To assess the effect of exercise on bone mass in pre- and postmenopausal women.

MEDLINE was searched from 1966 to January 1998 for publications in any language, using the MeSH terms 'exercise', 'bone mineral density' 'and 'osteoporosis'. The reference lists of all the identified RCTs and some reviews were examined.

Study designs of evaluations included in the review

Randomised controlled trials (RCTs) were eligible. Studies were subsequently excluded from the meta-analysis for the following reasons: the studies involved both impact and non-impact exercises; the results for the spine or femoral neck were not reported; or the results were reported in insufficient detail. One study was excluded as it randomised one upper limb of each woman to exercise with the other limb acting as the control.

Specific interventions included in the review

Comparisons of exercise and control interventions were eligible. The exercise interventions were classified as impact or non-impact exercises. Impact interventions included walking, running, aerobics, dancing, using treadmills, and heel drops. Non-impact exercises included resistance training, strength training and weightlifting. The duration of the programmes ranged from 3 to 60 months.

Participants included in the review

Women were eligible. The included participants were premenopausal and postmenopausal women; their mean age ranged from 20 to 44 years, and 45 to 76 years, respectively.

Outcomes assessed in the review

The inclusion criteria were not defined in terms of the outcomes. Changes in bone density over the study duration were assessed in the review.

How were decisions on the relevance of primary studies made?

The authors do not state how the papers were selected for the review, or how many of the reviewers performed the selection.

The included studies were restricted to RCTs but no formal validity assessment was performed.

The authors do not state how the data were extracted for the review, or how many of the reviewers performed the data extraction.

The following information were tabulated in the review: author; the mean age of the participants; intervention details; the number of women with data; the duration of the exercise programme; drop-out rate; compliance rate; and measurement sites. The mean difference in the rate of bone loss (or gain) between the exercise and control groups was calculated for each study. In studies with more than one exercise group, only the results from the most intensive exercise programme were included. If the study included separate impact and non-impact groups, each exercise group was included separately in the relevant meta-analysis. In studies where values were not reported, the data were estimated from graphs.

How were the studies combined?

A pooled, mean percentage bone loss and 95% confidence intervals (CIs) were calculated using the fixed-effect model described by Petitti, with weighting by the inverse of the variance (see Other Publications of Related Interest no.1). When standard deviations were not reported, they were estimated from the two studies of the same bone site with the closest sample sizes (see Other Publications of Related Interest no.2). The data were pooled separately for menopausal and postmenopausal women, for impact and non-impact exercise, and for femoral neck and lumbar spine bone density. A meta-regression analysis was performed in postmenopausal women to explore the effect of age, the duration of the exercise programme, and the type of exercise, on the change in lumbar spine bone density.

How were differences between studies investigated?

Statistical heterogeneity was assessed using the methods described by Petitti (see Other Publication of Related Interest no.1). When significant heterogeneity was detected, the influence of each study on the heterogeneity was assessed.

Thirty-two RCTs (data from 2,032 women) were included. Of these, 24 studies were conducted in premenopausal women and 8 were conducted in postmenopausal women.

Exercise programmes varied considerable between studies in terms of their content and duration. Compliance rates, defined generally as the percentage of sessions that were completed, ranged from 39 to 100%. The drop-out rates ranged from 0 to 47%.

Postmenopausal women (24 RCTs, 1,647 women).

Impact and non-impact exercise had a positive effect on bone mass at both the lumbar spine and the femoral neck. Significant heterogeneity was detected in all of the reported analyses. At the lumbar spine, the difference in bone density between the exercise and control groups was 1.6% (95% CI: 1.0, 2.2) for impact exercise, and 1.0% (95% CI: 0.4, 1.6) for non-impact exercise. Heterogeneity was due to one RCT in each analysis.

At the femoral neck, the difference in bone density between the exercise and control groups was 0.9% (95% CI: 0.5, 1.3). Heterogeneity was due to one RCT. The difference in bone density between the non-impact exercise and control groups was 1.4% (95% CI: 0.2, 2.6).

Premenopausal women (8 RCTs, 385 women).

Impact and non-impact exercise had a positive effect on bone mass at the lumbar spine. There was no evidence of heterogeneity for any of the reported analyses (p>0.25).

There was no statistically significant effect of impact exercise on bone density at the femoral neck. The improvement in bone density was 0.9% (95% CI: -0.2, -2.0, p>0.05). There were insufficient data to calculate the effect of non-impact exercise on the femoral neck.

In the meta-regression analysis of lumbar spine density in postmenopausal women, the coefficient for impact versus non-impact exercise just failed to reach statistical significance (Beta = -1.7; p=0.06). No significant effect was found for the duration of the exercise programme (p=0.24).

The results from the RCTs showed that exercise slows bone loss from the lumbar spine, and probably the neck of the femur. Exercise probably has a similar effect in premenopausal women but more studies are needed to reach a firm conclusion.

The aims were stated and the inclusion criteria were defined in terms of the study design, intervention, and participants. The inclusion criteria were not defined in terms of the outcome. No language restrictions were applied, but only one relevant database was searched and there was no attempt to locate unpublished material, thus raising the possibility of publication bias. The methods used to select the studies were not described. The included studies were restricted to RCTs but their validity was not formally assessed. Relevant data were tabulated but the methods used to extract the data were not described.

Significant statistical heterogeneity was detected in several analyses and for these analyses, the meta-analysis was used inappropriately to combine the data. Where significant heterogeneity was found, the contribution of each study to the heterogeneity appears to have been assessed and a meta-regression analysis was performed to investigate the influence of some variables on the results. The authors highlighted the considerable differences between the studies, and the variable compliance and drop-out rates in the discussion. However, no mention was made of the validity of the methods used to assess bone density.

The results from the review should be interpreted cautiously in view of these limitations, in particular, the significant heterogeneity among the studies of postmenopausal women.

Practice: The authors state that impact exercises probably have a positive effect at the femoral neck in postmenopausal women.

Research: The authors state that RCTs are required to compare different exercise regimens on both bone and fall-related outcomes.

1. Petitti DB. Meta-analysis, decision analysis, and cost-effectiveness analysis: methods for quantitative synthesis in medicine. New York: Oxford University Press; 1984. 2. Hedges L, Olkin I. Statistical methods for meta-analysis. Florida: Academic Press; 1985.

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.