Background
Osteoporosis is a major public health problem affecting an estimated 200 million women worldwide[
1]. Congruent with osteoporosis is an increased risk for osteoporosis-related fractures, especially in women during the postmenopausal years, generally considered to begin around 50 years of age[
2]. Comparatively, the lifetime risk of an osteoporosis-related fracture in women is equivalent to the risk of developing cardiovascular disease[
3]. The two most common sites for osteoporosis-related fractures are the hip and the spine, with an estimated worldwide prevalence of 1.1 million and 862,000, respectively, in women 50 years of age and older in the year 2000[
2]. In the United States, the total annual costs associated with osteoporosis-related fractures were more than $19 billion in 2005 with a predicted increase to $25.3 billion in 2025[
4]. The majority of the costs in 2005 were attributed to fractures of the hip (72%) followed by the spine (6%)[
4].
Prevention of osteoporosis has focused on maximizing bone mineral density (BMD) during childhood and adolescence and maintaining BMD during adulthood[
5,
6]. Preventive measures include adequate calcium and vitamin D intake as well as avoiding cigarette smoking and excessive alcohol intake[
5,
6]. In addition, ground reaction (for example, jogging) and joint reaction (for example, strength training) force exercise has been recommended across the lifespan[
5‐
8]. However, the results of previous randomized controlled exercise intervention trials have reached conflicting and underwhelming conclusions regarding the effects of ground reaction and/or joint reaction force exercise on BMD at the femoral neck (FN) and lumbar spine (LS) in postmenopausal women[
9‐
33]. For example, using the vote-counting approach, only 29% of the exercise versus control group differences in FN BMD have been reported as statistically significant and in the direction of benefit while even fewer (11%) have been reported at the LS[
9‐
33]. Based on these findings, one might reach the general conclusion that ground and joint reaction force exercise have little or no effect on FN and LS BMD. However, reliance on a vote-counting approach based on statistical significance can be extremely misleading since the absence of a statistically significant effect does not mean that an effect is absent[
34]. In contrast, meta-analysis allows one to go beyond statistical significance and focus on the magnitude of effect. It is a quantitative approach for combining the results of studies. The strengths of meta-analysis include: (1) increased power, (2) improved estimates of effect size (ES), and (3) the potential to resolve disagreements between studies[
35].
While a number of meta-analyses have been conducted on the effects of exercise on FN and LS BMD in adults[
36‐
54], fewer have focused, or partitioned data, according to randomized controlled trials in postmenopausal women[
37,
39,
48‐
51,
53]. One meta-analysis that included studies published up to December, 1995 found a statistically significant exercise minus control group benefit of 0.73% in LS BMD as a result of joint and/or ground reaction force exercise in postmenopausal women[
39]. Another meta-analysis that included studies published up to January, 1998 reported a statistically significant benefit in FN and LS BMD ranging from 0.9% to 1.6% as a result of impact and non-impact exercise among postmenopausal women[
53]. However, both meta-analyses were limited to studies published more than 14 years ago. Since that time, additional randomized controlled trials with inconsistent findings have been published[
10‐
12,
14‐
17,
19‐
22,
24‐
27,
30,
32,
33]. In addition, guidelines for the improved conduct of meta-analysis have been developed[
47].
A modality-specific, joint reaction force meta-analysis that included studies published up to December 2004 found a statistically significant benefit of 0.006 g/cm
2 in LS BMD and a non-significant benefit of 0.010 g/cm
2 in FN BMD as a result of high-intensity resistance exercise in postmenopausal women[
49]. Another modality-specific meta-analysis by the same research group which included studies published through December 2006 reported a non-statistically significant benefit in FN and LS BMD in postmenopausal women as a result of walking[
50]. These findings suggest that walking, a lower impact, ground reaction force exercise, may have little benefit on FN and LS BMD in postmenopausal women. The same research group also published another meta-analysis that included studies published to 2008[
51]. When limited to randomized controlled trials and a random-effects model, a statistically significant benefit of 0.004 g/cm
2 was found for FN BMD with no statistically significant benefit observed at the LS as a result of exercise in postmenopausal women[
51]. More recently, a Cochrane systematic review by Howe et al. reported a statistically significant exercise minus control group benefit of 0.85% in LS BMD but no significant change in FN BMD (−0.08%) as a result of joint and/or ground reaction force exercise in postmenopausal women[
37]. However, this systematic review did not appear to be limited to studies in which participants had been previously participating in exercise levels below that currently recommended for bone health[
8]. Consequently, the benefits of exercise could have been underestimated. Another meta-analysis reported a statistically significant benefit of 0.014 g/cm
2 and 0.012 g/cm
2, respectively, for both FN and LS BMD in females 60 years of age and older[
48]. However, similar to the work of Howe et al.[
37], participants did not appear to be limited to those who were participating in exercise levels below that currently recommended for bone health[
8]. In addition, all studies were coded by one person, thereby increasing the risk for coding errors[
47]. A potential reason for the discrepancy in findings for FN BMD between the Howe et al.[
37] and Marques et al.[
48] reviews may be accounted for by the fact that the latter meta-analysis limited studies to those in adults 60 years of age and older. This raises the possibility that older postmenopausal women may have more to gain from a regular exercise program. Finally, because the number of analyses aimed at trying to establish the association between selected covariates and changes in FN and LS BMD was limited for all of the previously described meta-analyses, potentially important covariates could have been missed.
A need exists for an updated and thorough meta-analysis on the effects of different ground and joint reaction force exercises, either alone or in combination, on FN and LS BMD in postmenopausal women not participating in exercise levels currently recommended for bone health[
8]. Therefore, the purpose of this study was to use the aggregate data meta-analytic approach to determine the effects of ground and/or joint reaction force exercise on BMD at the FN and LS in postmenopausal women not participating in exercise levels currently recommended for bone health[
8].
Discussion
The purpose of this study was to use the aggregate data meta-analytic approach to determine the effects of ground and/or joint reaction force exercise on BMD at the FN and LS in postmenopausal women participating in exercise levels below that currently recommended for bone health[
8]. The overall results suggest that ground and joint reaction force exercise may result in clinically important benefits in FN and LS BMD, with results more convincing for FN BMD. These findings are similar to those from three[
48,
51,
53] of four[
37,
48,
51,
53] previous meta-analyses for FN BMD and four[
37,
39,
48,
53] of five[
37,
39,
48,
51,
53] previous meta-analyses for LS BMD, all of which included both ground and joint reaction force exercises from randomized controlled trials in postmenopausal women. Further support for the overall findings of the current meta-analysis were strengthened by the robustness of results when data were collapsed so that only one
g represented each study as well as when examined for publication bias. When each study was deleted from the model once, results remained statistically significant for FN BMD across all deletions but were no longer statistically significant for LS BMD (
p = 0.12) when one study was deleted from the model[
11]. From a stability perspective, the statistical significance of findings has been consistent over a longer period of time for BMD at the FN (2000) versus LS (2009). Thus, the changes in BMD appear to be more convincing for FN versus LS BMD. This may have to do with the possibility that the exercise protocols employed were more specific to the FN versus LS.
While random-effects models that incorporate heterogeneity into the analysis were used, it is still important to point out that heterogeneity was observed for both FN and LS BMD. The existence of heterogeneity in meta-analysis is not only common[
79], but also important, as there is no need to combine studies exactly alike since their findings, within statistical error, would be the same[
80]. In addition, prediction intervals for estimating the expected results of a new trial included zero for both FN and LS BMD. However, these values should not be confused with confidence intervals since prediction intervals are based on a random mean effect while confidence intervals are not[
73]. Nevertheless, these prediction intervals may be beneficial for future researchers interested in conducting randomized controlled intervention trials addressing the effects of ground and/or joint reaction force exercise on FN and LS BMD in postmenopausal women.
While the magnitude of change in FN and LS BMD might be considered small at the FN and trivial at the LS, they appear to be clinically important. For example, based on previous prediction models[
81], the exercise-induced changes in BMD observed at the FN and LS in the current meta-analysis would reduce the 20-year relative risk of osteoporotic fracture at any site by approximately 11% and 10%, respectively. However, the observed benefits of exercise on FN (g = 0.29) and LS (g = 0.18) BMD in the current meta-analysis were smaller than those previously reported for pharmacologic interventions (alendronate, calcitonin, etidronate, hormone therapy, raloxifine, risedronate) at both the hip (range of
g = 0.64 to 5.74) and LS (range of
g = 0.90 to 8.90)[
82]. The exercise-induced benefits on FN and LS BMD also appear to be similar to or smaller than those observed for calcium and vitamin D supplementation (
g for calcium = 0.45 at the hip and 1.57 at the LS;
g for vitamin D = 0.47 at the hip and 0.20 at the LS)[
82]. However, the use of pharmacological and nutritional interventions should be considered with respect to several factors. These include: (1) the potential adverse effects of pharmacologic agents[
83], (2) that participants included in previous pharmacological and nutritional intervention studies had generally lower initial levels of BMD than participants included in the current exercise meta-analysis[
83], and (3) that exercise results in numerous other benefits not realized with pharmacologic and nutritional interventions[
84], for example, increases in balance and a subsequent reduction in falls[
85]. Given the former, the current recommendations of lifestyle changes such as exercise and adequate calcium and vitamin D intake prior to pharmacological intervention appear to be appropriate[
6].
The focus of the present meta-analysis has been on the use of the traditional alpha value for statistical significance (p ≤ 0.05) and 95% CI. However, it has been suggested that rather than focus on the term statistically significant and alpha value cutpoints, one should report the exact alpha value and use 90% CI to determine clinical relevance within the range of the 90% interval[
86]. Using the 90% CI approach, the interval no longer included zero (0) for changes in LS BMD (0.026 to 0.332) and ranged from 0.132 to 0.444 for changes in FN BMD.
No statistically significant between-group differences were found when mixed-effects ANOVA was conducted for changes in FN and LS BMD partitioned by a large number of categorical variables. However, while no statistically significant between-group differences were noted, changes in FN BMD were smaller for ground (g = 0.088) versus joint (g = 0.420) and combined joint and ground reaction force exercise (g = 0.398).
Several interesting associations were found when simple meta-regression was performed for changes in FN and LS BMD. For ease of reading, statistically significant findings (
p < 0.05) as well as trends for statistical significance (>0.05 but ≤ 0.10) are discussed collectively. For both FN and LS BMD, greater increases were associated with both greater intensity and compliance in the strength training (joint-reaction force) groups. These findings suggest that greater loads per repetition as well as greater adherence may provide greater benefit to FN and LS BMD. Greater improvements in both FN and LS BMD were also associated with increases in static balance. These associations may be especially important for reducing the risk of falling as well as subsequent fracture risk. Greater increases in both FN and LS BMD were also associated with decreases in BMI, body weight and percent body fat. In addition, increases in LS BMD were associated with increases in LBM. All of these associations may be reflective of greater exercise effort. The inverse association between increases in both FN and LS BMD with poorer compliance to aerobic and strength training protocols may be nothing more than the play of chance. Alternatively, studies with poorer compliance may have yielded greater benefits in FN and LS BMD because of the greater overall volume of training prescribed. For LS BMD, the positive association between increases in LS BMD and older age as well as a greater number of years postmenopausal may be the result of lower initial levels of BMD. However, we found no association between baseline LS BMD and changes in LS BMD. The negative associations between increases in LS BMD with shorter duration and total minutes of training per week for aerobic exercise studies may help to reinforce the belief that shorter duration activities such as jumping may be more beneficial to LS BMD than activities such as walking[
7]. One potential reason for this negative association may be the result of calcium loss from excessive sweating in longer duration and/or higher intensity activities[
87,
88]. This causes a decrease in serum calcium followed by an increase in serum parathyroid hormone, which then stimulates bone resorption[
87,
88]. While these findings are interesting, further research is needed before any firm conclusions can be drawn.
In addition to changes in FN and LS BMD, statistically significant improvements were found for several secondary outcomes. These included increases in BMD (total hip, trochanteric, intertrochanteric), aerobic fitness, dynamic and static balance, lean body mass and both upper and lower body strength. Statistically significant decreases in percent body fat were also found. These findings reinforce the many benefits that can be derived from exercise programs[
84]. The former notwithstanding, the results for secondary outcomes should be interpreted with caution since they were only included if FN and/or LS BMD data were reported. Consequently, secondary outcomes in meta-analysis may not comprise a representative sample.
A major interest of the investigative team was to examine the dose–response relationship between changes in FN and LS BMD and exercise load ratings in postmenopausal women. While we found no significant association between changes in FN and LS BMD and load ratings, these associations were based on general categorical estimates versus estimates specific to each activity[
59]. The decision to use categorical estimates was based on the inability to accurately calculate load ratings for those studies that involved multiple types of activities. In addition, the algorithm used requires further testing, improvement and validation[
59]. Future research should also focus on developing formulas for accurately calculating load ratings from data typically provided in randomized controlled intervention trials. Ideally, individual studies should collect and report force data in all exercise interventions. However, the accurate measurement of such may be challenging for some activities[
7]. Until additional dose–response research is conducted, it would appear plausible to suggest that postmenopausal women adhere to the exercise guidelines from the American College of Sports Medicine[
8]. These include weight-bearing endurance activities 3 to 5 times per week as well as resistance exercise 2 to 3 times per week[
8]. However, it will be particularly important for future dose–response studies to determine whether increased duration of aerobic exercise diminishes the potential skeletal benefits, as suggested by the current regression analyses.
The results of this meta-analysis should be viewed with respect to several potential limitations. First, because studies are not randomly assigned to covariates, they are considered to be observational in nature. Therefore, the results of moderator and regression analyses conducted in this or any other meta-analysis do not support causal inferences[
78]. Second, because a large number of statistical tests were conducted, some statistically significant results could have been nothing more than the play of chance. However, as suggested by Rothman[
89], no adjustment was made for multiple tests because of the concern about missing possibly important findings. Third, because of a lack of data, a common occurrence in meta-analysis, the research team was unable to examine several variables, thereby compromising the thoroughness of the study. With the former in mind, it is suggested that future randomized controlled trials addressing the effects of ground and/or joint reaction force exercise on FN and LS BMD in postmenopausal women include information regarding study design (allocation concealment, incomplete outcome data, verification that all outcomes planned to be assessed are reported), participant characteristics (adverse events, whether the participants had osteoporosis, cigarette smoking, alcohol consumption, change in exercise habits outside the intervention) and exercise intervention characteristics (intensity, how exercise was delivered). Fourth, future studies should provide more specific information regarding their exercise cutpoints for enrolling participants in their studies. The heterogeneity of reporting found in the current meta-analysis is not surprising. In a systematic review of the different definitions of sedentary for screening participants for entrance into physical activity intervention trials, Bennett et al.[
90], found that the definition of sedentary ranged from less than 20 to less than 150 minutes per week minutes of physical activity and that few studies reported the type and intensity of physical activity used to screen participants. While such varied definitions may make it difficult to generalize findings, the current meta-analysis, to the best of the authors’ knowledge, is the first one on exercise and BMD in women to limit the inclusion of studies to those in which participants were not currently meeting exercise recommendations for bone health[
8]. Fifth, given the potential advantage of high resolution peripheral quantitative computed tomography (HR-pQCT) for detecting microarchitectural changes in bone[
91], it would appear plausible to suggest that future exercise intervention studies should use this technology so as to better understand the exercise-induced changes that may occur in bone. Finally, consistent with recommendations from the 2008 Physical Activity Guidelines Report, there continues to be a need for large randomized controlled trials to determine whether fracture incidence is decreased as a result of ground and/or joint reaction force exercise[
7].
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
GAK was responsible for the conception and design, acquisition of data, analysis and interpretation of data, drafting the initial manuscript and revising it critically for important intellectual content. KSK was responsible for the conception and design, acquisition of data, and reviewing all drafts of the manuscript. WMK was responsible for the conception and design, interpretation of data and reviewing all drafts of the manuscript. All authors read and approved the final manuscript.