Does Exercise Reduce the
Burden of Fractures?
- a review
Department of Orthopaedics, Malmö University Hospital, SE –205 02, Malmo, Sweden
Correspondence: Dr Magnus Karlsson, Department of Orthopaedics, Malmo University Hospital, SE – 205 02 Malmo, Sweden .
Tel + 46 40 333843; Fax + 46 40 336200; E-mail: email@example.com
The null hypothesis that exercise has no effect on fracture rates in old age cannot be rejected on the basis of any published, randomized, prospective data. The notion that exercise reduces the number of fractures relies on prospective and retrospective, observational cohort studies and case-control studies, all hypothesis-generating, not hypothesis-testing. Consistently replicated sampling bias may produce the same observation when evaluating other than randomized prospective studies. Better health, better muscle function, higher muscle mass, better co-ordination may lead to exercise. The causal relationship could be between better health and exercise and better health and fewer fractures, not exercise and fewer fractures. The hypothesis should be tested in prospective, randomized studies evaluating hip fractures, spine fractures and other fragility fractures separately. Blinded studies evaluating the effect of exercise can obviously not be executed, but open trials can and should be undertaken to increase the level of evidence within the evidence-based system.
There are firm data supporting the notion that exercise during growth builds a stronger skeleton. Exercise during growth seems to result in high peak BMD and high muscle strength. However, the Achilles heel of exercise is it´s cessation. Are the skeletal and muscular benefits achieved during growth retained after the cessation of exercise and can any residual benefits be found in old age, the period when fragility fractures arise exponentially? Does exercise during adulthood produce any biologically important reduction in surrogate end- points for fractures other than BMD, as BMD can only marginally be influenced by exercise after completion of growth?
Recommendations for exercise should be based on evidence, not on opinion. Could continued recreational exercise maintain some of the benefits in BMD and muscle function achieved in youth? What level of recreational exercise should be performed to retain these benefits, if not fully, then at least to some extent? Dose-response relationships need to be quantified. Furthermore, the effect of exercise on independent, surrogate end- points for fractures, such as bone size, shape, architecture, muscle function, fall frequency and frequency of injurious falls during defined periods in the life cycle must be determined. Absence of evidence is not evidence of absence of effect, but if we recommend exercise then should this be to children, adults, elderly, men and women with fractures, all individuals? What type of exercise? For how long? Lifelong? If exercise could be implemented for most individuals within the society, would this reduce the number of fractures? Would the increased costs associated with the efforts to increase the activity level be lower than the reduced costs associated with any reduction in fractures? Our inability to answer these questions must be acknowledged before recommendations are made at the community level.
Does exercise reduce the number of fractures?
Half of all women and one third of all men will during their lifetime sustain a fragility fracture (Cooper et al., 1992). Increased morbidity, mortality and costs associated with the increased fracture incidence makes it imperative to implement prevention strategies in the community (Cooper et al., 1993; Poor et al., 1995). Hip and vertebral fractures in women are the fractures most commonly discussed, but also other fragility fractures create enormous problems (Ray et al., 1997). In addition, as fragility fractures in men increases, we must in the future also discuss the fracture problem also in this cohorts (Center et al., 1999; Kannus et al., 1996; Seeman, 1995).
In recent years, data have become available indicating that drugs reduce the fracture risk by about half in elderly women with bone mineral density (BMD) 2.5 SD below BMD in young healthy women, the definition of osteoporosis advocated by the World Health Organisation (SBU95, 1995; WHO, 1994). As the evidence-based decision for drug treatment is mainly based on trials including elderly, osteoporotic women with or without fractures, it is unclear whether also women with a more modest deficit in BMD benefit from drug treatment. Drug treatment probably also reduces the fracture rate in men with low BMD, but treatment strategies in men are less well defined (Orwoll et al., 2000; WHO, 1994).
General screening for detection of low BMD is not considered to be cost- beneficial, as a modest deficit in BMD implies a low absolute risk of sustaining a fracture (SBU95, 1995). Drug treatment in these groups would imply a large number needed to treat to save 1 fracture event, an approach that are not regarded as being evidence-based. Instead, when the aim of the health services is to reduce the fracture rate in the community, intervention programs are needed that are effective in preventing fractures, widely accessible, inexpensive with no adverse side effects. Exercise could have these benefits, but the question arises – does evidence-based information imply that exercise reduces the number of fractures? The final and only acceptable endpoint for evaluating the effect of exercise, are fractures, not surrogate end points as BMD, balance, muscle strength or fall frequency. However, a low absolute incidence of falls with an even lower incidence of fractures among the fallers creates a formidable challenge when randomized exercise intervention studies are planned with fracture as end-point. When designing a study with hip fracture as end-point, a 5-year study with = 0.05 and = 0.20, a control-group with a hip fracture incidence among 75-year-old women of 3-6% over a 5 year period and with risk reduction of 25 % with exercise, sample sizes need to be close to 7000 individuals to achieve the statistical power to detect a fracture reducing effect of exercise. Moreover, increasing the groups by 25% of due to drop-outs and non-responders is also recommended. Thus, prospective, randomized controlled studies to evaluate the effect of exercise on hip fracture rate are difficult and costly to perform (Gregg et al., 1998), and no such studies exist today. A prospective study evaluating if exercise during growth and adolescence protect against fragility fractures in old ages would due to compliance problem and drop-outs be virtually impossible to execute. With this background we have to use a lower level of evidence within the evidence-based hierarchy. The purpose of my review is to evaluate whether previous or current exercise affects fracture rate and surrogate end-points for fracture. Finally, it must be emphasised that exercise may confer a variety of health-related effects but in this survey I only discuss the effects on fracture rate and the muscular-skeletal system.
How strong are data, suggesting that exercise reduces the risk of sustaining a fracture?
There is no hypothesis proven evidence (randomized, prospective, controlled trial) that exercise reduces the fracture risk. No double-blinded trials can be done as there is no possibility to keep the investigator or participant blinded to exercise. Additionally, there has never been an unblinded, randomized prospective trial, an unrandomized, prospective trial or an uncontrolled trial showing that exercise reduces the fracture risk, mainly due to the large cohorts needed (Gregg, et al., 1998). However, lack of data from randomised trails is not proof of lack of efficacy. Going down in the evidence based hierarchy to non-interventionist, observational, case-control studies and prospective and retrospective cohort studies, there are data which support the notion that exercise reduces the fracture risk (Paganini-Hill et al., 1991). As these types of studies are the highest available evidence, we must lean on these data, not forgetting that causality can never be proven in observational or case/control studies. Even meta-analyses can not exclude the risk of sampling bias, as individuals with higher muscular capacity and function usually perform better in sports with probably a higher likelihood to chose a physically active lifestyle. The genetically inherited larger muscle mass and stronger bone may confer the lower fracture risk, not the high activity level.
Does exercise reduce the risk of sustaining a hip fracture?
In the following sections, odds ratios (all significant unless otherwise stated) for brevity are presented without confidence intervals, Most reports consistently suggest that individuals with a history of a low activity level at present or in the past have a higher incidence of hip fractures than individuals with a higher activity level (Gregg, et al., 1998; Wickham et al., 1989). Current activity such as daily standing, climbing stairs and walking are associated with a lower risk of sustaining a hip fracture (Cooper et al., 1988; Coupland et al., 1993). The Study of Osteoporotic Fracture (SOF), a longitudinal study following 9704 women aged 65 and over and for 4 years, revealed a 30% reduction in hip fracture risk associated with walking (Cummings et al., 1995). The same cohort followed for a mean of 8 years suggested that the hip fracture incidence was reduced by 42% among the women within the highest quintile of current activity compared to the least active quintile (Gregg, et al., 1998). There was a dose relationship in the activity, with 2 hours or more/ day of exercise reducing the hip fracture risk by 53% compared to less than 2 hours activity/day who reduced the incidence by 25 % compared to sedentary individuals. Also, sitting >9 hours/day increased the hip fracture risk by 43 % compared to individuals who sat <6 hours/day (Gregg, et al., 1998). The Leisure World Study (Paganini-Hill, et al., 1991), a prospective cohort study following 8600 postmenopausal women for 7 years, reported that exercise more than 1 hour/day reduced the hip fracture risk by 38 % compared to an activity level of less than ½ an hour/day. One study following 3595 non-institutionalized men and women over the age of 40 years in a population-based, longitudinal study for 10 years (NHANES I) suggested that no or a minimal activity level during recreation was associated with 90 % higher hip fracture risk compared to recreational exercisers (Farmer et al., 1989). These observations are supported in at least 6 other prospective cohort studies (Cummings, et al., 1995; Farmer, et al., 1989; Gregg, et al., 1998; Joakimsen et al., 1998; Meyer et al., 1993; Paganini-Hill, et al., 1991) and several case-control studies (Cooper, et al., 1988; Coupland, et al., 1993; Johnell et al., 1995). Although non-randomized, existing data consistently indicate that exercise during growth and adulthood is associated with a reduced hip fracture risk, selection bias cannot be excluded to explain the results. Finding a dose-response relationship in several published studies, with the risk reduction varying between 86 % (Coupland, et al., 1993) to 30 % (Paganini-Hill, et al., 1991) when comparing the most active with the least active individuals, strengthen the notion that moderate activity reduces the hip fracture risk in women (Gregg et al., 2000).
Data who support that exercise reduce the fracture risk in men are much weaker, as small cohorts and short follow-up periods increase the risk of a type II error. However, studies with the power to evaluate the exercise-induced, hip-fracture reducing effect support data presented in women. A longitudinal, cohort study of 3262 50-year-old Finnish men followed for 21 years reported that vigorous physical activity at baseline reduced the hip fracture risk by 58% (Kujala et al., 2000). The Leisure World Study included 5049 men aged 73 years, followed for 7 years presented an inverse relationship between exercise and hip fracture risk (Paganini-Hill, et al., 1991). Exercise more than 1 hour/day reduced the risk by 49% compared to exercise for less than ½ hour/day. The exercise-induced, hip-fracture reducing effect in men has so far been verified in at least 4 prospective, cohort studies with adequate sample sizes (Farmer, et al., 1989; Joakimsen, et al., 1998; Meyer, et al., 1993; Paganini-Hill, et al., 1991) but also in case-control studies (Cooper, et al., 1988; Gregg, et al., 2000; Grisso et al., 1991).
Does exercise reduce the risk of sustaining vertebral or other fragility fractures?
The SOF study reported that moderate to vigorous activity (> 2 hours/ day) reduced the vertebral fracture risk by 33 % relative to no activity (Gregg, et al., 1998). The European Vertebral Osteoporosis Study (EVOS) (Silman et al., 1997), including 6646 women aged 50 - 79 years, of whom 884 had a vertebral deformity, reported that current walking or cycling for more than 30 minutes each day resulted in a 20 % reduced risk to sustain a vertebral deformity compared to non-active women. In contrast, there are also authors who suggest that a longer duration of exposure to the risk of falling during activity, may increase some types of fractures. The risk of forearm fractures was non-significantly higher in women with walking as their leisure time activity compared to sedentary women (Mallmin et al., 1994; O'Neill et al., 1996) and the SOF study reported the same tendency with a non-significant increase in the risk of sustaining forearm fracture related to exercise (Kelsey et al., 1992) and a 13% increase in the risk of sustaining wrist fracture (NS) in the most active individuals (Gregg, et al., 1998).
Data supporting the contention that exercise reduces the incidence of vertebral deformities in men are weak. The EVOS prospective study (Silman, et al., 1997) including 5922 men, of whom 809 had a vertebral deformity, reported a 10 % reduced vertebral fracture prevalence with activity (NS). Two case-control studies with adequate sample sizes reported a tendency that physical activity reduced vertebral deformities, albeit a non-significant reduction (Chan et al., 1996; Greendale et al., 1995). When including all types of fragility fractures, the Dubbo epidemiological cohort study (Nguyen et al., 1996), suggests that each standard deviation of increased leisure time activity reduced all types of osteoporotic fractures by 14%, also after adjustment for differences in bone mass.
Does past exercise reduce the incidence of fractures?
What is the situation concerning fracture risks with reduced activity level after a period of active lifestyle during growth and adolescence, the scenario for many middle aged and elderly individuals? There were more individuals among two hundred and eighty-four former male soccer players now over the age of 48 who had fractures during their active career (before age 35) than controls (23 versus 16 %; p < 0.05), while after retirement (after age 35 years), the number of former soccer players with fractures were not fewer than controls with fractures (20 versus 21 %, NS) (Karlsson et al., 2000). Furthermore, the number of former soccer players who had sustained low energy fragility fractures after the age of 50 was not lower than controls (2 versus 4 %, NS), in absolute numbers only half in former athletes, but the power to detect a significant difference was low (Figure 1). The data are supported in other studies reporting more individuals with fractures among 2622 former female college athletes now 20 - 80 years compared to 2776 controls (40 % versus 32 %; p < 0.001) with no different fracture risk after retirement (Wyshak et al., 1987). Also the Leisure World Study (Paganini-Hill, et al., 1991) supports the findings when reporting that individuals with an activity level of more than 1 hour/day had a reduced risk of hip fracture compared to those active for less than ½ hour/day, but this effect was lost with further reduced activity level.
In summary, reports consistently suggest that exercise reduces the risk of hip fractures in men and women. Presenting a dose-response effect of exercise in several cohorts support this. Data suggesting that exercise reduces also other types of fractures related to osteoporosis are weaker. Existing studies consistently suggest that exercise in youth does not protect against fractures after retirement. As exercise during adulthood is reported to at best increase BMD by a biologically non-significant magnitude, the question remains - what is the mechanism behind the eventually reduced fracture rate? Is the quality of the skeleton improved? Are balance or muscle strength improved? Is the incidence of falls or injuries from falls reduced?
Does exercise during growth increase the accrual of bone mass and bone size?
The skeletal effects of exercise may differ in young and old individuals. The mechanical threshold for old rats was higher than in young rats but that, once activated, their cells had the same capacity as those of younger rats to enhance bone formation (Turner et al., 1995). The relative bone formation rate in the elderly rats was 16- fold less, and the relative bone forming surface 5-fold less compared to younger rats under similar loads (Turner et al., 1994; Turner, et al., 1995). Similar results have been presented by other trails, showing a dramatic reduction in responsiveness of the ulnae of old turkeys to applied mechanical loads compared to young turkeys (Rubin et al., 1992). Even if data in animals not uncritically can be transformed into humans, it seems as the skeletal response to exercise must be viewed separately in young and old individuals.
Data suggesting that exercise during growth increases mineralization and/or bone size are strong. Studies of young tennis and squash players have increased our understanding of the exercise induced skeletal effects by comparing the dominant and non-dominant arm. This approach eliminates the risk of selection bias among the athletes. Tennis players were early reported to have bigger bone, 10 – 35% greater cortical thickness and higher bone mass in the playing than in the non-playing arm (Huddleston et al., 1980; Jones et al., 1977). This observation was later confirmed in several independent reports that bone mass was up to four times higher in the playing versus the non-playing arm in female players who began their tennis training five years before menarche compared to those starting 15 years after menarche (Haapasalo et al., 1996; Kannus et al., 1994) (Figure 2). Including competitive athletes who began training early also suggests that exercise during growth and adolescence can substantially increase BMD (Bass et al., 1998; Dyson et al., 1997; Fuchs et al., 2001; Karlsson et al., 1993a; 1993b; 1996; 2000; 2001). Furthermore, cross sectional data consistently suggest that BMD is increased by 10 – 20 % with exercise only in weight-loaded skeletal regions. Pre-pubertal gymnasts had 10 – 30 % higher BMD compared to controls, with the greatest difference reported in the arms, a weight-bearing site in these athletes (Bass, et al., 1998) (Figure 3). Similarly, male weight lifters had 10 – 20 % higher BMD in the arms compared to controls (Karlsson, et al., 1993a; 1993b; 1996 ). Both male and female soccer players had no higher BMD in the arms, a region minimally loaded during soccer exercise, while BMD in the legs was 10 - 20 % higher compared to controls, a discrepancy of the same magnitude as in weight-lifters (Figure 3) (Duppe et al., 1996; Karlsson, et al., 2000). In fact, data even suggest a lower BMD to occur in unloaded skeletal regions in athletes compared to controls, suggesting that a redistribution of bone occurs from unloaded to weight-loaded skeletal regions during high activity with a reverse distribution with reduced activity level (Figure 3) (Karlsson, et al., 1996; Magnusson et al., 2001a; 2001b; Ramnemark et al., 1999).
Currently, 7 controlled, intervention studies, some randomized some unrandomized, comprising pre- and peripubertal boys and girls, have been published (Bass, et al., 1998; Blimkie et al., 1996; Bradney et al., 1998; Fuchs, et al., 2001; McKay et al., 2000; Morris et al., 1997). One study included the exercise intervention within the school curriculum (McKay, et al., 2000), the rest as leisure time activity on a voluntary basis. The intervention studies were short-term, 7-10 months with increased exercise 3 * 20-30 minutes extra per week. During this period, BMD increased 1.3–5% more in the legs in the active than in the sedentary children, only 2 studies reported an increased bone mineralisation in the spine. When a similar exercise program was conducted in peripubertal children, the effect on the skeleton was smaller or non-significant.
Data from prospective and retrospective cohort studies support this view in reports that physically active children have higher BMD than sedentary controls (Bailey et al., 1999; Cooper et al., 1995; Slemenda et al., 1994). However, these observational studies may be confounded by selection bias; exercise during leisure time could be preferred by children with larger muscle mass accompanied by a larger bone size and higher BMD due to shared genetic regulation, not that exercise confers high BMD. Most prospective studies indicate only a 1–4% higher increase in BMD in active individuals whether cross sectional studies often report 10-20 % higher BMD in athletes compared to controls. This could be due to a cumulative long term effect in athletes while most prospective studies at maximum span 2 years. It is also unknown if this increase in BMD are followed by fracture reduction and if so at what magnitude. For example, Raloxifen treatment increase BMD by 3% but reduced the lumbar fracture risk by 38% (Sarkar et al., 2002).
Does exercise during adulthood increase BMD and bone strength?
Low or moderate impact exercise has little effect in increasing BMD during young adulthood. Most studies report that aerobic exercise at best stops bone loss or increases BMD by less than 3%, probably of minor importance for the fracture risk (Bouxsein and Marcus, 1994; Drinkwater, 1993; Forwood and Burr, 1993). The outcome of weight-training produces similar discouraging results, with most studies reporting a BMD increase of no more than 2% (Friedlander et al., 1995; Gleeson et al., 1990; Lohman et al., 1995; Rockwell et al., 1990; Snow-Harter et al., 1992).
Similar beneficial effects in magnitude has been confirmed in numerous randomized, prospective, short-term studies in premenopausal women (Bassey and Ramsdale, 1995; Heinonen et al., 1996). Prospective intervention studies in peri- and postmenopausal women vary from 6 to 24 months and evaluate activities such as walking, stepping up and down, running, jumping and strength training. The studies report in general a beneficial effect in spine BMD by less than 3% compared to sedentary controls with the adaptive changes at the femoral neck described as less (Bravo et al., 1996; Grove and Londeree, 1992; Hatori et al., 1993; Nelson et al., 1994; Revel et al., 1993). Several published review articles during the last decade, describing 10-20 different prospective, randomized or non-randomized exercise studies, report exercise-induced beneficial skeletal effects in three-quartes of the studies in peri- and postmenopausal women (Bailey and McCulloch, 1990; Berard et al., 1997; Gutin and Kasper, 1992; Wallace and Cumming, 2000). One review evaluated the effect of exercise in women between age 46-76 in 35 different randomized, prospective studies and found that 6-36 months of both impact and non-impact exercise prevented bone loss by 1-2% in lumbar spine in peri- and postmenopausal women and that impact exercise seemed to have a similar effect, also in magnitude, on femoral neck BMD (Ebrahim et al., 1997; Friedlander, et al., 1995; Preisinger et al., 1995; Prince et al., 1995).
The outcome of exercise intervention in the elderly are equally discouraging . Six to 24 months of exercise in 65-80 years old women was reported with unchanged bone loss in spite of increased exercise in one third of the studies and a BMD gain but at a maximum of 2% during the study period in some (Ebrahim, et al., 1997; Hartard et al., 1996; Kelley, 1998; Lau et al., 1992; McMurdo et al., 1997; Prince, et al., 1995; Pruitt et al., 1995). Also in this age group, most positive results found in the spine while femoral neck BMD was reported with a less beneficial outcome (Ebrahim, et al., 1997; Lau, et al., 1992). If exercise confer other effects of the skeleton as changes in bone size, skeletal geometry or matrix properties, maybe influencing bone strength is not known and no intervention studies in adult men have so far been published.
No randomized, prospective studies exist evaluating the skeletal effects of lifelong exercise. The Rancho Bernardo Study (Greendale, et al., 1995) reported both current and lifetime exercise to correlate with hip BMD. Differences in BMD between individuals within the highest and lowest categories of exercise were 5 and 8%, respectively. Brahm et al. (1998) found the same, when report high lifetime occupational and leisure time activity to be associated with high BMD in 61 women and 61 men aged 22-85 years.