HomeLarge Type Edition
HOME ARCHIVE SEARCH TABLE OF CONTENTS
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Services
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
PubMed
Right arrow PubMed Citation
The Journals of Gerontology Series A: Biological Sciences and Medical Sciences 57:M599-M604 (2002)
© 2002 The Gerontological Society of America

Exercise and Lumbar Spine Bone Mineral Density in Postmenopausal Women

A Meta-Analysis of Individual Patient Data

George A. Kelleya, Kristi S. Kelleya and Zung Vu Tranb

a Clinical Investigation Program, MGH Institute of Health Professions, Boston, Massachusetts
b Department of Preventive Medicine and Biometrics, University of Colorado Health Sciences Center, Denver

George A. Kelley, FACSM, Associate Professor, Graduate Program in Clinical Investigation, Director, Meta-Analytic Research Group, MGH Institute of Health Professions, Charlestown Navy Yard, 36 First Avenue, Boston, MA 02129 E-mail: gakelley{at}bics.bwh.harvard.edu.


   Abstract
Top
 Abstract
Methods
 Results
 Discussion
 References
 
Background. Low bone mineral density (BMD) at the lumbar spine is a major public health problem among postmenopausal women. We conducted a meta-analysis of individual patient data (IPD) to examine the effects of exercise on lumbar spine BMD in postmenopausal women.

Methods. IPD were requested from a previously developed database of summary means from randomized and nonrandomized trials dealing with the effects of exercise on BMD. Two-way analysis of variance tests with pairwise comparisons (p <= .05) and 95% confidence intervals (CIs) were used to determine the statistical significance for changes in lumbar spine BMD.

Results. Across 13 trials that included 699 subjects (355 exercise, 344 control), a statistically significant interaction was found between test and group (F = 15.232, p = .000). Pairwise comparisons (Bonferroni t tests) revealed a statistically significant increase in final minus initial BMD for the exercise group ( ± SD = 0.005 ± 0.043 g/cm2, t = 2.46, p = .014, 95% CI = 0.001–0.009) and a statistically significant decrease in final minus initial BMD for the control group ( ± SD = -0.007 ± 0.045 g/cm2, t = -3.051, p = .002, 95% CI = -0.012–-0.002). Changes were equivalent to an approximate 2% benefit in lumbar spine BMD (exercise, +1%, control, -1%).

Conclusions. The results of this IPD meta-analysis suggest that exercise helps to improve and maintain lumbar spine BMD in postmenopausal women.

IT has been estimated that approximately 26.2 million postmenopausal women have either osteoporosis or osteopenia (1). As a result of having osteoporosis or osteopenia, a person is at an increased risk for fracture, particularly at the vertebrae, hip, and distal forearm (2). Of these three sites, fractures of the vertebrae, which represent approximately 56% of all fractures, are the most common, with an estimated 700,000 per year (2). The health-care costs associated with vertebral fractures were estimated to be approximately $746 million in 1995 and are expected to increase substantially in the future (3).

One of the potential interventions for increasing and/or maintaining vertebral bone mineral density (BMD) in postmenopausal women is exercise, a low-cost, nonpharmacologic intervention that is available to most individuals. We have recently conducted meta-analytic work in which we reported improvements in lumbar spine BMD because of exercise in postmenopausal women (4)(5). This work was based on the most commonly used approach for conducting meta-analytic work, that is, the abstraction of summary means from studies meeting specified inclusion criteria. However, the use of individual patient data (IPD) versus summary means from eligible studies represents the most comprehensive approach for conducting meta-analytic work, including the potential for increased statistical power as well as a more thorough examination of potential covariates (6)(7)(8). Given the health-care consequences of low BMD at the lumbar spine, the possible benefit of exercise for improving and/or maintaining lumbar spine BMD, and the potential for a meta-analysis of IPD to provide more thorough information regarding the effect of exercise on lumbar spine BMD, we sought to examine the effects of exercise on lumbar spine BMD in postmenopausal women by conducting a meta-analysis using IPD.


   Methods
Top
 Abstract
 Methods
Results
 Discussion
 References
 
Data Sources
From a previously developed meta-analytic database that included the summary means from 76 studies dealing with the effects of exercise on BMD, we sought to obtain IPD. Briefly, IPD were requested by sending a cover letter and data request sheet to authors via postal mail. For those who did not respond to our initial request, a follow-up letter was sent via postal mail approximately 5 weeks later.

Study Selection
From the database of 76 studies, we included studies that met the following criteria: (i) randomized and nonrandomized trials that included a comparative control (nonexercise) group, (ii) exercise lasting at least 16 weeks, (iii) postmenopausal women only, (iv) journal articles, dissertations, and masters theses published in the English-language literature, (v) studies published between January 1966 and December 1998, (vi) BMD (relative value of bone mineral per measured bone area) assessed at the lumbar spine, and (vii) ability to obtain IPD from authors. Despite the fact that methods to assess BMD (dual-photon absorptiometry [DPA], dual-energy x-ray absorptiometry [DEXA]) have only been widely available since the 1980s, we searched back to 1966 to ensure that there was no comparative technology that we might have missed. We did not include studies from non-English-language journals because of the potential for error in the translation and interpretation of findings. If more than one study included the same subjects, for example, a dissertation and refereed journal article, we retrieved and referenced both studies to extract the maximum amount of information but only included this as one data set.

Data Abstraction
All data were abstracted on a coding sheet that could hold up to 91 pieces of information from each study. All data were coded and verified for accuracy and consistency by George A. Kelley. Blinding of the coder to the identity and institutional affiliation of the authors as well as study results was not performed because it has been shown that these procedures have neither a statistically significant nor a clinically important effect on the results (9). The major categories coded included study, subject, BMD assessment, and training program characteristics as well as primary and secondary outcomes.

Statistical Analysis
Initial subject characteristics.-- Potential differences between initial subject characteristics for exercise and control groups were examined using independent t tests and 95% confidence intervals (CIs) for continuous variables and 2 x 2 chi-square tests for categorical variables.

Primary outcomes.-- Initial and final values for lumbar spine BMD between exercise and control groups were examined by using a two-way analysis of variance (ANOVA) test with repeated measures on one factor (test). Because this was an unbalanced design, a General Linear Model was used. Pairwise comparison tests (Bonferroni t tests) were used to identify the specific location of the observed interaction between test (final vs initial) and group (exercise vs control). To examine for outliers, ANOVAs were performed with each study deleted from the model once. Because of missing data, we were unable to include potential covariates in the ANOVA model. Consequently, we used Pearson-Product moment correlations to examine for potential associations between changes in BMD and age, height, body weight, years postmenopausal, cigarette smoking, alcohol consumption, calcium and vitamin D intake, compliance (percentage of exercise sessions attended), length of training (weeks), type of BMD assessment (DEXA, DPA), and study design (randomized vs nonrandomized controlled trial). We were unable to partition the data according to the different types of exercise because of the various interventions employed.

Because of the inability to retrieve IPD from all eligible studies, we also examined whether our results differed between studies according to the availability of IPD. To include all eligible studies in the analysis, we used the standardized difference effect size (ES) calculated from the summary data reported in the studies and corrected for small sample bias (10). In general, an ES of 0.20 is considered a small effect, 0.50 a moderate effect, and 0.80 a large effect (11). An ES of 0.80, for example, means that the exercise group differed from the control group by eight-tenths of a standard deviation in favor of the exercise group. We then compared ES differences between those studies in which IPD were provided versus those in which they were not using an ANOVA-like random effects model developed for meta-analytic research (10). This was accomplished by examining the between (Qb) and within (Qw) group differences for the ESs and their variances from each group.

Secondary outcomes.-- Secondary outcomes (body weight, calcium intake, and vitamin D intake) were analyzed using the same ANOVA procedures that were used to evaluate changes in lumbar spine BMD. We used independent t tests to analyze initial differences between exercise and control groups for these variables because more data were available for initial values versus final values and we wanted to capture as much data as possible in our analyses.

Descriptive Statistics and Alpha Level
Means and standard deviations ( ± SD) were used to describe continuous variables, whereas frequencies and percentages were used for categorical variables. The alpha level for statistical significance was set at p <= .05. Ninety-five percent CIs that did not cross 0.00 were also considered statistically significant.


   Results
Top
 Abstract
 Methods
 Results
Discussion
 References
 
Study Characteristics
Of the 32 studies that met our criteria for inclusion, we were able to retrieve IPD from 13 (41%) (12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26). Note that the number of references exceeds the number of studies because two were published in dissertation (17)(24) and two in journal format (18)(25). A description of the studies is shown in Table 1 . The 13 studies represented a total of 30 groups (17 exercise, 13 control) and 699 subjects (355 exercise, 344 control). Seven of the trials were randomized controlled trials, and the remaining six were nonrandomized controlled trials. The length of the studies ranged from 24 to 104 weeks ( ± SD = 56 ± 8 weeks). Thirteen of the exercise groups included some type of weight-bearing exercise, two appeared to perform nonweight-bearing exercise, and the remaining two participated in weight training. Compliance, defined as the percentage of exercise sessions attended, averaged 75 ± 17%. Seven of the studies assessed lumbar spine BMD using DEXA, whereas the remaining six used DPA.


View this table:
[in this window]
[in a new window]
  		Table 1. Characteristics of Bone Mineral Density Studies (gm/cm2) in Which IPD Were Provided for Postmenopausal Women at the Lumbar Spine

 
Initial Subject Characteristics
Initial subject characteristics for continuous and categorical variables can be found in Table 2 and Table 3 , respectively. For continuous variables, the number of years that the subjects were postmenopausal was significantly greater in the control versus exercise groups, whereas calcium intake was greater in the exercise versus control groups. No statistically significant differences between exercise and control groups were observed for any other continuous or categorical variables.


View this table:
[in this window]
[in a new window]
  		Table 2. Initial Characteristics of Subjects for Continuous Variables

 

View this table:
[in this window]
[in a new window]
  		Table 3. Initial Characteristics of Subjects for Categorical Variables

 
Primary Outcomes
As can be seen in Table 4 , there was an increase in lumbar spine BMD in the exercise groups and a decrease in the control groups. The mean difference between the two groups was 0.013 ± 0.079 g/cm2, 95% CI = 0.007–0.019. These changes were equivalent to an approximate 2% benefit in lumbar spine BMD (exercise, +1%, control, -1%). The ANOVA results in Table 5 show a statistically significant main effect difference between group and an interaction between group and test. Pairwise comparison tests for the Group x Test interaction revealed a statistically significant increase in final versus initial BMD for the exercise groups (t = 2.464, p = .014), a statistically significant decrease in final versus initial BMD for control groups (t = -3.051, p = .002), and greater initial as well as final values for exercise groups compared to control groups (initial, t = 2.544, p = .011; final, t = 3.320, p = .001). Results were similar when each study was deleted from the model once. For the exercise groups, larger increases in lumbar spine BMD were associated with assessment of BMD using DEXA versus DPA (r = -0.126, p = .018, 95% CI = -0.227–-0.022). For control subjects, larger decreases in lumbar spine BMD were associated with younger age (r = 0.170, p = .002, 95% CI = 0.064–0.272), taller stature (r = -0.109, p = .048, 95% CI = -0.215–-0.005), absence of hormone replacement therapy (r = 0.152, p = .005, 95% CI = 0.047–0.254), assessment of BMD using DPA versus DEXA (r = -0.287, p = .000, 95% CI = -0.381–-0.187) and nonrandomized versus randomized controlled trials (r = 0.172, p = .001, 95% CI = 0.067–0.273). No other statistically significant or clinically relevant relationships were observed for the exercise or control groups. Finally, no statistically significant differences in lumbar spine BMD were found when we compared the 13 studies that included IPD (ES = 0.366 ± 0.423, 95% CI = 0.131–0.600) with the 19 studies that did not include IPD (ES = 0.219 ± 0.430, 95% CI = 0.059–0.379; Qb = 1.184, p = .277).


View this table:
[in this window]
[in a new window]
  		Table 4. Lumbar Spine BMD Results (g/cm2)

 

View this table:
[in this window]
[in a new window]
  		Table 5. ANOVA Summary Table for Lumbar Spine BMD (General Linear Model)

 
Secondary Outcomes
No statistically significant main effects or interactions were found for body weight, calcium intake, or vitamin D intake.


   Discussion
Top
 Abstract
 Methods
 Results
 Discussion
References
 
The primary purpose of meta-analysis is to reach some general conclusions about a body of research. The overall results of this study suggest that exercise helps to increase and maintain lumbar spine BMD in postmenopausal women. This supports our previous meta-analytic work of summary means and lumbar spine BMD (4)(5), but is in contrast to our more recent meta-analytic work using IPD in which we found no statistically significant difference in femoral neck BMD (27). Although these are important findings, the clinical importance of an approximate 2% benefit, especially as it relates to fracture risk, cannot be elucidated at this time. However, although beyond the scope of this study, the increased strength, balance, and ambulatory skills that may be realized from a regular program of exercise may also help reduce the risk of falling and suffering subsequent fractures (28). Although we were unable to identify specific exercise programs for optimizing lumbar spine BMD, it would appear plausible to suggest that one adhere to the recent National Institutes of Health Consensus Statement that recommends participation in regular exercise, especially resistance and high-impact activities (28).

Our finding that larger decreases in BMD in the control groups were associated with younger age is not surprising given the fact that bone loss is most rapid during the early postmenopausal years (29). In addition, the observed association between the absence of hormone replacement therapy and greater decreases in lumbar spine BMD was also not surprising because hormone replacement therapy is an established therapeutic intervention for preserving BMD among postmenopausal women (28). However, we can offer no biological explanation regarding the observed association between greater decreases in lumbar spine BMD and taller stature. This is especially because it is generally believed that shorter women are considered more osteoporotic than taller women. Given this currently held notion, caution is warranted in the interpretation of this finding. Indeed, it may be that our observed association was nothing more than the play of chance given the large number of statistical tests that were conducted in our study.

Meta-analysis, like any other type of review, is limited by the availability of data and the limitations of the included studies. Thus, in addition to making the best of the existing data and trying to reach some overall conclusions regarding a body of research, it is also the meta-analyst's responsibility to identify areas of weakness to provide directions for future research.

For example, because we were unable to categorize the different types of exercise interventions, we would suggest that future researchers provide a better description of their exercise programs, especially as it relates to the forces employed during the exercise intervention. Consequently, exercise programs that provide optimal benefits to lumbar spine BMD can be recommended.

We were surprised that data on calcium intake were available for only 56% of the subjects included in this analysis. Because calcium intake is important for maintaining and/or increasing BMD in humans, it would seem reasonable to suggest that data on calcium intake be assessed and reported. In addition, because vitamin D intake is also important for the absorption of calcium and data on vitamin D intake were available for only 17% of the subjects included in this analysis, the assessment and reporting of this information also appears warranted.

Although white, non-Hispanic women are disproportionately affected with osteoporosis and low bone mass, the effect on other races is also significant. For example, the National Osteoporosis Foundation has reported that approximately 10% of black women older than 50 years have osteoporosis, and 29% have low bone mass. Additionally, 16% of American-Indian and Hispanic women aged 50 and older have osteoporosis, and 36% have low bone mass (30). Because approximately 93% of the subjects in this study were white and the responses to exercise in relation to BMD may vary by race, it is recommended that future studies include women from other ethnic groups.

Because data on the number of years that the subjects were postmenopausal were available for only 43% of the subjects included in this analysis, future research needs to include this type of information because it may be a potential confounder in relation to exercise-induced changes in lumbar spine BMD in postmenopausal women.

The fact that the vast majority of studies included in our meta-analysis were published in refereed journal articles may have led to an overestimate of the benefits of exercise on BMD at the lumbar spine because there is a tendency for authors to submit, and editors to publish, studies that yield statistically significant and positive results, i.e., publication bias (10).

For both exercise and control subjects, greater decreases in lumbar spine BMD were associated with assessment of BMD using DPA versus DEXA. Because DEXA is generally considered to be a more valid assessment of BMD and is currently the most common method used to assess BMD at the lumbar spine, the results from studies using DEXA may be more valid. The finding that greater decreases in lumbar spine BMD were associated with nonrandomized versus randomized trials suggests that randomized trials may yield more valid results.

Although the above-described associations are interesting, they should be viewed with caution for the following reasons: (i) they may have been nothing more than the play of chance given the large number of statistical tests that were conducted, (ii) we were unable to examine for potential interrelationships between variables because of missing data, and (iii) the associations accounted for only a small proportion of the total variance.

In conclusion, the results of this IPD meta-analysis suggest that exercise improves and maintains lumbar spine BMD in postmenopausal women.


   Acknowledgments
 
This study was supported by a grant from the Department of Defense, United States Army, Medical Research and Materiel Command Award #17-98-1-8513 (G.A. Kelley, Principal Investigator).

We thank the following people for taking the time and effort to contribute their individual patient data to this project: S. Bloomfield, G. Bravo, K. Brooke-Wavell, G. Caplan, S. Ebrahim, J. Iwamoto, K. Little, S. Lord, D. Martin, R. Prince, and A. Ryan.

Received January 8, 2002

Accepted April 2, 2002


   References
Top
 Abstract
 Methods
 Results
 Discussion
 References
 

  1. Kanis JA, Melton LJ, Christiansen C, Johnston CC, Khaltaev N, 1994. The diagnosis of osteoporosis. J Bone Miner Res 9:1137-1141. [Medline]
  2. Riggs BL, Melton LJ, 1995. The worldwide problem of osteoporosis: insights afforded by epidemiology. Bone 17:505S-511S. [Medline]
  3. Ray NF, Chan JK, Thamer M, Melton LJ, 1997. Medical expenditures for the treatment of osteoporotic fractures in the United States in 1995: report from the National Osteoporosis Foundation. J Bone Miner Res 12:24-35. [Medline]
  4. Kelley GA, 1998. Aerobic exercise and lumbar spine bone mineral density in postmenopausal women: a meta-analysis. J Am Geriatr Soc 46:143-152. [Medline]
  5. Kelley GA, Kelley KS, Tran ZV, 2001. Resistance training and bone mineral density in women: a meta-analysis of controlled trials. Am J Phys Med Rehabil 80:65-77. [Medline]
  6. Stewart LA, Parmar MKB, 1993. Meta-analysis of the literature or of individual patient data: is there a difference?. Lancet 341:418-422. [Medline]
  7. Stewart LA, Clarke MJ, 1995. Practical methodology of meta-analyses (overviews) using updated individual patient data. Stat Med 14:2057-2079. [Medline]
  8. Sylvester R, Collette L, Duchateau L, 2000. The role of meta-analyses in assessing cancer treatments. Eur J Cancer 36:1351-1358.
  9. Berlin JA, 1997. Does blinding of readers affect the results of meta-analyses?. Lancet 350:185-186. [Medline]
  10. Hedges LV, Olkin I. Statistical Methods for Meta-Analysis. San Diego, Calif: Academic Press; 1985.
  11. Cohen J, 1992. A power primer. Psychol Bull 112:155-159. [Medline]
  12. Bloomfield SA, Williams N, Lamb D, Jackson R, 1993. Non-weightbearing exercise may increase lumbar spine bone mineral density in healthy postmenopausal women. Am J Phys Med Rehabil 72:204-209. [Medline]
  13. Bravo G, Gauthier P, Roy PM, et al. 1996. Impact of a 12-month exercise program on the physical and psychological health of osteopenic women. J Am Geriatr Soc 44:756-762. [Medline]
  14. Brooke-Wavell KSF, Jones PRM, Hardman AE, 1997. Brisk walking reduces calcaneal bone loss in post-menopausal women. Clin Sci (Colch) 92:75-80. [Medline]
  15. Caplan GA, Ward JA, Lord SR, 1993. The benefits of exercise in postmenopausal women. Aust J Public Health 17:23-26. [Medline]
  16. Ebrahim S, Thompson PW, Baskaran V, Evans K, 1997. Randomized placebo-controlled trial of brisk walking in the prevention of postmenopausal osteoporosis. Age Ageing 26:253-260. [Abstract/Free Full Text]
  17. Grove CA. The Effects of High Impact Exercise Versus Low Impact Exercise on Bone Density in Postmenopausal Women [dissertation]. Columbia, Mo: University of Missouri; 1990.
  18. Grove KA, Londeree BR, 1992. Bone density in postmenopausal women: high impact versus low impact exercise. Med Sci Sports Exerc 24:1190-1194. [Medline]
  19. Iwamoto J, Takeda T, Otani T, Yabe Y, 1998. Effect of increased physical activity on bone mineral density in postmenopausal osteoporotic women. Keio J Med 47:157-161. [Medline]
  20. Little KD. Effect of Exercise Mode on Bone Mineral Mass in Recently Postmenopausal Women [dissertation]. Kent, Ohio: Kent State University; 1992.
  21. Lord SR, Ward JA, Williams P, Zivanovic E, 1996. The effects of community exercise program on fracture risk factors in older women. Osteoporos Int 6:361-367. [Medline]
  22. Martin D, Notelovitz M, 1993. Effects of aerobic training on bone mineral density of postmenopausal women. J Bone Miner Res 8:931-936. [Medline]
  23. Prince R, Devine A, Criddle A, et al. 1995. The effects of calcium supplementation (milk powder or tablets) and exercise on bone density in postmenopausal women. J Bone Miner Res 10:1068-1075. [Medline]
  24. Pruitt L. The Effect of an Isotonic Weight-Training Program Upon Bone Density and Selected Indices of Bone Turnover in Early Postmenopausal Women [dissertation]. Columbus, Ohio: Ohio State University; 1989.
  25. Pruitt LA, Jackson RD, Bartels RL, Lehnhard HJ, 1992. Weight-training effects on bone mineral density in early postmenopausal women. J Bone Miner Res 7:179-185. [Medline]
  26. Ryan AS, Nicklas BJ, Dennis KE, 1998. Aerobic exercise maintains regional bone mineral density during weight loss in postmenopausal women. J Appl Physiol 84:1305-1310. [Abstract/Free Full Text]
  27. Kelley GA, Kelley KS, Tran ZV. Exercise and bone mineral density at the femoral neck in postmenopausal women: a meta-analysis of controlled clinical trials using individual patient data. Dis Management Clin Outcomes. In press.
  28. NIH Consensus Development Panel on Osteoporosis Prevention Diagnosis, and Therapy2001. Osteoporosis prevention, diagnosis, and therapy. JAMA 285:785-795. [Abstract/Free Full Text]
  29. Beck BR, Shoemaker MR, 2000. Osteoporosis: understanding key risk factors and therapeutic options. Physician Sports Med 28:69-84.
  30. National Osteoporosis Foundation. 1996 and 2015 Osteoporosis Prevalence Figures: State by State Report. Washington, DC: National Osteoporosis Foundation; 1997.



This article has been cited by other articles:


Home page
 JAMAHome page
S. R. Cummings
A 55-Year-Old Woman With Osteopenia
JAMA, December 6, 2006; 296(21): 2601 - 2610.
[Abstract] [Full Text] [PDF]

Home page
 Biol Res NursHome page
J. V. Jessup, C. Horne, R. K. Vishen, and D. Wheeler
Effects of Exercise on Bone Density, Balance, and Self-Efficacy in Older Women
Biol Res Nurs, January 1, 2003; 4(3): 171 - 180.
[Abstract] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Services
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via HighWire
PubMed
Right arrow PubMed Citation

HOME ARCHIVE SEARCH TABLE OF CONTENTS