In both men and women, bone mineral density (BMD) and bone mineral content (BMC) increase from early childhood and throughout adolescence, and ascend to a peak, usually in the third decade. This is followed by a slow decrease in BMC and BMD in men, and in women the effect is magnified after the menopause. Approximately 90% of bone mass is present at the end of skeletal maturation (18), and the peak bone mass is an important determinant of bone mass in later life. BMD can explain between 75 and 85% of the variation in bone strength (19).
In addition to its numerous other health benefits, weight-bearing physical activity is essential for the normal development and maintenance of a healthy skeleton, and sedentary individuals who become more active may increase bone mass slightly, or reduce the rate of bone loss (1). Mechanical loads cause bone tissue within the loaded region to deform, stimulating adaptation. It has been suggested that the specific response to any bone strain depends on a threshold for that bone, and there exists a minimum effective strain for adjusting bone mass and architecture (6). Bone strains exceeding the minimum strain for modeling—adjustment of skeletal architecture and bone mass to keep a bone’s strains below the microdamage threshold—will result in a net increase in bone, whereas those falling below the minimum strain for remodeling—the turnover of bone in basic multicellular units (BMUs)—will result in a net loss of bone.
Different types of exercise cause different effects on bone remodeling, and apart from accidental trauma, the largest strains on bone appear to be the result of large muscle groups working against several times body weight using inefficient levers to cause movement under the earth’s gravity (6). Removing gravity virtually eliminates strain on bone, and astronauts may lose up to 19% of their weight-bearing bone on extended missions (10). Under weightless conditions, exercise countermeasures are employed in an attempt to minimize the loss of muscle and bone. Although no activity-specific countermeasure adequately prevent musculoskeletal deficiencies, resistance exercise appears to be more important than endurance exercise in reducing bone loss (3). Frequency of strain appears relatively unimportant in stimulating bone adaptation. Animal studies suggest strain rate to be important in stimulating new bone formation (17) and increasing strain magnitude has been shown to produce a dose-dependent increase in bone mass (16). In humans the case remains unproven, although a variety of related evidence now exists. Impact activity such as running can induce strain as a result of the momentary loading of several times body weight. By contrast, activities such as road cycling and swimming are weight supported, involve minimal strain, and involve a prone orientation for considerable periods of time. Although self-selection of preferred genotypes could explain why athletes of specific sports have different bone structure (17), it is likely that specific training for certain sports plays a pivotal role in skeletal adaptation. Understanding such effects could inform exercise advice given to young athletes. The hypothesis of this study was that running would increase bone density, whereas cycling would have little or no effect. This was tested by comparing athletes of three groups with controls for differences in BMD and BMC.
Male Caucasian nonsmoking volunteers between the ages of 18 and 43 volunteered for the study. They comprised competitive athletes and controls and were divided into groups of runners (N = 12), cyclists (N = 14), those who competed in both running and cycling events (“both”;N = 13); and healthy, nonexercising controls (N = 23). All athletes were regularly competitive at club level in their sports, and international athletes were present in each group. The runners undertook no cycling training, and similarly the cyclists undertook no running or running-based training. Those of the “both” group were participants in multisport events of running and cycling (N = 2) or, in addition, swimming (N = 9) or kayaking (N = 3). Controls undertook no deliberate exercise or sport. Athletes or controls engaged in manual work, or those with a family history of osteoporosis were excluded. All subjects gave written, informed consent, and the study was approved by the Lothian Health ethics committee for healthy volunteers, Edinburgh, U.K.
All subjects were measured for height and weight, and questioned by an administered questionnaire as to their postchildhood physical activity, competitive history in their chosen sport, and weekly training. The results are summarized in Table 1. Subjects were either fasted or ate only lightly on the day of the test, and presented fully hydrated and voided. Scans of the whole body and of the lumbar spine (L1–L4) were performed on each of the subjects using a Hologic QDR 1000W scanner (Hologic Inc., Waltham, MA). Results were analyzed using software version 4.47 for the spine and enhanced mode version 5.55 for the whole body. This allowed a % fat comparison in addition to bone mineral measurements. Scans were analyzed for regional bone density and total body composition. Because two athletes (one cyclist and one runner) and one control were too tall to fit onto the scanning table without bending their knees slightly (a practice which would artificially enlarge leg BMD values), leg BMC/total body weight was calculated in addition to leg BMD, to provide a mass-dependent rather than a density-dependent measure.
The study had 80% power to detect an effect size of 1.0 assuming a one-tailed α of 0.05. An effect size of 1.0 was selected due to the World Health Organization’s designation of osteopenia (significantly reduced BMD) as 1 SD below peak bone mass. Unpaired t-tests were used to determine whether differences between groups were significant. Statistical significance was accepted at the P < 0.05 level. Regression analysis was performed to assess the contribution of various factors to BMD variance.
The physical characteristics of subjects are summarized in Table 1. Compared with nonexercising controls, the athletes (treated as one large group) showed no significant differences in age or height (P > 0.05) but had lower weight and BMI (P < 0.01) and % fat (P < 0.001). However, more differences were observed when the athletic groups were considered separately. Compared with controls, all athletic groups had lower % fat (P < 0.001). Runners and cyclists had a lower BMI (P < 0.01), and runners alone were lighter than controls (P < 0.01). Compared with controls, the “both” group had no difference in BMI (P > 0.05). When the athletic groups were compared with one another, there were no significant differences in age, height, weight, BMI, % fat, or hours of training (P > 0.05).
Bone data are summarized in Table 2 and illustrated in Figures 1–5. Compared with controls, runners had greater total BMD (P < 0.01), greater leg BMD (P < 0.01), and greater leg BMC corrected for total body weight (P < 0.001). Compared with controls, cyclists had less spine BMD (P < 0.05). Compared with controls, the “both” group had greater total BMD (P < 0.05), arm BMD (P < 0.05), and leg BMC corrected for weight (P < 0.01). Compared with runners, cyclists had less total BMD (P < 0.001), less spine BMD (P < 0.01), less leg BMD (P < 0.01), and less leg BMC corrected for weight (P < 0.001). Compared with runners, athletes of the “both” group had greater arm BMD (P < 0.01), but less leg BMC corrected for weight (P < 0.01). Compared with cyclists, athletes of the “both” group had greater total BMD (P < 0.01), spine BMD (P < 0.001), arm BMD (P < 0.001), and leg BMC corrected for weight (P < 0.01).
Regression analysis was performed to evaluate the extent to which the variance in total and spine BMD was explained by weight alone, or weight together with running status (RS) or cycling status (CS), each of which was assigned a 1 for presence and 0 for absence. When athletes and controls were combined (N = 62), weight alone explained 18% of total BMD, whereas adding the factors explained 52%. Weight alone explained 12% of the variation in spine BMD, increasing to 38% when the other factors were included (P < 0.001). When athletes were treated as one separate group (N = 39), weight alone explained 21% of the variation in total BMD, rising to 60% with the factors included. Weight explained 14% of the variation in spine BMD, rising to 50% with the other factors included (P < 0.001).
This study shows that each of the athletic groups displayed a significant difference from each other and from controls with respect to BMD or weight corrected BMC. As there were no age or anthropometric differences between any of the athletic groups (P > 0.05), it is likely that these differences can be attributed to differences in exercise and training, which has placed different loading on the skeleton.
The mechanism of bone remodeling in adult life responds to commonly experienced strains and has been referred to as the “mechanostat,” similar to the principle of a thermostat (5). The sequence of remodeling occurs in a period of time between 100 d and 1 yr via negative feedback loops involving systemic hormones. Thus, although the general configuration of the skeleton is genetically determined, the internal structure is in the process of dynamic change according to external stimuli. In this respect, running and cycling appear to cause different adjustments to the “mechanostat.” Experimental data show dynamic weight-bearing activity using high strains appears to be most effective in promoting skeletal development in animals (13,15) and that in humans, power athletes have superior bone density to endurance athletes, although the bone response to mechanical loading is site specific (4). Heinonen et al. (7) reported gains in BMD of 1.4–3.7% in a period of 18 months of “jump training,” which induces impacts approaching 6 times body weight. Their study was succeeded by a follow-up that divided the original training group into training and control subgroups. Despite the mean age of 40 yr, BMD continued to rise over the subsequent 26 months in the training subgroup. Marcus (12) suggests jogging can produce forces of 3–4 times body weight and suggests that the extra intensity of activities that stress the skeleton may also be sufficiently rigorous to compromise safety.
Studies investigating the bone density of competitive cyclists are rare. One study of female athletes of different endurance sports, weight lifters, and controls (8) showed cyclists had no difference in BMD from controls at any site, whereas orienteers had higher BMD at distal femur and proximal tibia, suggesting a site-specific osteogenic influence of running.
Rico et al. (14) investigated 22 young male cyclists who trained 10 h·wk−1 for 2 yr. Using a Norland XR 26 bone densitometer (Norland Medical Systems Inc., Fort Atkinson, WI), they found no difference in total body bone mineral (TBBM) between the cyclists and age-matched controls but found the cyclists had less leg bone mineral than controls. This difference was eliminated when results were normalized for body weight. Although it has been suggested that more than 90% of adult bone mineral is present at the end of skeletal maturation (18), the mean height of the cyclists in Rico et al.’s study was only 171 cm, 4 cm less than the controls, suggesting further growth of the cyclists was probable. This could also explain why their leg BMC data were lower in cyclists than controls. In the present study, the height of the cyclists was not different from that of controls (P > 0.05), and both cyclists and controls had virtually identical proportions of total weight as leg mass, and leg BMC divided by total weight or leg weight (P > 0.05).
Warner and Dalsky (21) investigated 30 elite competitive male cyclists with similar age and height to the cyclists of the present study. When compared with recreationally active male controls, there was no difference in BMD or weight-adjusted BMD at any regional site. The ethnicity and physical activity of the recreational athletes was not presented, nor was it stated that the cyclists undertook no impact activity. The significant difference in total weight and % fat would suggest that their cyclists trained harder than their recreational athletes. Despite their study being performed on a Lunar DPX-L (Lunar Corporation, Madison, WI), and the spine scan being L2–L4, rather than L1–L4 as in the present study, the difference in findings between their work and the present study cannot be accounted for by reported inter-scanner differences (20). The results of the present study, which suggest that cycling is associated with reduced bone mineral, are at odds with their findings.
Non weight-bearing activities such as swimming and cycling have not been promoted for increasing BMD (1), and the lack of weight-bearing exercise is considered harmful to the skeleton (2). Although this could explain why cyclists have no different bone density from controls, why cycling should exert a negative influence on BMD independent of weight is something of a paradox. Possible explanations for cyclists’ apparent low bone content include the failure of cycling to produce strains above the remodeling threshold and the fact that body weight is distributed horizontally in the axial skeleton. Given that the average weekly training time in this position is 11 h, this represents a substantial proportion of total waking hours during which the skeleton is exposed to minimal strain, whereas the spine is exposed to the equivalent of bed rest. Such a flat cycling position is likely to cause the arms to assume a proportion of upper body weight. However, because this loading is largely static in nature, the increase in strain on arm bones is probably minimal, and no commensurate increase in arm BMD relative to runners or controls was observed.
Individuals who engaged in manual work or other impact sports such as tennis, were excluded from the study in an attempt to focus the investigation on cycling or running. However, the lack of other strain-producing activity in the cyclists is equally crucial to this argument. Although the regional and total BMD disparity between cyclists and controls disappeared when corrected for weight (P > 0.05), the coefficient of cycling status remained negative in the regression. Thus, the fact that cyclists had lower BMI than controls could explain most, but not all, of their lower BMD. However, given that all subjects in the study were at or close to their theoretical peak bone mass, a low value may predispose increased fracture risk later in life if other risk factors for osteoporosis are present. The mean spine T score (number of standard deviations from a peak BMD reference range supplied by the manufacturer) of the cyclists was −1.16, which would be diagnosed as osteopenic (significantly reduced BMD). One cyclist aged 36, with no family history of osteoporosis, had a T score of −2.72 (classified by the World Health Organization as osteoporotic). Although endocrine data were not collected as part of this study, this subject, after further clinical assessment, was shown to have normal endocrine function. Without further evidence, it would be misleading to suggest that cycling per se causes the observed low bone density, but the occurrence of reduced bone density appears greater than expected among those who cycle to the exclusion of physical activity involving impact.
Excluding all other exercise except cycling and running would have enabled a more complete understanding of the interactive effect of both activities. The fact that athletes who both cycled and ran largely engaged in significant additional exercise (11 of 13 individuals), means the contribution of upper body exercise could be viewed as a confounding variable. Nevertheless, the greater arm bone density of “both” group athletes is highly suggestive of the influence of upper body exercise on overall skeletal adaptation. The heavier build of these athletes, coupled with the diversity of their training, can also explain that their weight-corrected leg BMC was less than that of runners. However, the fact that such upper body exercise is nonimpact suggests a different mechanism for adjusting bone architecture may operate that can be implemented at lower bone strains. Evidence of human bone strain in vivo is extremely difficult to collect, although limited evidence (9) suggests not only mechanical loading but the mechanism for strain sensing differs in different skeletal regions.
Seen comparatively, the data on all three athletic groups suggest cycling in the absence of other sports is associated with a reduction in bone, whereas running is associated with an increase in bone at load-bearing sites. The effect of running would appear to counteract the effect of cycling in those athletes who do both, and upper body exercise appears to increase arm and spine bone significantly.
What remains to be investigated is the extent to which cyclists can improve their skeletal mass by impact bearing activity. Recovery of lost bone is possible but is probably site specific (11). Future work could usefully concentrate on the dose-response of exercise regimens, together with more specific in vivo measurements. Nonetheless, it would appear that caution is advised on cycling to the exclusion of other physical activity, particularly in young competitive athletes who are ascending toward their peak bone mass.
The authors wish to thank R. Elton for statistical advice and services, and P. Tothill for advice and assistance.
1. American College of Sports Medicine. Position stand on osteoporosis and exercise. Med. Sci. Sports Exerc. 27:i–vii, 1995.
2. Bailey, D. A., and R. G. McCulloch. Bone tissue and physical activity. Can. J. Sport. Sci. 15:229–239, 1990.
3. Baldwin, K., T. White, S. Arnaud, et al. Musculoskeletal adaptations to weightlessness and development of effective countermeasures. Med. Sci. Sports Exerc. 28:1247–1253, 1996.
4. Bennell, K., S. Malcolm, K. Khan, et al. Bone mass and bone turnover in power athletes
, endurance athletes
and controls: a 12 month longitudinal study. Bone 20:477–484, 1997.
5. Frost, H. Bone mass and the mechanostat: a proposal. Anat. Rec. 219:1–9, 1987.
6. Frost, H. Why do marathon runners have less bone than weight lifters? A vital biomechanical view and explanation. Bone 20:183–189, 1997.
7. Heinonen, A., P. Kannus, H. Sievanen, et al. Randomized controlled trial of effect of high impact exercise on selected risk factors for osteoporotic fractures. Lancet 348:1343–1347, 1996.
8. Heinonen, A., P. Oja, P. Kannus, H. Sievanen, A. Manttari, and I. Vuori. bone mineral density of female athletes
in different sports. Bone Miner. 23:1–14, 1993.
9. Hillam, R. A., M. Jackson, A. E. Goodship, and T. M. Skerry. Regional differences in human bone strain in vivo (Abstract). J. Bone Miner. Res. 10:(Suppl.)S443, 1995.
10. Hughes-Fulford, M., and M. Lewis. Effects of microgravity on osteoblast growth activation. Exp. Cell Res. 224:103–109, 1996.
11. Leblanc, A., and V. Schneider. Can the adult skeleton recover lost bone? Exp. Gerontol. 26:189–201, 1991.
12. Marcus, R. Skeletal “impact” of exercise. Lancet 348:1326–1327, 1996.
13. O’Connor, J., L. Lanyon, and H. MacFie. The influence of strain rate on adaptive bone remodeling. J. Biomech. 15:767–781, 1982.
14. Rico, H., M. Revilla, F. Hernandez, F. Gomez-Castresana, and L. Villa. Bone mineral content and body composition in postpubertal cyclist boys. Bone 14:93–95, 1993.
15. Rubin, C., and L. Lanyon. Regulation of bone formation by applied dynamic loads. J. Bone Joint Surg. 66A:411–417, 1984.
16. Rubin, C., and L. Lanyon. Osteoregulatory nature of mechanical stimuli: function as a determinant for adaptive remodeling in bone. J. Orthoped. Res. 5:300–310, 1987.
17. Skerry, T. Mechanical loading and bone: what sort of exercise is beneficial to the skeleton? Bone 20:179–181, 1997.
18. Slemenda, C., T. Reister, S. Hui, J. Miller, J. Christian, and C. Johnston. Influences on skeletal mineralization in children and adolescents: evidence for varying effects of sexual maturation and physical activity. J. Pediatr. 125:201–207, 1994.
19. Smith, C. B., and D. A. Smith. Relations between age, mineral density and mechanical properties of human femoral compacta. Acta. Orthop. Scand.
20. Tothill, P., A. Avenell, and D. Reid. Precision and accuracy of measurements of whole-body bone mineral: comparisons between Hologic, Lunar and Norland dual-energy X-ray absorptiometers. Br. J. Radiol. 67:1210–1217, 1994.
21. Warner, S. E., and G. P. Dalsky. Bone mineral density of elite male cyclists (Abstract). Med. Sci. Sports Exerc. 29:S5, 1997.