). The concentration of biochemical markers in blood or urine reflects multiple factors beyond the endogenous rate of bone turnover. Since the anaerobic group trained at significantly higher intensity than did the aerobic group, a number of relevant perturbations in the balance of fluid, energy, or acid–base status may have occurred. Altered renal clearance due to hydration state would seem to have particular importance, since some of these changes might persist for hours or days beyond the activity burst itself. The observed changes in turnover markers may therefore reflect a physiological artifact rather than a real change in bone turnover. A second issue of concern is that marker activity in blood and urine reflects whole body bone turnover, whereas the loads applied during both categories of exercise primarily affect the legs and axial skeleton. Thus, considerable additional work must be done to clarify these and other issues before the relationship of turnover markers to bone adaptive response will be resolved.
The exercise studies presented in this issue of the JBMR provide insights into the nature of skeletal response. However, we must consider them within the larger context of ultimately preventing fragility-related fractures. Improving bone mass certainly remains a laudable goal, particularly if there is a possibility to produce a durable increase in peak bone mass. However, work over the past decade has demonstrated that exercise-induced changes in bone mass potentially achieved by people of average capabilities (as opposed to elite athletes) are modest at best and that exercise cannot be viewed as a panacea for preventing bone fragility. For frail, older people already at high risk of fracture, prevention of falls is of paramount importance. Since muscle weakness is an important modifiable determinant of falls risk, a program aimed simply at improving leg strength might be an attractive intervention, even if changes in BMD did not accrue. Several groups have examined the effect of strength training in older men and women.(12–15) The degree to which muscle strength improves with training, even during the 9th decade, is extraordinary,(12) and it is encouraging to see that substantial increases in muscle strength, muscle fiber size, and even neuromuscular performance are achieved by exercise programs that are less rigorous than those generally thought necessary in younger people.(16)
Finally, we must remember that we remain a fundamentally sedentary society. Space does not permit enumerating the evidence that people in Western societies on average exercise less and perform less well on physical fitness tests than in the past. Thus, it is most unlikely that these people will widely embrace the notion that they must do one sort of exercise to benefit the heart, another to increase muscle strength, and yet another to fortify bones. If health educators can convince people to carry out one general-purpose exercise program to satisfy multiple goals, the public will have been well served.