Exercise for preventing and treating osteoporosis in postmenopausal women

Abstract

Background

Osteoporosis is a condition resulting in an increased risk of skeletal fractures due to a reduction in the density of bone tissue. Treatment of osteoporosis typically involves the use of pharmacological agents. In general it is thought that disuse (prolonged periods of inactivity) and unloading of the skeleton promotes reduced bone mass, whereas mechanical loading through exercise increases bone mass.

Objectives

To examine the effectiveness of exercise interventions in preventing bone loss and fractures in postmenopausal women.

Search methods

During the update of this review we updated the original search strategy by searching up to December 2010 the following electronic databases: the Cochrane Musculoskeletal Group's Trials Register; the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library, 2010 Issue 12); MEDLINE; EMBASE; HealthSTAR; Sports Discus; CINAHL; PEDro; Web of Science; Controlled Clinical Trials; and AMED. We attempted to identify other studies by contacting experts, searching reference lists and searching trial registers.

Selection criteria

All randomised controlled trials (RCTs) that met our predetermined inclusion criteria.

Data collection and analysis

Pairs of members of the review team extracted the data and assessed trial quality using predetermined forms. For dichotomous outcomes (fractures), we calculated risk ratios (RRs) using a fixed‐effect model. For continuous data, we calculated mean differences (MDs) of the percentage change from baseline. Where heterogeneity existed (determined by the I² statistic), we used a random‐effects model.

Main results

Forty‐three RCTs (27 new in this update) with 4320 participants met the inclusion criteria. The most effective type of exercise intervention on bone mineral density (BMD) for the neck of femur appears to be non‐weight bearing high force exercise such as progressive resistance strength training for the lower limbs (MD 1.03; 95% confidence interval (CI) 0.24 to 1.82). The most effective intervention for BMD at the spine was combination exercise programmes (MD 3.22; 95% CI 1.80 to 4.64) compared with control groups. Fractures and falls were reported as adverse events in some studies. There was no effect on numbers of fractures (odds ratio (OR) 0.61; 95% CI 0.23 to 1.64). Overall, the quality of the reporting of studies in the meta‐analyses was low, in particular in the areas of sequence generation, allocation concealment, blinding and loss to follow‐up.

Authors' conclusions

Our results suggest a relatively small statistically significant, but possibly important, effect of exercise on bone density compared with control groups. Exercise has the potential to be a safe and effective way to avert bone loss in postmenopausal women.

Plain language summary

Exercise for preventing and treating osteoporosis in postmenopausal women

This summary of a Cochrane review presents what we know from research about the effect of exercise on bone mass in postmenopausal women.

The review shows that for postmenopausal women

‐        Exercise will improve bone mineral density slightly.

‐        Exercise will reduce the chances of having a fracture slightly.

These results might have happened by chance.

What is osteoporosis and exercise

Bone is a living, growing part of your body. Throughout your lifetime, new bone cells grow and old bone cells break down to make room for the new, stronger bone. When you have osteoporosis, the old bone breaks down faster than the new bone can replace it. As this happens, the bones lose minerals (such as calcium). This makes bones weaker and more likely to break even after a minor injury, like a little bump or fall.

Exercise interventions are typically those that stress or mechanically load bones (when bones support the weight of the body or when movement is resisted for example when using weights) and include aerobics, strength training, walking and tai chi.

Best estimate of what happens to postmenopausal women who exercise

Bone mineral density at the spine

People who exercised had on average 0.85% less bone loss than those who didn't exercise.

People who engaged in combinations of exercise types had on average 3.2% less bone loss than those who did not exercise.

Bone mineral density at the hip

People who exercised had on average 1.03% less bone loss than those who didn't exercise.

People who exercised by strength training had on average 1.03% less bone loss.

Fractures

4 less women out of 100 who did exercise had a fracture.  (Absolute difference 4%).

7 women out of 100 who exercised had a fracture.

11 women out of 100 who did not exercise had a fracture.

Summary of findings

Summary of findings for the main comparison. Exercise for preventing and treating osteoporosis in postmenopausal women.

Exercise for preventing and treating osteoporosis in postmenopausal women
Patient or population: preventing and treating osteoporosis in postmenopausal women Settings: Intervention: exercise
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No. of Participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Exercise
Total number of fractures radiographs	Study population		OR 0.61 (0.23 to 1.64)	539 (4 studies)	⊕⊕⊕⊕ high	Absolute difference = 4% Not statistically significant.
	106 per 1000	67 per 1000 (27 to 163)
Bone mineral density % change: spine	The mean bone mineral density % change: spine ranged across control groups from ‐4.38 to 1.05 %	The mean bone mineral density % change: spine in the intervention groups was 0.85 % higher (0.62 to 1.07 higher)		1441 (24 studies)	⊕⊕⊕⊕ high
Bone mineral density % change: femoral neck	The mean bone mineral density % change: femoral neck ranged across control groups from ‐3.19 to 3.12 %	The mean bone mineral density % change: femoral neck in the intervention groups was 0.08 % lower (1.08 lower to 0.92 higher)		1338 (19 studies)	⊕⊕⊝⊝ low1
Bone mineral density % change: total hip	The mean bone mineral density % change: total hip ranged across control groups from ‐2.18 to 2.61 %	The mean bone mineral density % change: total hip in the intervention groups was 0.41 % higher (0.64 lower to 1.45 higher)		863 (13 studies)	⊕⊕⊕⊕ high
Bone mineral density % change: trochanter	The mean bone mineral density % change: trochanter ranged across control groups from ‐1.62 to 2.94 %	The mean bone mineral density % change: trochanter in the intervention groups was 1.03 % higher (0.56 to 1.49 higher)		815 (10 studies)	⊕⊕⊕⊕ high
Adverse events: Falls	see comment	see comment	not estimable	378 (3 studies)	see comment	Reported as adverse events there were 75 falls reported in the exercise groups and 55 in the control groups2
Other adverse events	see comment	see comment	not estimable	907 (11 studies)	see comment	Events included muscle soreness, joint pain, headache, itching There were 60 events reported in the exercise groups and 5 in the control groups3
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; OR: Odds ratio
GRADE Working Group grades of evidence High quality: further research is very unlikely to change our confidence in the estimate of effect Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate Very low quality: we are very uncertain about the estimate

¹ Significant heterogeneity observed even with random effects model.

² Reported as total number of falls but individuals may have had more than one fall.

³ Events reported as adverse for participants in intervention groups, generally no mention of event monitoring in control groups.

Background

Description of the condition

Osteoporosis is a condition resulting in an increased risk of skeletal fractures due to a reduction in the density of bone tissue (CDC 1991). The most common clinical manifestations of osteoporosis are fractures of the hip, vertebrae or wrist. Osteoporotic‐related fractures are responsible for excess mortality, morbidity, chronic pain, reduction in quality of life, admission to long‐term care and health and social care costs (Papaioannou 2010). For women at the age of 50 years in developed countries, the remaining lifetime possibility of osteoporotic fractures exceeds 40%; the remaining lifetime probability for hip fracture alone exceeds 20% (Bessette 2008). The excess mortality associated with a hip fracture has been estimated to be 20% (Cooper 1993). The number of osteoporotic‐related fractures is certain to increase as a result of the ageing population (WHO 1994).

Prevention of osteoporotic‐related fractures is based on the ability to estimate fracture probability by means of risk factor assessment. The quality of bone, (the total characteristics of the bone that influence the bone’s resistance to fracture), is determined by a number of factors including bone geometry, cortical thickness and porosity, trabecular bone morphology and intrinsic properties of bony tissue.  Low bone mass, detected by bone densitometry, is one of the most important risk factors. Bone mineral density (BMD) measured by dual X‐ray absorptiometry (DXA) is reported to account for 60% to 70% of the variation in bone strength (Ammann 2003), and prospective studies have documented that the lifetime risk of an osteoporotic‐related fracture increases 1.5 to 3 times with each standard deviation (SD) decrease in bone density (Cummings 1993). However Zebaze 2010 reported that most bone loss is cortical, not trabecular, and occurs after the age of 65 years; and that the resulting structural decay, including the magnitude of intracortical remodelling and intracortical porosity, are poorly captured by current measurement methods. Although measurements of BMD contribute to the prediction of fracture risk they cannot identify individuals who will have a fracture (Marshall 1996) as many fractures, particularly in older populations, are results of falls which are influenced by environmental and other medical causes, e.g. impaired visual function, muscle strength and balance (Gillespie 2009).

The treatment of hip fracture and the hospitalisation required following all types of fracture account for most of the economic costs associated with osteoporosis (Cooper 1993). Therefore, the prevention of fractures is the primary goal of intervention. See other Cochrane systematic reviews for pharmacological interventions for osteoporosis: alendronate (Wells 2008a); etidronate (Wells 2008b); fluoride (Haguenauer 2004); risedronate (Wells 2008c); and strontium renalate (Cranney 2006).

Description of the intervention

Various exercise interventions, designed to stimulate bone growth and preserve bone mass have been described and evaluated. Interventions are typically those that stress or mechanically load bones (when bones support the weight of the body or when movement is resisted, for example when using weights) and include aerobics, weight bearing and resistance exercises.

How the intervention might work

In general, it is thought that disuse (not using the limbs or prolonged periods of inactivity) and unloading of the skeleton promotes reduced bone mass (Zerwekh 1998), whereas loading promotes increased bone mass. The effects of mechanical loading have been demonstrated in athletes undertaking high‐impact exercise (Taaffe 1997) and in rats (Robling 2002). Mechanical loading through exercise has the potential to be a safe and effective way to avert or delay the onset of osteoporosis in postmenopausal women. The previous version of this review (Bonaiuti 2002) concluded that exercise has beneficial effects on bone density of the hip and spine, although long term‐studies including fracture data are rare. In addition, strength and balance exercises contribute to fracture risk reduction through their efficacy in reducing falls risk (Gillespie 2009).

Why it is important to do this review

The high prevalence and staggering costs (Burge 2007) of osteoporotic‐related fractures in postmenopausal women means prevention and management of this disease is important. There continues to be much interest in the effects of exercise on bone as a nonpharmacological intervention. A systematic review is required to identify the number of new trials in this area and summarise the evidence for healthcare professionals, policy makers, researchers and others with an interest in this area.

Objectives

To examine the effectiveness of exercise in preventing bone loss in postmenopausal women by determining whether or not exercise slows bone loss and has a beneficial effect on the axial (the skull, spine and rib cage) and appendicular (the bones of the limbs and pelvis) bone density in postmenopausal women.

Methods

Criteria for considering studies for this review

Types of studies

We considered all randomised controlled clinical trials (RCTs) of exercise in healthy postmenopausal women.

Types of participants

We included studies where participants were healthy postmenopausal women (including those with previous fractures) aged between 45 and 70 years.

Types of interventions

We included all RCTs with an exercise programme (e.g. walking, callisthenics and resisted strengthening) assumed to be adequate to improve aerobic capacity, or both aerobic capacity and muscle strength compared to standard therapy (e.g. usual activity or placebo with or without pharmacological consumption).

Types of outcome measures

Major outcomes

• Number of incident fractures: vertebral and non‐vertebral (hip and wrist).

Secondary outcomes

• Bone mass including BMD, bone mineral content (BMC), or calcium bone index (CaBI) immediately postintervention and at follow‐up.

BMD was measured by Single‐Photon Absorptiometry (SPA), Dual‐Photon Absorptiometry (DPA), Quantitative Computerised Tomography (QCT) or DXA at baseline, immediately postintervention and at follow‐up. All results were converted to the percentage change of BMD from baseline values. The difference between the percentage lost in the exercise group and the percentage lost in the control group was used as the measure of effect in pooling the data.

• Serious adverse events including death.

• Minor adverse events including falls.

Search methods for identification of studies

Electronic searches

To identify exercise trials, we searched the following five electronic databases: the Cochrane Musculoskeletal Group's Trials Register; the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library); MEDLINE; EMBASE; and Current Contents from 1966 to January 2000 with no language restrictions, according to the methods suggested by Dickersin 1994 and Haynes 1994 and the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). During this current update of the review, we undertook a further search. We searched the following electronic databases; the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library, 2010); MEDLINE (Appendix 1); EMBASE (Appendix 2); HealthSTAR; Sports Discus; CINAHL (Appendix 3); PEDro (Appendix 4); Web of Science; Controlled Clinical Trials (Appendix 5); and AMED (Appendix 6) (all to December 2010).

Searching other resources

In addition, we searched the reference lists of included trials and trials registers, and contacted content experts for additional studies and data.

Data collection and analysis

Selection of studies

Following an a priori protocol, at least two review authors (BS, TH, LD and FD) independently reviewed the eligibility criteria for abstracts for inclusion in this review. We screened all titles and/or abstracts generated by the searches for potentially relevant studies based on the following criteria: the type of study; type of participants; type of intervention; and type of outcome measurements. We assessed the full‐length articles of the selected titles and/or abstracts for eligibility (for a full description see Criteria for considering studies for this review). We resolved disagreements by consensus or third‐party adjudication.

Data extraction and management

Pairs of members of the review team used a customised data extraction tool, tested prior to use, to independently extract data. We resolved disagreements by consensus or third‐party adjudication. We attempted to contact authors of studies where there was inadequate reporting of data, to enable clarification and where appropriate, to allow pooling. Where available and appropriate, we presented quantitative data for the outcomes listed in the inclusion criteria in the analyses. Where studies reported standard errors (SEs) of the means, we obtained SDs by multiplying SEs of means by the square root of the sample size.

In order to assess efficacy, we extracted raw data for outcomes of interest (means and SDs for continuous outcomes and number of events for dichotomous outcomes) where available in the published reports. Wherever we converted or imputed reported data, we recorded this in the notes section of the Characteristics of included studies. All trials reported continuous outcomes as end‐point scores (i.e. mean and SD of the variable at follow‐up, assuming baseline comparability).

Assessment of risk of bias in included studies

We assessed risk of bias for each included study using the Cochrane Collaboration's 'Risk of bias' tool (Higgins 2011). Pairs of members of the review team reported the following six key domains: sequence generation; allocation concealment; blinding; incomplete outcome data; selective outcome reporting; and "other bias" (comparability of treatment and control group at entry, and appropriateness of duration of surveillance). In cases of disagreement between the review authors, we used consensus to make a decision. The final assessments for all included studies are presented in a 'risk of bias' table (see Characteristics of included studies).

Measures of treatment effect

For each trial, we calculated risk ratios (RRs) and 95% CIs for dichotomous outcomes, and mean differences (MDs) and 95% CIs for continuous outcomes (reporting mean and SD or standard error (SE) of the mean). Where appropriate, we pooled results of comparable groups of trials using the fixed‐effect model and 95% CIs.

Unit of analysis issues

We reported the level at which randomisation occurred in the included studies as specified by the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2011). Possible variations in study designs include cluster randomised trials, cross‐over trials, multiple observations, re‐occurring events, multiple treatments and multiple intervention groups.

Dealing with missing data

If we discovered missing data during data extraction, we attempted to contact the original investigators of the study to request the required information. We anticipated that it may also have been necessary to conduct a sensitivity analysis if assumptions were made (Deeks 2011). We have also described the potential effect of missing data upon conclusions drawn from this review.

Assessment of heterogeneity

We tested heterogeneity between comparable trials using a standard Chi² test and considered heterogeneity statistically significant at P < 0.1 after due consideration of the value of the I² statistic; a value greater than 50% may indicate substantial heterogeneity.

Assessment of reporting biases

If there were sufficient studies, we intended to assess the possibility of publication bias with funnel plots.

Data synthesis

We used Cochrane Review Manager software to meta‐analyse the statistics as described below (RevMan 2011). We used 95% CIs for all outcomes.

Continuous outcomes

We calculated mean differences (MDs) using a fixed‐effect model as we measured outcomes on standard scales. We explored possible reasons for heterogeneity in terms of prevention versus treatment studies (primary versus secondary prevention), trial quality, differing populations and exercise programmes. We used a random‐effects model to further analyse the results which were determined to indicate substantial heterogeneity (i.e. after due consideration of the value of the I² statistic, a value greater than 50%).

Dichotomous outcomes

For interpretation of the dichotomous outcome measures in this review, we selected the RR using a random‐effects model, since this is the most appropriate statistic for the interpretation of pooled data where the event is common and where there is statistical heterogeneity between trials (Deeks 1998).

We performed appropriate statistical analysis using Review Manager (RevMan 2011) in accordance with the Cochrane Handbook for Systematic Reviews of Interventions (Deeks 2011).

Subgroup analysis and investigation of heterogeneity

In the presence of heterogeneity, we pooled the results of comparable groups using the random‐effects model and 95% CIs.

We performed separate outcome analyses to determine the effectiveness of different categories of exercise interventions versus control.

Where the data allowed, we also anticipated performing separate outcome analyses to test the following hypotheses:

1. effectiveness is not dependent on the duration and/or intensity of the physical activity/exercise intervention;

2. effectiveness is not dependent on the setting in which the physical activity/exercise intervention is delivered; and

3. effectiveness is not dependent on the level or type of supervision of the physical activity/exercise intervention.

Sensitivity analysis

We anticipated that we would undertake sensitivity analyses, when indicated, to investigate the effects of methodological quality, for example, allocation concealment and intention‐to‐treat (ITT) analysis or where cluster randomised trials are combined with each other or with other studies in a meta‐analysis.

Grading of evidence and summary of findings table

Major outcomes (including benefits and adverse events) are presented in the Table 1 which provides information on the quality of evidence and the magnitude of the intervention effect, as well as a summary of the main outcome data. We have also presented an assessment of the overall quality of evidence per outcome (high, moderate, low and very low) using the GRADE approach as outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2011).

Results

Description of studies

Forty‐three RCTs (27 new in this update) with 4320 participants met the inclusion criteria. Although included in the original version of this review, we excluded (Kerr 1996) from this update because participants were their own control, with one side of body randomised to a different exercise type. Forty‐one of the 43 included studies were published in English, one in German (Von Stengel 2009) and one in Italian (Tolomio 2009).

Exercise interventions

For details of the content of individual interventions see Characteristics of included studies. Many factors influence the effectiveness of exercise interventions:

1. attendance;

2. adherence to the programme;

3. accuracy of the assessment system;

4. type of exercise; and

5. duration, intensity, frequency and length of exercise programme.

The World Health Organization (WHO) defines adherence as "the extent to which a person's behavior such as taking medication, following a diet, executing lifestyle changes like exercising, corresponds with agreed recommendations from a health care provider", (WHO 2003) this is clearly different to attendance. The quality of the reporting of these factors varied in the 43 studies included in this review.

Compliance/adherence

Thirty‐three studies reported compliance/adherence to the programmes and 10 did not (Bravo 1996; Brentano 2008; Chuin 2009; Hatori 1993; Iwamoto 2005; Lau 1992; Russo 2003; Sakai 2010; Sinaki 1989; Tolomio 2009). The attendance rate ranged from 48% (Preisinger 1995) to 93% (Bemben 2000) and the adherence rate to the exercise programmes ranged from 39% (Prince 1995) to 95% (Bocalini 2009).

Thirteen studies reported the accuracy of the assessing instrument (i.e. the test‐retest reliability) (Grove 1992; Hatori 1993; Lau 1992; Lord 1996; Martin 1993; Nelson 1994; Preisinger 1995; Prince 1991; Prince 1995; Pruitt 1996; Revel 1993; Sinaki 1989; Smidt 1992). Instrument accuracy ranged from 0.4% (Hatori 1993) to 3% (Grove 1992).

Setting

The studies included participants from North America (Bravo 1996; Bemben 2000; Chilibeck 2002; Chow 1987; Chubak 2006; Chuin 2009; Going 2003; Grove 1992; Maddalozzo 2007; Martin 1993; Metcalfe 2001; Nelson 1994; Newstead 2004; Papaioannou 2003; Pruitt 1996; Rubin 2004; Russo 2003; Sinaki 1989; Smidt 1992), Australia (Kerr 2001; Lord 1996; Prince 1991; Prince 1995), Europe (Bergstrom 2008; Cheng 2002; Ebrahim 1997; Englund 2005; Karinkanta 2007; Korpelainen 2006; Preisinger 1995; Revel 1993; Tolomio 2009; Uusi‐Rasi 2003; Verschueren 2004; Von Stengel 2009), Japan (Hatori 1993; Iwamoto 2001; Iwamoto 2005; Sakai 2010), China (Chan 2004; Lau 1992) and Brazil (Bocalini 2009; Brentano 2008).

Length of exercise programmes

The length of the exercise programmes varied in the included studies; 10 were less than 12 months long (Bemben 2000; Bocalini 2009; Brentano 2008; Chuin 2009; Hatori 1993; Lau 1992; Russo 2003; Sakai 2010; Tolomio 2009; Verschueren 2004); 26 were 12 months long (Bergstrom 2008; Bravo 1996; Chan 2004; Chilibeck 2002; Chow 1987; Chubak 2006; Englund 2005; Going 2003; Grove 1992; Iwamoto 2001; Iwamoto 2005; Karinkanta 2007; Kerr 2001; Lord 1996; Maddalozzo 2007; Martin 1993; Metcalfe 2001; Nelson 1994; Newstead 2004; Papaioannou 2003; Pruitt 1996; Revel 1993; Rubin 2004; Smidt 1992; Uusi‐Rasi 2003; Von Stengel 2009); and seven were greater than 12 months (Ebrahim 1997; Kerr 2001; Korpelainen 2006; Preisinger 1995; Prince 1991; Prince 1995; Sinaki 1989).

Number of years postmenopausal

The number of years postmenopausal was reported in 15 studies (Bemben 2000; Bravo 1996; Chan 2004; Chilibeck 2002; Chubak 2006; Going 2003; Grove 1992; Maddalozzo 2007; Nelson 1994; Preisinger 1995; Prince 1991; Revel 1993; Russo 2003; Uusi‐Rasi 2003; Verschueren 2004).

Exercise interventions

All the exercise interventions described were land based except in one study (Tolomio 2008) which included both land and water based exercise. We categorised the exercise interventions into the following six categories.  

• Static weight bearing (SWB); including single leg standing.

• Dynamic weight bearing exercise low force (DWBLF); including walking and Tai chi.

• Dynamic weight bearing exercise high force (DWBHF); including jogging, jumping, running, dancing and vibration platform.

• Non‐weight bearing exercise low force (NWBLF); e.g. low load, high repetition strength training.

• Non‐weight bearing exercise high force (NWBHF); e.g. progressive resisted strength training.

• Combination (COMB); more than one of the above exercise interventions.

Exercise training consisted of: DWBLF in 12 studies (Bravo 1996; Chan 2004; Chow 1987; Ebrahim 1997; Grove 1992; Hatori 1993; Lau 1992; Lord 1996; Martin 1993; Preisinger 1995; Prince 1991; Prince 1995); DWBHF in 11 studies (Cheng 2002; Going 2003; Grove 1992; Iwamoto 2005; Karinkanta 2007; Maddalozzo 2007; Newstead 2004; Rubin 2004; Russo 2003; Uusi‐Rasi 2003; Verschueren 2004); NWBLF in six studies (Bemben 2000; Brentano 2008; Kerr 2001; Pruitt 1996; Revel 1993; Sinaki 1989); NWBHF in nine studies (Bemben 2000; Brentano 2008; Bocalini 2009; Chilibeck 2002; Chuin 2009; Kerr 2001; Nelson 1994; Pruitt 1996; Smidt 1992); COMB in 11 studies (Bergstrom 2008; Chow 1987; Chubak 2006; Englund 2005; Iwamoto 2001; Karinkanta 2007; Korpelainen 2006; Metcalfe 2001; Papaioannou 2003; Tolomio 2009; Von Stengel 2009); and SWB in one study (Sakai 2010).

It should be noted that some studies included more than one exercise intervention arm.

Frequency of the exercise intervention

The frequency of the sessions for the majority of studies was two or three times per week. The exception being three studies where participants had daily exercise sessions (Iwamoto 2001; Revel 1993; Sakai 2010) and seven who had four to six sessions per week (Bergstrom 2008; Chan 2004; Cheng 2002; Lau 1992; Prince 1995; Sinaki 1989; Smidt 1992).

Content of the exercise intervention

The content of the training session was reported in five studies (Lau 1992; Lord 1996; Preisinger 1995; Prince 1991; Prince 1995) and two studies assessed the effect of the repetition of only one exercise (Revel 1993; Sinaki 1989). There were five studies that described that exercise intensity was determined by maximal heart rate (Bravo 1996; Chow 1987; Ebrahim 1997; Hatori 1993; Martin 1993), and three studies reported how the maximum strength was measured (Nelson 1994; Pruitt 1996; Smidt 1992). One study measured the exercise intensity in relation to body weight (Grove 1992).

Controls

In all but twelve studies the controls were invited to continue their usual activity without any exercise prescription. One study reported the control group performing 45 minute stretching sessions once a week (Chubak 2006); one performing upper limb exercises only (Ebrahim 1997); and one performing gentle exercise and relaxation (Von Stengel 2009). In five studies controls were divided into usual activity with drug interventions or usual activity alone (Lau 1992; Martin 1993; Maddalozzo 2007; Prince 1991; Prince 1995). A placebo device was used in only one study (Rubin 2004), and there were three studies which did not report on the control group (Englund 2005; Hatori 1993; Papaioannou 2003).

Outcome measures

Fracture rate was a primary outcome measure in only one trial (Iwamoto 2005), but was reported as an adverse event at follow‐up in three (Chan 2004; Karinkanta 2007; Korpelainen 2006). Falls were reported as adverse events in four trials (Chan 2004; Ebrahim 1997; Iwamoto 2005; Karinkanta 2007).

BMD was measured at the lumbar spine in 30 studies (Bemben 2000; Bergstrom 2008; Bocalini 2009; Bravo 1996; Chan 2004; Chilibeck 2002; Chuin 2009; Ebrahim 1997; Englund 2005; Going 2003; Grove 1992; Hatori 1993; Iwamoto 2001; Kerr 2001; Lau 1992; Lord 1996; Maddalozzo 2007; Martin 1993; Metcalfe 2001; Nelson 1994; Newstead 2004; Papaioannou 2003; Prince 1995; Pruitt 1996; Revel 1993; Rubin 2004; Sinaki 1989; Smidt 1992; Uusi‐Rasi 2003; Von Stengel 2009).

BMD was measured at the hip in 30 studies (Bemben 2000; Bergstrom 2008; Bocalini 2009; Bravo 1996; Brentano 2008; Chan 2004; Cheng 2002; Chilibeck 2002; Chuin 2009; Ebrahim 1997; Englund 2005; Going 2003; Kerr 2001; Korpelainen 2006; Lau 1992; Lord 1996; Maddalozzo 2007; Metcalfe 2001; Nelson 1994; Newstead 2004; Papaioannou 2003; Prince 1995; Pruitt 1996; Rubin 2004; Sakai 2010; Smidt 1992; Tolomio 2009; Uusi‐Rasi 2003; Verschueren 2004; Von Stengel 2009). These were subdivided into the following sites: femoral neck (Bemben 2000; Bocalini 2009; Bravo 1996; Brentano 2008; Chan 2004; Chuin 2009; Ebrahim 1997; Englund 2005; Going 2003; Kerr 2001; Korpelainen 2006; Lau 1992; Lord 1996; Maddalozzo 2007; Nelson 1994; Newstead 2004; Papaioannou 2003; Pruitt 1996; Sakai 2010; Tolomio 2009; Uusi‐Rasi 2003); intertrochanteric (Brentano 2008; Sakai 2010); trochanter (Bemben 2000; Brentano 2008; Chan 2004; Chilibeck 2002; Englund 2005; Going 2003; Kerr 2001; Korpelainen 2006; Lord 1996; Maddalozzo 2007; Sakai 2010; Smidt 1992; Uusi‐Rasi 2003); Ward's Triangle (Bemben 2000; Brentano 2008; Chilibeck 2002; Englund 2005; Lau 1992; Pruitt 1996; Sakai 2010; Smidt 1992); and total hip (Bemben 2000; Bergstrom 2008; Chilibeck 2002; Kerr 2001; Korpelainen 2006; Maddalozzo 2007; Newstead 2004; Pruitt 1996; Tolomio 2009; Verschueren 2004; Von Stengel 2009).

BMD was also measured at the distal radius (Korpelainen 2006; Preisinger 1995; Rubin 2004; Uusi‐Rasi 2003); forearm (Kerr 2001; Martin 1993; Prince 1991); tibia (Chan 2004; Cheng 2002); ankle (Prince 1995); and total body (Bemben 2000; Chilibeck 2002; Chubak 2006; Englund 2005; Going 2003; Kerr 2001; Newstead 2004; Verschueren 2004).

Other outcome measures included: BMC (Englund 2005; Karinkanta 2007; Nelson 1994; Uusi‐Rasi 2003); cortical bone density (Cheng 2002; Karinkanta 2007); trabecular bone density (Russo 2003); CaBI (Chow 1987); body mass (Martin 1993); muscle strength (Metcalfe 2001); and rate of falls (Von Stengel 2009).

Other adverse events were reported by 11 studies (Chow 1987; Ebrahim 1997; Grove 1992; Karinkanta 2007; Korpelainen 2006; Nelson 1994; Pruitt 1996; Revel 1993; Rubin 2004; Russo 2003; Uusi‐Rasi 2003) and included muscle soreness, joint pain, headache and itching.

Results of the search

From the search we found 936 references to potential studies. We discussed and resolved disagreements by reading the full text of the paper. We retrieved a total of 90 potential relevant trials for further classification (see PRISMA flow chart Figure 1).

1.

Study flow diagram.

Included studies

Forty‐three RCTs (27 new in this update) with 4320 participants met the inclusion criteria. On further scrutiny of the included studies from the original version (Mayoux‐Benhamou 1997) was actually follow‐up data for another included study (Revel 1993). Forty‐one of these 43 included studies were published in English, one in German (Von Stengel 2009) and one in Italian (Tolomio 2009). Three studies are awaiting classification and one trial is ongoing. We assessed all study designs as not affecting unit of analysis, for example we reported no cluster randomised trials or cross‐over trials.

Excluded studies

Thirty‐one studies did not meet the inclusion criteria as published in our a priori protocol. On further scrutiny of the included studies from the original version we excluded one study (Kerr 1996) because participants were their own control, with one side of body randomised to a different exercise type.

Risk of bias in included studies

Pairs of review authors judged the following key domains as 'low risk’, ‘high risk’ or ‘unclear risk of bias’:

• random sequence generation;

• allocation concealment;

• incomplete outcome data;

• selective reporting;

• blinding (participant);

• blinding (assessor); and

• "other bias" (comparability of treatment and control group at entry, and appropriateness of duration of surveillance).

In cases of disagreement between the review authors, we made a decision based on consensus. The methodological quality summary for each included study is presented in Figure 2 and the review authors' judgements about each methodological quality item are presented as percentages across all included studies in Figure 3. We assessed the overall risk of bias as 'low' for 13 studies (Bergstrom 2008; Bravo 1996; Cheng 2002; Chilibeck 2002; Chow 1987; Chubak 2006; Ebrahim 1997; Karinkanta 2007; Korpelainen 2006; Lau 1992; Lord 1996; Uusi‐Rasi 2003; Verschueren 2004) and 'high risk of bias' for four studies (Metcalfe 2001; Prince 1991; Prince 1995; Rubin 2004).

2.

Methodological quality summary: review authors' judgements about each methodological quality item for each included study.

3.

Methodological quality graph: review authors' judgements about each methodological quality item presented as percentages across all included studies.

Allocation

In the original version of the review only one study had clearly used a proper method of randomisation (Lau 1992). In this update 16 studies adequately described sequence generation ('low risk of bias') (Bergstrom 2008; Bravo 1996; Cheng 2002; Chow 1987; Chubak 2006; Ebrahim 1997; Karinkanta 2007; Kerr 2001; Korpelainen 2006; Lau 1992; Prince 1991; Prince 1995; Rubin 2004; Russo 2003; Sakai 2010; Verschueren 2004) and eleven studies adequately described allocation concealment ('low risk of bias') (Chilibeck 2002; Chow 1987; Chubak 2006; Ebrahim 1997; Karinkanta 2007; Korpelainen 2006; Lau 1992; Lord 1996; Prince 1991; Prince 1995; Rubin 2004). The other studies did not clearly describe these methods ('unclear risk of bias').

Blinding

It is difficult to ensure blinding of participants in studies of exercise interventions. Only one study adequately blinded participants for the type of exercise intervention ('low risk of bias') (Rubin 2004), there was an 'unclear risk of bias' in 15 studies (Bravo 1996; Brentano 2008; Cheng 2002; Chilibeck 2002; Chubak 2006; Ebrahim 1997; Going 2003; Hatori 1993; Iwamoto 2001; Karinkanta 2007; Lau 1992; Lord 1996; Martin 1993; Nelson 1994; Uusi‐Rasi 2003) and participants were not blinded in 27 studies ('high risk of bias').

Fourteen studies adequately blinded assessors to type of exercise intervention ('low risk of bias') (Bocalini 2009; Bravo 1996; Chilibeck 2002; Chow 1987; Chubak 2006; Hatori 1993; Korpelainen 2006; Papaioannou 2003; Preisinger 1995; Revel 1993; Rubin 2004; Sinaki 1989; Uusi‐Rasi 2003; Verschueren 2004), there was an 'unclear risk of bias' in 28 studies and assessors were not blinded in one study ('high risk of bias') (Ebrahim 1997).

Incomplete outcome data

Fifteen studies were judged as appropriately addressing incomplete outcome data ('low risk of bias') (Bergstrom 2008; Bravo 1996; Chow 1987; Chubak 2006; Ebrahim 1997; Going 2003; Karinkanta 2007; Korpelainen 2006; Lord 1996; Metcalfe 2001; Nelson 1994; Sinaki 1989; Tolomio 2009; Uusi‐Rasi 2003; Von Stengel 2009), there was an 'unclear risk of bias' for 26 studies (Bocalini 2009; Brentano 2008; Chan 2004; Cheng 2002; Chilibeck 2002; Chuin 2009; Englund 2005; Grove 1992; Iwamoto 2001; Iwamoto 2005; Kerr 2001; Lau 1992; Maddalozzo 2007; Martin 1993; Newstead 2004; Papaioannou 2003; Preisinger 1995; Prince 1991; Prince 1995; Pruitt 1996; Revel 1993; Rubin 2004; Russo 2003; Sakai 2010; Smidt 1992; Verschueren 2004) and two studies were judged as not addressing incomplete outcome data ('high risk of bias') (Bemben 2000; Hatori 1993).

Selective reporting

Insufficient information was available to permit judgement of 'low risk’ or ‘high risk of bias’ for selective reporting for any of the 43 studies.

Other potential sources of bias

Treatment and control groups were comparable at entry in 35 studies ('low risk of bias'), it was unclear in six studies ('unclear risk of bias') (Bergstrom 2008; Brentano 2008; Grove 1992; Metcalfe 2001; Newstead 2004; Sinaki 1989) and significant differences were present in two studies indicating a 'high risk of bias' (Prince 1991; Rubin 2004). Most studies only provided data at the end of the intervention. Only eight studies provided follow‐up data (Englund 2005; Karinkanta 2007; Korpelainen 2006; Metcalfe 2001; Preisinger 1995; Prince 1991; Revel 1993; Uusi‐Rasi 2003).

Effects of interventions

See: Table 1

1. All exercise types versus control (Analyses 1.1 to 1.8)

Thirty‐one studies examining exercise programmes versus control reported data for the seven outcomes selected as important to decision making:

• total number of fractures;

• percentage change in BMD at the spine;

• percentage change in BMD at the femoral neck;

• percentage change in BMD in total hip;

• percentage change in BMD at the trochanter;

• adverse events (falls); and

• other adverse events (muscle soreness, joint pain, headache and itching).

These results are reported in Table 1. However, the nature of the exercise programmes were heterogeneous and thus the results should be considered with caution. Among these studies 13 were considered to have 'low risk of bias'.

Meta‐analyses revealed significant differences between the exercise and control groups in favour of exercise for percentage change in BMD at the spine (MD 0.85; 95% CI 0.62 to 1.07), 24 studies and 1441 participants (Analysis 1.2); and trochanter (MD 1.03; 95% CI 0.56 to 1.49), 10 studies and 815 participants (Analysis 1.6).

1.2. Analysis.

Comparison 1 Any exercise versus control, Outcome 2 Bone mineral density % change: spine.

1.6. Analysis.

Comparison 1 Any exercise versus control, Outcome 6 Bone mineral density % change: trochanter.

The risk of fracture in exercise groups was not significantly different than that in controls (OR 0.61; 95% CI 0.23 to 1.64), four studies and 539 participants. There was no significant difference between the exercise and control groups for: percentage change in BMD at the femoral neck (MD ‐0.08; 95% CI ‐1.08 to 0.92), 19 studies and 1338 participants; or total hip (MD 0.41; 95% CI ‐0.64 to 1.45), 13 studies and 863 participants. Only one study (Uusi‐Rasi 2003), 76 participants, reported percentage change in BMC for the spine and femoral neck and no significant differences were observed between the exercise and control groups.

Three studies (Chan 2004; Ebrahim 1997; Karinkanta 2007) (n = 378) reported falls as adverse events. There were 75 falls reported in the exercise groups and 55 in the control groups. Other adverse events were reported in 11 studies (Chow 1987; Ebrahim 1997; Grove 1992; Karinkanta 2007; Korpelainen 2006; Nelson 1994; Pruitt 1996; Revel 1993; Rubin 2004; Russo 2003; Uusi‐Rasi 2003); 60 events were reported in the exercise groups and five in the control groups, and included muscle soreness, joint pain, headache and itching.

2. Static weight bearing (SWB) (Analysis 2.1)

Only one study Sakai 2010 involving 31 participants examined SWB (standing on one leg for three minutes per day). A significant difference was reported for percentage change in BMD at the hip (Analysis 2.1).

2.1. Analysis.

Comparison 2 Static weight bearing exercise versus control, Outcome 1 Bone mineral density % change: hip.

3. Dynamic weight bearing exercise low force (DWBLF) (Analyses 3.1 to 3.7)

Nine studies examining DWBLF exercise (including walking and Tai chi) reported data on 705 participants. Among these studies five were considered to have 'low risk of bias' (Bravo 1996; Chow 1987; Ebrahim 1997; Lau 1992; Lord 1996). The compliance with exercise programmes, when reported, varied from 39% (Prince 1995) to 79.2% (Martin 1993).

The results of meta‐analysis showed that there was a statistically significant effect on percentage change in BMD of the spine (MD 0.87; 95% CI 0.26 to 1.48), seven studies and 1119 participants (Analysis 3.1), in favour of exercise. However there was no effect on the femoral neck (MD ‐1.20; 95% CI ‐4.45 to 2.05), five studies and 585 participants; trochanter (MD 0.39; 95% CI ‐0.59 to 1.38), two studies and 241 participants; or number of fractures (OR 0.92; 95% CI 0.21 to 3.96), two studies and 229 participants.

3.1. Analysis.

Comparison 3 Dynamic weight bearing exercise low force versus control, Outcome 1 Bone mineral density % change: spine.

Results of single studies indicate a significant difference in favour of exercise for percentage change of mean regression slope in BMD at the wrist (MD 1.40; 95% CI 0.85 to 1.95), 103 participants (Preisinger 1995), and in favour of the control group for percentage change in BMD in Ward's triangle (MD ‐3.60; 95% CI ‐5.48 to ‐1.72), 23 participants (Lau 1992).

4. Dynamic weight bearing exercise high force (DWBHF) (Analyses 4.1 to 4.16)

Ten studies examining DWBHF exercise (including jogging, jumping, running, dancing and vibration platform) reported data on 568 participants. Among these studies four were considered to have 'low risk of bias' (Cheng 2002; Karinkanta 2007; Uusi‐Rasi 2003; Verschueren 2004). The compliance with exercise programmes, when reported, varied from 82.6% (Grove 1992) to 86.2% (Maddalozzo 2007).

The results of meta‐analysis showed that there was a statistically significant effect on percentage change in BMD of the hip (MD 1.55; 95% CI 1.41 to 1.69), four studies and 179 participants (Analysis 4.2); and trochanter (MD 1.23; 95% CI ‐0.01 to 2.47), two studies and 188 participants (Analysis 4.6) in favour of exercise.

4.2. Analysis.

Comparison 4 Dynamic weight bearing exercise high force versus control, Outcome 2 Bone mineral density % change: hip.

4.6. Analysis.

Comparison 4 Dynamic weight bearing exercise high force versus control, Outcome 6 Bone mineral density % change: trochanter.

There was no effect on the percentage change in BMD of the spine (MD ‐1.20; 95% CI ‐4.45 to 2.05); mid femur (MD 0.12; 95% CI ‐4.84 to 5.08); proximal tibia (MD 3.31; 95% CI ‐20.22 to 26.84); femoral neck (MD 1.06; 95% CI ‐0.32 to 2.45); or on CaBI at the trunk and upper thighs (MD 5.30; 95% CI ‐7.50 to 18.10).

Results of a single study (Uusi‐Rasi 2003) with 76 participants did not indicate any effect on percentage change in BMC at the spine (MD 1.43; 95% CI ‐9.18 to 12.04); femoral neck (MD 0.00; 95% CI ‐9.11 to 9.11); or wrist (MD ‐3.41; 95% CI ‐15.64 to 8.82).

5. Non‐weight bearing exercise low force (NWBLF) (Analyses 5.1 to 5.6)

Five of six studies examining NWBLF exercise (e.g. low load, high repetition strength training) reported data on 231 participants. Among these studies none were considered to have 'low risk of bias'. The compliance with exercise programmes, when reported, varied from 65% (Pruitt 1996) to 90% (Kerr 2001). No significant differences were observed for any outcome.

6. Non‐weight bearing exercise high force (NWBHF) (Analayses 6.1 to 6.6)

Nine studies examining NWBHF exercise (e.g. progressive resisted strengthening exercise) reported data on 292 participants. Among these studies one was considered to have 'low risk of bias' (Chilibeck 2002). The compliance with exercise programmes, when reported, varied from 65% (Pruitt 1996) to 92% (Kerr 2001).

The results of meta‐analysis showed that there was a statistically significant effect on percentage change in BMD of the spine (MD 0.86; 95% CI 0.58 to 1.13), eight studies and 246 participants (Analysis 6.1); and neck of femur (MD 1.03; 95% CI 0.24 to 1.82), eight studies and 247 participants (Analysis 6.3). No significant differences were observed for any other outcome.

6.1. Analysis.

Comparison 6 Non‐weight bearing exercise high force versus control, Outcome 1 Bone mineral density % change: spine.

6.3. Analysis.

Comparison 6 Non‐weight bearing exercise high force versus control, Outcome 3 Bone mineral density % change: femoral neck.

7. Combination (COMB) (Analyses 7.1 to 7.9)

Ten studies examining combinations of exercise types (more than one of the above exercise interventions) reported data on 823 participants. Among these studies five were considered to have 'low risk of bias' (Bergstrom 2008; Chow 1987; Chubak 2006; Karinkanta 2007; Korpelainen 2006). The compliance with exercise programmes, when reported, varied from 62% (Papaioannou 2003) to 95% (Bergstrom 2008).

The results of meta‐analysis showed that the risk of fractures in exercise groups was significantly lower than that in controls (OR 0.33; 95% CI 0.13 to 0.85), two studies and 236 participants (Analysis 7.9): in percentage change in BMD of the spine (MD 3.22; 95% CI 1.80 to 4.64), four studies and 258 participants (Analysis 7.1); trochanter (MD 1.31; 95% CI 0.69 to 1.92), two studies and 200 participants (Analysis 7.3); and neck of femur (MD 0.45; 95% CI 0.08 to 0.82), three studies and 325 participants (Analysis 7.6).

7.9. Analysis.

Comparison 7 Combination versus control, Outcome 9 Fractures.

7.1. Analysis.

Comparison 7 Combination versus control, Outcome 1 Bone mineral density % change: spine.

7.3. Analysis.

Comparison 7 Combination versus control, Outcome 3 Bone mineral density % change: trochanter.

7.6. Analysis.

Comparison 7 Combination versus control, Outcome 6 Bone mineral density % change: neck of femur.

However the results of meta‐analysis showed that there was a statistically significant effect in favour of control in percentage change in BMD of the total hip (MD ‐1.07; 95% CI ‐1.58 to ‐0.56), four studies and 468 participants (Analysis 7.2).

7.2. Analysis.

Comparison 7 Combination versus control, Outcome 2 Bone mineral density % change: total hip.

8. Exercise and pharmacological products versus control and pharmacological products (Analyses 8.1 to 15.2)

Ten studies examining exercise and pharmacological products versus control and pharmacological products reported data on 598 participants. Among these studies four were considered to have 'low risk of bias'. The pharmacological products were hormone replacement therapy (HRT) (Cheng 2002; Maddalozzo 2007; Going 2003), bisphosphonates (Chilibeck 2002; Iwamoto 2005; Uusi‐Rasi 2003), antioxidants (Chuin 2009), calcium (Kerr 2001; Lau 1992; Prince 1995) and calcium plus vitamin D (Martin 1993). Exercise types included DWBHF (Cheng 2002; Going 2003; Iwamoto 2005; Maddalozzo 2007; Uusi‐Rasi 2003), DWBLF and NWBLF (Kerr 2001), and NWBHF (Chilibeck 2002; Chuin 2009; Kerr 2001). Compliance with exercise programmes, when reported, varied from 39% (Prince 1995) to 92% (Kerr 2001).

On the whole data for these eight comparisons comprised small single studies. DWBHF plus HRT versus HRT demonstrated a significant effect in favour of exercise in percentage change in BMD at the trochanter (MD 1.86; 95% CI 0.60 to 3.13), 2 studies and 203 participants; spine NWBHF plus bisphosphonates versus bisphosphonates (Chilibeck 2002), 26 participants; distal tibia and Ward's triangle DWBLF plus calcium versus calcium (MD 0.60; 95% CI 0.46 to 0.74) (Prince 1995) and (MD 14.50; 95% CI 10.05 to 18.95) (Lau 1992). However a significant difference in favour of calcium was seen for DWBLF plus calcium versus calcium (MD ‐1.02; 95% CI ‐1.36 to ‐0.68) (Lau 1992).

No significant differences were observed for any outcome for any exercise type and pharmacological products versus control and pharmacological products as listed above.

Discussion

Summary of main results

We have summarised the data from 43 RCTs comparing exercise with usual activity and exercise plus pharmacological products versus pharmacological products. We also separately examined the effect of different categories of exercise.

Our results suggest a relatively small statistically significant, but possibly important, effect of exercise on bone density in postmenopausal women compared with control groups. The risk of fracture in exercise groups was not significantly different than that in controls (OR 0.61; 95% CI 0.23 to 1.64). Our inference is strengthened by the consistency of significant findings in favour of exercise for percentage change in BMD across three sites; spine (MD 0.85; 95% CI 0.62 to 1.07); total hip (MD 0.41; 95% CI ‐0.64 to 1.45); and trochanter (MD 1.03; 95% CI 0.56 to 1.49). This inference is, however, weakened by methodological limitations such as small sample sizes, loss to follow‐up in most studies and by the unexplained heterogeneity of results across studies. Few studies provided follow‐up data to determine the effect of exercise beyond the end of the intervention. Falls were reported as adverse events as numbers of falls rather than number of fallers; one study reported more events than people, indicating repeat falls. Other adverse events were reported for participants mainly in intervention groups and included muscle soreness, joint pain, headache and itching. There was generally no mention of event monitoring in control groups.

COMB exercise programmes (comprising more than one exercise type) had a significant effect on BMD at three sites: neck of femur; spine; and trochanter. The risk of fracture in exercise groups was lower than that of controls. However there was a significant difference in BMD in total hip measurement in favour of the control group.

DWBHF exercise (jogging, jumping and vibration platforms) had a significant effect on BMD at two sites; total hip and trochanter, compared to the control groups, but no effect at any other site (neck of femur, spine, mid femur, tibia, trunk and thighs, or BMC at neck of femur and spine).

DWBLF exercise (walking or Tai chi) had a significant effect on BMD at two sites; spine and wrist compared to the control groups.

NWBHF exercise (progressive resistance exercise) had a significant effect on BMD at two sites; neck of femur and spine compared to the control groups.

SWB exercise (e.g. single leg standing) had a significant effect on BMD at the hip.

NWBLF exercise (low load high repetition strength training) had no significant difference on any outcomes reported.

On the whole, comparisons of exercise plus pharmacological products versus pharmacological products were small single studies and thus conclusions on their findings are limited.

The population included in these studies was heterogeneous but results should be reviewed with caution due to differences in ethnicity. However, the lack of reporting of exercise characteristics (type, intensity, frequency, duration and mode) of the study exercise interventions also limits the conclusions that can be drawn from this review.

Overall completeness and applicability of evidence

It is important for future research in this area to have standardised recommendations for conducting exercise interventions and reporting of exercise outcomes. We cannot currently determine if the effect of the varied exercises was different in the first and in the second period of the postmenopausal time period. The short time of the follow‐up of all the studies limits our ability to predict the long‐term effects exercise may or may not have on bone loss.

Quality of the evidence

The 43 studies (27 new studies in this update) included in this review were predominantly in the English language and originate mainly from North America and Europe (n = 32). Whilst this may be seen to limit the applicability of the evidence to these healthcare systems and social environments the evidence has potential generalisability; the majority of participants were healthy community‐dwelling women.

The overall quality of the included studies was variable and should be taken into account when interpreting the results of this review. The risk of bias was assessed as 'low' for only 13 studies (Bergstrom 2008; Bravo 1996; Cheng 2002; Chilibeck 2002; Chow 1987; Chubak 2006; Ebrahim 1997; Karinkanta 2007; Korpelainen 2006; Lau 1992; Lord 1996; Uusi‐Rasi 2003; Verschueren 2004). Only one study included in this review blinded the patients (Rubin 2004) and few blinded the assessors (Bocalini 2009; Bravo 1996; Chilibeck 2002; Chow 1987; Chubak 2006; Hatori 1993; Korpelainen 2006; Papaioannou 2003; Preisinger 1995; Revel 1993; Rubin 2004; Sinaki 1989; Uusi‐Rasi 2003; Verschueren 2004).

It is very difficult and maybe impossible to blind patients and care providers in exercise therapy. Nevertheless, this bias is unlikely to have influenced the BMD measurements.

Potential biases in the review process

The review was restricted to RCTs; we excluded clinical controlled trials CCTs) thus limiting the potential for bias. All studies described themselves as randomised mostly without giving details of how the randomisation sequence was generated and what precautions were taken in relation to concealment of allocation. Only eight studies adequately deal with incomplete outcome data (Englund 2005; Karinkanta 2007; Korpelainen 2006; Metcalfe 2001; Preisinger 1995; Prince 1991; Revel 1993; Uusi‐Rasi 2003) the remainder reporting the results for only those participants who completed all post‐treatment assessments. Fractures, falls and other adverse events were reported as adverse events and were mainly monitored for the exercise groups only.

Agreements and disagreements with other studies or reviews

Few meta‐analyses have been published in this area (Berard 1997; Hind 2007; Wolff 1999). These meta‐analyses included RCTs and CCTs, one (Wolff 1999) also included studies on premenopausal women. All authors subdivided the studies depending on the exercise characteristics and meta‐analysed all exercise programmes together. While Wolff 1999 concluded that exercises prevent femoral and lumbar bone loss, Berard 1997 concluded that weight bearing exercises are effective mainly on the spine, but there may be some evidence of efficacy on femoral bone and forearm bone. Weight bearing exercise also appears to enhance bone mineral accrual in children, particularly during early puberty (Hind 2007). The results of a meta‐analysis in Nikander 2010 indicate that exercise can significantly enhance bone strength at loaded sites in children but not in adults.

Other reviews with limited quality evidence suggest that Tai chi (Wayne 2007) and physical activity (Schmitt 2009) may be effective and safe methods of maintaining BMD in postmenopausal women. Furthermore an overview of the literature (Winett 2001) purports the benefits of resistance training on BMD. The results of this current review suggest that high force resistance training increases BMD whereas low force does not, indicating the importance of intensity of exercise.

This current review reported on bone mass as an outcome of effectiveness and included BMC, CaBI and BMD. Zebaze 2010 reported that most bone loss is cortical, not trabecular, and occurs after 65 years of age, and the resulting structural decay, including the magnitude of intracortical remodelling and intracortical porosity are poorly captured by current measurement methods. Nikander 2010 recommends that further research is required to quantify the effects of exercise on whole bone strength and its structural determinants throughout life.

Authors' conclusions

Implications for practice.

Our results suggest a relatively small statistically significant, but possibly important, effect of exercise on bone density in postmenopausal women compared with control groups. The most effective type of exercise intervention on BMD for the neck of femur appears to be NWBHF exercise such as progressive resistance strength training for the lower limbs. The most effective intervention for BMD at the spine was COMB exercise programmes (comprising more than one exercise type) (MD 3.22; 95% CI 1.80 to 4.64), with a change of over 3% compared with control groups. However the risk of fracture across all exercise groups was not significantly different than that in controls with 4 more women out of 100 who did not exercise sustaining a fracture (absolute difference 4%).

These exercise types should be considered as preferred interventions in clinical practice, however, it remains unclear as to what constitutes an optimal exercise programme.

Implications for research.

On the whole the quality of the reporting of studies in the meta‐analyses in this review was low, in particular in the areas of sequence generation, allocation concealment, blinding and loss to follow‐up. Future research needs to focus on standardised outcome measures and exercise programmes, better reporting of all the parameters of exercise programmes and, the accuracy of measurements. Fracture and fall data, and adverse event reporting for all participants regardless of group allocation should be included for all future studies. Measuring the BMD changes for both the hip (particularly neck of femur) and spine is important as fractures at these sites result in high mortality and morbidity. Adequate follow‐up of participants is required to determine long‐term effects of exercise.

What's new

Date	Event	Description
17 May 2011	New search has been performed	The methodology has been updated to include risk of bias and summary of findings tables. The exercise interventions have been re‐categorised which included reclassification of interventions from the original studies. The search was updated to include vibrations plates and included all studies up to Jan 2011. Of the 43 included studies 27 are new studies; and on further scrutiny of the included studies from the original version one was excluded and a further study was actually follow‐up data for another included study. New comparisons are based on the reclassification of exercise categories. All analyses are new, based on the new categories of exercise interventions and the conclusions have changed.
17 May 2011	New citation required and conclusions have changed	This review update involves new authors and conclusions have changed

History

Protocol first published: Issue 3, 1997
 Review first published: Issue 3, 2002

Date	Event	Description
27 September 2008	Amended	Converted to new review format. CMSG ID: C035‐R

Appendices

Appendix 1. MEDLINE search strategy

MEDLINE (searched 10/12/2010)

 

1              osteoporosis/

2              osteoporo$.mp.

3              osteopenia.mp.

4              bone density/

5              bone densit$.mp.

6              exp "bone and bones"/

7              bone loss$.mp.

8              bone mass$.mp.

9              bone mineral densit$.mp.

10            bone mineral content$.mp.

11            bone age.mp.

12            bone defect$.mp.

13            bone deminerali?ation.mp.

14            bone mineral$.mp.

15            bone strength.mp.

16            decalcifi$.mp.

17            deminerali?ed bone.mp.

18            or/1‐17

19            randomized controlled trial.pt.

20            controlled clinical trial.pt.

21            randomi?ed.ab.

22            placebo.ab.                     

23            clinical trials as topic.sh.

24            randomly.ab.

25            trial.ti.

26            or/19‐25

27            (animals not (humans and animals)).sh.

28            26 not 27

29            exp exercise/

30            exp exercise therapy/

31            exerci$.mp.

32            exp sports/

33            sport$.mp.

34            physical fitness/

35            physical fitness.mp.

36            physical activit$.mp.

37            vibration/tu

38            vibration therap$.mp.

39            or/29‐38

40            exp menopause/

41            (menopaus$ or postmenopaus$ or post‐menopaus$ or (post adj menopause$)).mp.

42            or/40‐41

43            18 and 28 and 39 and 42

44            limit 43 to yr="2000‐Current"

Appendix 2. EMBASE search strategy

EMBASE (searched 10/12/2010)

1              osteoporosis/

2              osteoporo$.mp.

3              osteopenia.mp.

4              bone density/

5              bone mass/

6              bone densit$.mp.

7              exp bone/

8              bone loss$.mp.

9              bone mass$.mp.

10            bone mineral/

11            bone mineral densit$.mp.

12            bone mineral content$.mp.

13            or/1‐12

14            Randomized Controlled Trial/

15            Double Blind Procedure/

16            Single Blind Procedure/

17            Triple Blind Procedure/

18            randomi?ed.ti,ab.

19            randomisation/

20            Placebo/

21            placebo$.mp.

22            ((controlled or comparative or placebo or randomi?ed) adj3 (trial or study)).mp.

23            (random$ adj7 (allocat$ or allot$ or assign$ or basis$ or divid$ or order$)).mp.

24            ((singl$ or doubl$ or trebl$ or tripl$) adj7 (blind$ or mask$)).mp.

25            or/14‐24

26            limit 25 to human

27            exp exercise/

28            exp sport/

29            fitness/

30            exp physical activity/

31            sport$.mp.

32            exercis$.mp.

33            physical fitness.mp.

34            physical activit$.mp.

35            vibration therapy/

36            vibration therap$.mp.

37            or/27‐36

38            postmenopause/

39            postmenopause osteoporosis/

40            menopause/

41            (menopaus$ or postmenopaus$ or post‐menopaus$ or (post adj menopause$)).mp.

42            or/38‐41

43            42 and 37 and 26 and 13

44            limit 43 to yr="2000 ‐Current"

Appendix 3. CINAHL search strategy

CINAHL (Searched 17/12/2010)

1              (MH "Osteoporosis+")

2              TI osteoporosis OR AB osteoporosis

3              TI osteopenia OR AB osteopenia

4              MH bone density

5              TI bone densit* OR AB bone densit*

6              (MH "Bone and Bones+")

7              TI bone loss* OR AB bone loss*

8              TI bone mass* OR AB bone mass*

9              TI bone mineral densit* OR AB bone mineral densit*

10            TI bone mineral content* OR AB bone mineral content*

11            TI bone age OR AB bone age

12            TI bone defect* OR AB bone defect*

13            TI bone deminerali?ation OR AB bone deminerali?ation

14            TI bone mineral* OR AB bone mineral*

15            TI bone strength OR AB bone strength

16            TI decalcifi* OR AB decalcifi*

17            TI deminerali?ed bone OR AB deminerali?ed bone

18            S1 or S2 or S3 or S4 or S5 or S6 or S7 or S8 or S9 or S10 or S11 or S12 or S13 or S14 or S15 or S16 or S17

19            (MH “double‐blind studies”)

20            (MH “single‐blind studies”)

21            (MH “triple‐blind studies”)

22            TI randomi?ed OR AB randomi?ed

23            (MH “random assignment”)

24            (MH “crossover design”)

25            (MH “placebos”)

26            TI placebo* OR AB placebo*

27            TI controlled N3 trial OR TI controlled N3 study OR TI comparative N3 trial OR TI comparative N3 study OR TI placebo N3 trial OR TI placebo N3 study OR TI randomi?ed N3 trial OR TI randomi?ed N3 study

28            AB controlled N3 trial OR AB controlled N3 study OR AB comparative N3 trial OR AB comparative N3 study OR AB placebo N3 trial OR AB placebo N3 study OR AB randomi?ed N3 trial OR AB randomi?ed N3 study

29            TI random* N7 allocat* OR TI random* N7 allot* OR TI random* N7 assign* OR TI random* N7 basis* OR TI random* N7 divid* OR TI random* N7 order*

30            AB random* N7 allocat* OR AB random* N7 allot* OR AB random* N7 assign* OR AB random* N7 basis* OR AB random* N7 divid* OR AB random* N7 order*

31            TI singl* N7 blind* OR TI doubl* N7 blind* OR TI trebl* N7 blind* OR TI tripl* N7 blind* OR TI singl* N7 mask* OR TI doubl* N7 mask* OR TI trebl* N7 mask* OR TI tripl* N7 mask*

32            AB singl* N7 blind* OR AB doubl* N7 blind* OR AB trebl* N7 blind* OR AB tripl* N7 blind* OR AB singl* N7 mask* OR AB doubl* N7 mask* OR AB trebl* N7 mask* OR AB tripl* N7 mask*

33            S19 or S20 or S21 or S22 or S23 or S24 or S25 or S26 or S27 or S28 or S29 or S30 or S31 or S32

34            (MH “animals”)

35            33 not 34

36            (MH “exercise+”)

37            TI exerci* OR AB exerci*

38            (MH “sports+”)

39            TI sport* OR AB sport*

40            MH “physical fitness+”

41            MH physical activity

42            TI physical fitness OR AB physical fitness

43            TI physical activit* OR AB physical activit*

44            MH vibration/tu

45            TI vibration therap* OR AB vibration therap*

46            S36 or S37 or S38 or S39 or S40 or S41 or S42 or S43 or S44 or S45

47            MH menopause

48            MH postmenopause

49            TI menopaus* OR AB menopaus* OR TI post‐menopaus* OR AB post‐menopaus* OR TI postmenopaus* OR AB postmenopaus* OR TI post N1 menopaus* OR AB post N1 menopaus* 

50            S47 OR S48 OR S49

51            S18 AND S35 AND S46 AND S50 

52            S18 AND S35 AND S46 AND S50  Limiters ‐ Published Date from: 20000101‐20101231

Appendix 4. PEDro keyword and text word search strategy

PEDro (Searched December 15, 2010)

Search 1

Osteoporo* or bone in Abstract & Title AND

Fitness training in Therapy

Search 2

Osteoporo* or bone in Abstract & Title AND

Strength training in Therapy

Appendix 5. CCTR search strategy

CCTR (Searched 10/12/2010)

1              osteoporosis/

2              osteoporo$.mp.

3              osteopenia.mp.

4              bone density/

5              bone densit$.mp.

6              exp "bone and bones"/

7              bone loss$.mp.

8              bone mass$.mp.

9              bone mineral densit$.mp.

10            bone mineral content$.mp.

11            bone age.mp.

12            bone defect$.mp.

13            bone deminerali?ation.mp.

14            bone mineral$.mp.

15            bone strength.mp.

16            decalcifi$.mp.

17            deminerali?ed bone.mp.

18            or/1‐17

19            randomized controlled trial.pt.

20            controlled clinical trial.pt.

21            randomi?ed.ab.

22            placebo.ab.                     

23            clinical trials as topic.sh.

24            randomly.ab.

25            trial.ti.

26            or/19‐25

27            animals/

28            26 not 27

29            exp exercise/

30            exp exercise therapy/

31            exerci$.mp.

32            exp sports/

33            sport$.mp.

34            physical fitness/

35            physical fitness.mp.

36            physical activit$.mp.

37            vibration/tu

38            vibration therap$.mp.

39            or/29‐38

40            exp menopause/

41            (menopaus$ or postmenopaus$ or post‐menopaus$ or (post adj menopause$)).mp.

42            or/40‐41

43            18 and 28 and 39 and 42

44            limit 43 to yr="2000‐Current"

Appendix 6. AMED search strategy

AMED (Searched 10/12/2010)

1              osteoporosis/

2              osteoporo$.mp.

3              osteopenia.mp.

4              bone density/

5              bone densit$.mp.

6              exp bones/

7              bone loss$.mp.

8              bone mass$.mp.

9              bone mineral densit$.mp.

10            bone mineral content$.mp.

11            or/1‐10

12            randomized controlled trial.pt.

13            controlled clinical trial.pt.

14            randomi?ed.ti,ab.

15            Placebos/

16            placebo$.ti,ab.

17            double‐blind method/

18            single‐blind method/

19            random allocation/

20            ((controlled or comparative or placebo or randomi?ed) adj3 (trial or study)).mp.

21            (random$ adj7 (allocat$ or allot$ or assign$ or basis$ or divid$ or order$)).mp.

22            ((singl$ or doubl$ or trebl$ or tripl$) adj7 (blind$ or mask$)).mp.

23            or/12‐22

24            animals/

25            23 not 24

26            exp exercise/

27            exp Sports/

28            Physical Fitness/

29            sport$.mp.

30            exerci$.mp.

31            physical fitness.mp.

32            physical activit$.mp.

33            vibration/

34            vibration therapy.mp.

35            or/26‐34

36            exp menopause/

37            (menopaus$ or postmenopaus$ or post‐menopaus$ or (post adj menopause$)).mp.

38            or/36‐37

39            11 and 25 and 35 and 38

40            limit 39 to yr="2000‐Current"

Data and analyses

Comparison 1. Any exercise versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Total number of fractures	4	539	Odds Ratio (M‐H, Random, 95% CI)	0.61 [0.23, 1.64]
2 Bone mineral density % change: spine	24	1441	Mean Difference (IV, Fixed, 95% CI)	0.85 [0.62, 1.07]
3 Bone mineral density % change: femoral neck	19	1338	Mean Difference (IV, Random, 95% CI)	‐0.08 [‐1.08, 0.92]
4 Bone mineral density % change: Ward's triangle	6	185	Mean Difference (IV, Fixed, 95% CI)	‐2.67 [‐4.06, ‐1.28]
5 Bone mineral density % change: hip	13	863	Mean Difference (IV, Random, 95% CI)	0.41 [‐0.64, 1.45]
6 Bone mineral density % change: trochanter	10	815	Mean Difference (IV, Fixed, 95% CI)	1.03 [0.56, 1.49]
7 Bone mineral content % change: spine	1		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
7.1 Immediately postintervention	1	76	Mean Difference (IV, Fixed, 95% CI)	1.43 [‐9.18, 12.04]
7.2 Follow‐up at 15 months post intervention	1	50	Mean Difference (IV, Fixed, 95% CI)	2.44 [‐8.96, 13.84]
8 Bone mineral content % change: femoral neck	1		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
8.1 Immediately postintervention	1	76	Mean Difference (IV, Fixed, 95% CI)	0.0 [‐9.11, 9.11]
8.2 Follow‐up at 15 months postintervention	1	51	Mean Difference (IV, Fixed, 95% CI)	2.98 [‐7.41, 13.37]

1.1. Analysis.

Comparison 1 Any exercise versus control, Outcome 1 Total number of fractures.

1.3. Analysis.

Comparison 1 Any exercise versus control, Outcome 3 Bone mineral density % change: femoral neck.

1.4. Analysis.

Comparison 1 Any exercise versus control, Outcome 4 Bone mineral density % change: Ward's triangle.

1.5. Analysis.

Comparison 1 Any exercise versus control, Outcome 5 Bone mineral density % change: hip.

1.7. Analysis.

Comparison 1 Any exercise versus control, Outcome 7 Bone mineral content % change: spine.

1.8. Analysis.

Comparison 1 Any exercise versus control, Outcome 8 Bone mineral content % change: femoral neck.

Comparison 2. Static weight bearing exercise versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral density % change: hip	1	31	Mean Difference (IV, Fixed, 95% CI)	2.42 [0.73, 4.10]

Comparison 3. Dynamic weight bearing exercise low force versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral density % change: spine	7	519	Mean Difference (IV, Fixed, 95% CI)	0.87 [0.26, 1.48]
2 Bone mineral density % change: femoral neck	5	485	Mean Difference (IV, Random, 95% CI)	‐1.20 [‐4.45, 2.05]
3 Bone mineral density % change: trochanter	2	241	Mean Difference (IV, Fixed, 95% CI)	0.39 [‐0.59, 1.38]
4 Bone mineral density % change: Ward's triangle	1	23	Mean Difference (IV, Fixed, 95% CI)	‐3.6 [‐5.48, ‐1.72]
5 Bone mineral density % change: wrist	1	83	Mean Difference (IV, Fixed, 95% CI)	0.10 [‐1.30, 1.50]
6 Bone mineral density mean regression slope % change: wrist	1	103	Mean Difference (IV, Fixed, 95% CI)	1.4 [0.85, 1.95]
7 Fractures	2		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only
7.1 Vertebral year 1	2	229	Odds Ratio (M‐H, Random, 95% CI)	0.54 [0.11, 2.65]
7.2 Vertebral year 2	1	97	Odds Ratio (M‐H, Random, 95% CI)	4.18 [0.45, 38.82]
7.3 Total number of fractures	2	229	Odds Ratio (M‐H, Random, 95% CI)	0.92 [0.21, 3.96]

3.2. Analysis.

Comparison 3 Dynamic weight bearing exercise low force versus control, Outcome 2 Bone mineral density % change: femoral neck.

3.3. Analysis.

Comparison 3 Dynamic weight bearing exercise low force versus control, Outcome 3 Bone mineral density % change: trochanter.

3.4. Analysis.

Comparison 3 Dynamic weight bearing exercise low force versus control, Outcome 4 Bone mineral density % change: Ward's triangle.

3.5. Analysis.

Comparison 3 Dynamic weight bearing exercise low force versus control, Outcome 5 Bone mineral density % change: wrist.

3.6. Analysis.

Comparison 3 Dynamic weight bearing exercise low force versus control, Outcome 6 Bone mineral density mean regression slope % change: wrist.

3.7. Analysis.

Comparison 3 Dynamic weight bearing exercise low force versus control, Outcome 7 Fractures.

Comparison 4. Dynamic weight bearing exercise high force versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral density % change: spine	4	247	Mean Difference (IV, Fixed, 95% CI)	0.60 [‐0.23, 1.44]
2 Bone mineral density % change: hip	4	179	Mean Difference (IV, Fixed, 95% CI)	1.55 [1.41, 1.69]
3 Bone mineral density % change: mid femur	1	23	Mean Difference (IV, Fixed, 95% CI)	0.12 [‐4.84, 5.08]
4 Bone mineral density % change: proximal tibia	1	23	Mean Difference (IV, Fixed, 95% CI)	3.31 [‐20.22, 26.84]
5 Calcium bone index % change: trunk and upper thighs	1	32	Mean Difference (IV, Fixed, 95% CI)	5.3 [‐7.50, 18.10]
6 Bone mineral density % change: trochanter	2	188	Mean Difference (IV, Fixed, 95% CI)	1.23 [‐0.01, 2.47]
7 Bone mineral density % change: femoral neck	3	237	Mean Difference (IV, Fixed, 95% CI)	1.06 [‐0.32, 2.45]
8 Bone mineral content % change: spine	1		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
8.1 Immediately postintervention	1	76	Mean Difference (IV, Fixed, 95% CI)	1.43 [‐9.18, 12.04]
8.2 Follow‐up at 15 months postintervention	1	50	Mean Difference (IV, Fixed, 95% CI)	2.44 [‐8.96, 13.84]
9 Bone mineral content % change: femoral neck	1		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
9.1 Immediately postintervention	1	76	Mean Difference (IV, Fixed, 95% CI)	0.0 [‐9.11, 9.11]
9.2 Follow‐up at 15 months postintervention	1	51	Mean Difference (IV, Fixed, 95% CI)	2.98 [‐7.41, 13.37]
10 Bone mineral content % change: wrist	1		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
10.1 Immediately postintervention	1	76	Mean Difference (IV, Fixed, 95% CI)	‐3.41 [‐15.64, 8.82]
10.2 Follow‐up at 15 months postintervention	1	50	Mean Difference (IV, Fixed, 95% CI)	‐0.70 [‐14.96, 13.56]
11 Bone mineral content % change: ankle	1	76	Mean Difference (IV, Fixed, 95% CI)	2.07 [‐7.09, 11.23]
12 Bone mineral content % change: tibia	1	76	Mean Difference (IV, Fixed, 95% CI)	0.86 [‐6.22, 7.94]
13 Bone mineral density % change: total body	2	179	Mean Difference (IV, Fixed, 95% CI)	0.37 [‐0.00, 0.75]
14 Volumetric bone density % change: tibial trabecular	1	29	Mean Difference (IV, Fixed, 95% CI)	‐1.15 [‐20.30, 18.00]
15 Volumetric bone density % change: tibial cortical	1	29	Mean Difference (IV, Fixed, 95% CI)	0.49 [‐2.93, 3.91]
16 Fractures	1		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only
16.1 Total number of fractures	1	112	Odds Ratio (M‐H, Random, 95% CI)	1.56 [0.16, 15.56]

4.1. Analysis.

Comparison 4 Dynamic weight bearing exercise high force versus control, Outcome 1 Bone mineral density % change: spine.

4.3. Analysis.

Comparison 4 Dynamic weight bearing exercise high force versus control, Outcome 3 Bone mineral density % change: mid femur.

4.4. Analysis.

Comparison 4 Dynamic weight bearing exercise high force versus control, Outcome 4 Bone mineral density % change: proximal tibia.

4.5. Analysis.

Comparison 4 Dynamic weight bearing exercise high force versus control, Outcome 5 Calcium bone index % change: trunk and upper thighs.

4.7. Analysis.

Comparison 4 Dynamic weight bearing exercise high force versus control, Outcome 7 Bone mineral density % change: femoral neck.

4.8. Analysis.

Comparison 4 Dynamic weight bearing exercise high force versus control, Outcome 8 Bone mineral content % change: spine.

4.9. Analysis.

Comparison 4 Dynamic weight bearing exercise high force versus control, Outcome 9 Bone mineral content % change: femoral neck.

4.10. Analysis.

Comparison 4 Dynamic weight bearing exercise high force versus control, Outcome 10 Bone mineral content % change: wrist.

4.11. Analysis.

Comparison 4 Dynamic weight bearing exercise high force versus control, Outcome 11 Bone mineral content % change: ankle.

4.12. Analysis.

Comparison 4 Dynamic weight bearing exercise high force versus control, Outcome 12 Bone mineral content % change: tibia.

4.13. Analysis.

Comparison 4 Dynamic weight bearing exercise high force versus control, Outcome 13 Bone mineral density % change: total body.

4.14. Analysis.

Comparison 4 Dynamic weight bearing exercise high force versus control, Outcome 14 Volumetric bone density % change: tibial trabecular.

4.15. Analysis.

Comparison 4 Dynamic weight bearing exercise high force versus control, Outcome 15 Volumetric bone density % change: tibial cortical.

4.16. Analysis.

Comparison 4 Dynamic weight bearing exercise high force versus control, Outcome 16 Fractures.

Comparison 5. Non‐weight bearing exercise low force versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral density % change: spine	5	231	Mean Difference (IV, Fixed, 95% CI)	‐0.17 [‐1.13, 0.79]
2 Bone mineral density % change: total hip	3	99	Mean Difference (IV, Fixed, 95% CI)	‐0.03 [‐4.94, 4.89]
3 Bone mineral density % change: femoral neck	3	99	Mean Difference (IV, Fixed, 95% CI)	0.21 [‐6.02, 6.45]
4 Bone mineral density % change: Ward's triangle	2	33	Mean Difference (IV, Fixed, 95% CI)	2.75 [‐17.96, 23.47]
5 Bone mineral density % change: trochanter	2	81	Mean Difference (IV, Fixed, 95% CI)	0.05 [‐7.04, 7.14]
6 Bone mineral density % change: total body	2	81	Mean Difference (IV, Fixed, 95% CI)	1.27 [‐2.73, 5.27]

5.1. Analysis.

Comparison 5 Non‐weight bearing exercise low force versus control, Outcome 1 Bone mineral density % change: spine.

5.2. Analysis.

Comparison 5 Non‐weight bearing exercise low force versus control, Outcome 2 Bone mineral density % change: total hip.

5.3. Analysis.

Comparison 5 Non‐weight bearing exercise low force versus control, Outcome 3 Bone mineral density % change: femoral neck.

5.4. Analysis.

Comparison 5 Non‐weight bearing exercise low force versus control, Outcome 4 Bone mineral density % change: Ward's triangle.

5.5. Analysis.

Comparison 5 Non‐weight bearing exercise low force versus control, Outcome 5 Bone mineral density % change: trochanter.

5.6. Analysis.

Comparison 5 Non‐weight bearing exercise low force versus control, Outcome 6 Bone mineral density % change: total body.

Comparison 6. Non‐weight bearing exercise high force versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral density % change: spine	8	246	Mean Difference (IV, Fixed, 95% CI)	0.86 [0.58, 1.13]
2 Bone mineral density % change: total hip	5	165	Mean Difference (IV, Fixed, 95% CI)	0.11 [‐0.06, 0.29]
3 Bone mineral density % change: femoral neck	8	247	Mean Difference (IV, Fixed, 95% CI)	1.03 [0.24, 1.82]
4 Bone mineral density % change: Ward's triangle	4	108	Mean Difference (IV, Fixed, 95% CI)	‐1.77 [‐3.87, 0.33]
5 Bone mineral density % change: trochanter	4	149	Mean Difference (IV, Fixed, 95% CI)	0.40 [‐1.36, 2.17]
6 Bone mineral density % change: total body	3	100	Mean Difference (IV, Fixed, 95% CI)	0.55 [‐0.51, 1.62]

6.2. Analysis.

Comparison 6 Non‐weight bearing exercise high force versus control, Outcome 2 Bone mineral density % change: total hip.

6.4. Analysis.

Comparison 6 Non‐weight bearing exercise high force versus control, Outcome 4 Bone mineral density % change: Ward's triangle.

6.5. Analysis.

Comparison 6 Non‐weight bearing exercise high force versus control, Outcome 5 Bone mineral density % change: trochanter.

6.6. Analysis.

Comparison 6 Non‐weight bearing exercise high force versus control, Outcome 6 Bone mineral density % change: total body.

Comparison 7. Combination versus control.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral density % change: spine	4		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
1.1 immediately postintervention	4	258	Mean Difference (IV, Fixed, 95% CI)	3.22 [1.80, 4.64]
1.2 Follow‐up at 1 year	1	28	Mean Difference (IV, Fixed, 95% CI)	3.33 [1.13, 5.53]
1.3 Follow‐up at 5 years	1	34	Mean Difference (IV, Fixed, 95% CI)	‐1.60 [‐5.64, 2.44]
2 Bone mineral density % change: total hip	4	468	Mean Difference (IV, Fixed, 95% CI)	‐1.07 [‐1.58, ‐0.56]
3 Bone mineral density % change: trochanter	2		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
3.1 immediately postintervention	2	200	Mean Difference (IV, Fixed, 95% CI)	1.31 [0.69, 1.92]
3.2 Follow‐up at 5 years	1	34	Mean Difference (IV, Fixed, 95% CI)	‐3.50 [‐9.93, 2.93]
4 Bone mineral density % change: total body	2		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
4.1 immediately postintervention	2	213	Mean Difference (IV, Fixed, 95% CI)	0.14 [‐0.32, 0.60]
4.2 Follow‐up at 5 years	1	34	Mean Difference (IV, Fixed, 95% CI)	‐0.70 [‐2.19, 0.79]
5 Calcium bone index % change: trunk and upper thighs	1	31	Mean Difference (IV, Fixed, 95% CI)	9.04 [‐5.13, 23.21]
6 Bone mineral density % change: neck of femur	3		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
6.1 immediately postintervention	3	325	Mean Difference (IV, Fixed, 95% CI)	0.45 [0.08, 0.82]
6.2 Follow‐up at 5 years	1	34	Mean Difference (IV, Fixed, 95% CI)	0.70 [‐3.33, 4.73]
7 Bone mineral density % change: Ward's triangle	1		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
7.1 immediately postintervention	1	40	Mean Difference (IV, Fixed, 95% CI)	8.38 [‐7.27, 24.03]
7.2 Follow‐up at 5 years	1	34	Mean Difference (IV, Fixed, 95% CI)	‐2.0 [‐7.96, 3.96]
8 Bone mineral density % change: arms	1		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
8.1 immediately postintervention	1	40	Mean Difference (IV, Fixed, 95% CI)	0.02 [‐9.43, 9.47]
8.2 Follow‐up at 5 years	1	34	Mean Difference (IV, Fixed, 95% CI)	‐0.60 [‐3.65, 2.45]
9 Fractures	2		Odds Ratio (M‐H, Random, 95% CI)	Subtotals only
9.1 Total number of fractures	2	236	Odds Ratio (M‐H, Random, 95% CI)	0.33 [0.13, 0.85]

7.4. Analysis.

Comparison 7 Combination versus control, Outcome 4 Bone mineral density % change: total body.

7.5. Analysis.

Comparison 7 Combination versus control, Outcome 5 Calcium bone index % change: trunk and upper thighs.

7.7. Analysis.

Comparison 7 Combination versus control, Outcome 7 Bone mineral density % change: Ward's triangle.

7.8. Analysis.

Comparison 7 Combination versus control, Outcome 8 Bone mineral density % change: arms.

Comparison 8. Dynamic weight bearing exercise high force plus HRT versus HRT.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral density % change: proximal tibia	1	19	Mean Difference (IV, Fixed, 95% CI)	‐2.33 [‐21.77, 17.11]
2 Bone mineral density % change: hip	2	86	Mean Difference (IV, Fixed, 95% CI)	0.17 [‐6.37, 6.72]
3 Bone mineral density % change: mid femur	1	19	Mean Difference (IV, Fixed, 95% CI)	0.35 [‐3.33, 4.03]
4 Bone mineral density % change: spine	2	203	Mean Difference (IV, Fixed, 95% CI)	‐0.14 [‐0.87, 0.60]
5 Bone mineral density % change: trochanter	2	203	Mean Difference (IV, Fixed, 95% CI)	1.86 [0.60, 3.13]
6 Bone mineral density % change: femoral neck	2	203	Mean Difference (IV, Fixed, 95% CI)	0.59 [‐0.50, 1.67]
7 Bone mineral density % change: total body	1	136	Mean Difference (IV, Fixed, 95% CI)	0.01 [‐0.32, 0.34]

8.1. Analysis.

Comparison 8 Dynamic weight bearing exercise high force plus HRT versus HRT, Outcome 1 Bone mineral density % change: proximal tibia.

8.2. Analysis.

Comparison 8 Dynamic weight bearing exercise high force plus HRT versus HRT, Outcome 2 Bone mineral density % change: hip.

8.3. Analysis.

Comparison 8 Dynamic weight bearing exercise high force plus HRT versus HRT, Outcome 3 Bone mineral density % change: mid femur.

8.4. Analysis.

Comparison 8 Dynamic weight bearing exercise high force plus HRT versus HRT, Outcome 4 Bone mineral density % change: spine.

8.5. Analysis.

Comparison 8 Dynamic weight bearing exercise high force plus HRT versus HRT, Outcome 5 Bone mineral density % change: trochanter.

8.6. Analysis.

Comparison 8 Dynamic weight bearing exercise high force plus HRT versus HRT, Outcome 6 Bone mineral density % change: femoral neck.

8.7. Analysis.

Comparison 8 Dynamic weight bearing exercise high force plus HRT versus HRT, Outcome 7 Bone mineral density % change: total body.

Comparison 9. Non‐weight bearing exercise high force plus bisphosphonates versus bisphosphonates.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral density % change: spine	1	26	Mean Difference (IV, Fixed, 95% CI)	‐3.4 [‐5.79, ‐1.01]
2 Bone mineral density % change: hip	1	26	Mean Difference (IV, Fixed, 95% CI)	‐0.1 [‐1.90, 1.70]
3 Bone mineral density % change: femoral neck	1	26	Mean Difference (IV, Fixed, 95% CI)	‐0.3 [‐2.52, 1.92]
4 Bone mineral density % change: trochanter	1	26	Mean Difference (IV, Fixed, 95% CI)	‐0.5 [‐2.74, 1.74]
5 Bone mineral density % change: Ward's triangle	1	26	Mean Difference (IV, Fixed, 95% CI)	‐1.5 [‐6.52, 3.52]
6 Bone mineral density % change: total body	1	26	Mean Difference (IV, Fixed, 95% CI)	‐0.30 [‐1.96, 1.36]
7 Bone mineral content % change: total body	1	26	Mean Difference (IV, Fixed, 95% CI)	‐1.5 [‐3.35, 0.35]

9.1. Analysis.

Comparison 9 Non‐weight bearing exercise high force plus bisphosphonates versus bisphosphonates, Outcome 1 Bone mineral density % change: spine.

9.2. Analysis.

Comparison 9 Non‐weight bearing exercise high force plus bisphosphonates versus bisphosphonates, Outcome 2 Bone mineral density % change: hip.

9.3. Analysis.

Comparison 9 Non‐weight bearing exercise high force plus bisphosphonates versus bisphosphonates, Outcome 3 Bone mineral density % change: femoral neck.

9.4. Analysis.

Comparison 9 Non‐weight bearing exercise high force plus bisphosphonates versus bisphosphonates, Outcome 4 Bone mineral density % change: trochanter.

9.5. Analysis.

Comparison 9 Non‐weight bearing exercise high force plus bisphosphonates versus bisphosphonates, Outcome 5 Bone mineral density % change: Ward's triangle.

9.6. Analysis.

Comparison 9 Non‐weight bearing exercise high force plus bisphosphonates versus bisphosphonates, Outcome 6 Bone mineral density % change: total body.

9.7. Analysis.

Comparison 9 Non‐weight bearing exercise high force plus bisphosphonates versus bisphosphonates, Outcome 7 Bone mineral content % change: total body.

Comparison 10. Dynamic weight bearing exercise high force plus bisphosphonates versus bisphosphonates.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral content % change: spine	2		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
1.1 immediately postintervention	2	126	Mean Difference (IV, Fixed, 95% CI)	0.93 [‐7.70, 9.56]
1.2 Follow‐up at 15 months postintervention	1	51	Mean Difference (IV, Fixed, 95% CI)	‐2.33 [‐15.79, 11.13]
2 Bone mineral content % change: femoral neck	1		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
2.1 Immediately postintervention	1	76	Mean Difference (IV, Fixed, 95% CI)	‐0.65 [‐10.81, 9.51]
2.2 Follow‐up at 15 months postintervention	1	51	Mean Difference (IV, Fixed, 95% CI)	‐1.01 [‐12.37, 10.35]
3 Bone mineral content % change: wrist	1		Mean Difference (IV, Fixed, 95% CI)	Subtotals only
3.1 Immediately postintervention	1	76	Mean Difference (IV, Fixed, 95% CI)	0.52 [‐13.99, 15.03]
3.2 Follow‐up at 15 months postintervention	1	50	Mean Difference (IV, Fixed, 95% CI)	‐0.32 [‐18.97, 18.33]
4 Bone mineral content % change: distal tibia	1	76	Mean Difference (IV, Fixed, 95% CI)	0.18 [‐8.33, 8.69]
5 Bone mineral content % change: tibial shaft	1	76	Mean Difference (IV, Fixed, 95% CI)	0.4 [‐6.31, 7.11]

10.1. Analysis.

Comparison 10 Dynamic weight bearing exercise high force plus bisphosphonates versus bisphosphonates, Outcome 1 Bone mineral content % change: spine.

10.2. Analysis.

Comparison 10 Dynamic weight bearing exercise high force plus bisphosphonates versus bisphosphonates, Outcome 2 Bone mineral content % change: femoral neck.

10.3. Analysis.

Comparison 10 Dynamic weight bearing exercise high force plus bisphosphonates versus bisphosphonates, Outcome 3 Bone mineral content % change: wrist.

10.4. Analysis.

Comparison 10 Dynamic weight bearing exercise high force plus bisphosphonates versus bisphosphonates, Outcome 4 Bone mineral content % change: distal tibia.

10.5. Analysis.

Comparison 10 Dynamic weight bearing exercise high force plus bisphosphonates versus bisphosphonates, Outcome 5 Bone mineral content % change: tibial shaft.

Comparison 11. Non‐weight bearing exercise high force plus antioxidants versus antioxidants.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral density % change: spine	1	16	Mean Difference (IV, Fixed, 95% CI)	‐0.92 [‐18.73, 16.89]
2 Bone mineral density % change: femoral neck	1	16	Mean Difference (IV, Fixed, 95% CI)	‐2.24 [‐21.61, 17.13]

11.1. Analysis.

Comparison 11 Non‐weight bearing exercise high force plus antioxidants versus antioxidants, Outcome 1 Bone mineral density % change: spine.

11.2. Analysis.

Comparison 11 Non‐weight bearing exercise high force plus antioxidants versus antioxidants, Outcome 2 Bone mineral density % change: femoral neck.

Comparison 12. Dynamic weight bearing exercise low force plus Ca²⁺ versus Ca²⁺.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral density % change: femoral neck	2	111	Mean Difference (IV, Random, 95% CI)	4.44 [‐3.44, 12.32]
2 Bone mineral density % change: spine	1	27	Mean Difference (IV, Fixed, 95% CI)	‐1.02 [‐1.36, ‐0.68]
3 Bone mineral density % change: trochanter	2	111	Mean Difference (IV, Random, 95% CI)	4.51 [‐2.00, 13.03]
4 Bone mineral density % change: distal tibia	1	84	Mean Difference (IV, Fixed, 95% CI)	0.60 [0.46, 0.74]
5 Bone mineral density % change: Ward's triangle	1	27	Mean Difference (IV, Fixed, 95% CI)	14.5 [10.05, 18.95]

12.1. Analysis.

Comparison 12 Dynamic weight bearing exercise low force plus Ca²⁺ versus Ca²⁺, Outcome 1 Bone mineral density % change: femoral neck.

12.2. Analysis.

Comparison 12 Dynamic weight bearing exercise low force plus Ca²⁺ versus Ca²⁺, Outcome 2 Bone mineral density % change: spine.

12.3. Analysis.

Comparison 12 Dynamic weight bearing exercise low force plus Ca²⁺ versus Ca²⁺, Outcome 3 Bone mineral density % change: trochanter.

12.4. Analysis.

Comparison 12 Dynamic weight bearing exercise low force plus Ca²⁺ versus Ca²⁺, Outcome 4 Bone mineral density % change: distal tibia.

12.5. Analysis.

Comparison 12 Dynamic weight bearing exercise low force plus Ca²⁺ versus Ca²⁺, Outcome 5 Bone mineral density % change: Ward's triangle.

Comparison 13. Non‐weight bearing exercise low force plus calcium versus calcium.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral density % change: spine	1	66	Mean Difference (IV, Fixed, 95% CI)	0.33 [‐4.98, 5.64]
2 Bone mineral density % change: total hip	1	66	Mean Difference (IV, Fixed, 95% CI)	‐0.08 [‐5.32, 5.16]
3 Bone mineral density % change: femoral neck	1	66	Mean Difference (IV, Fixed, 95% CI)	0.14 [‐6.56, 6.84]
4 Bone mineral density % change: trochanter	1	66	Mean Difference (IV, Fixed, 95% CI)	‐0.01 [‐7.41, 7.39]
5 Bone mineral density % change: total body	1	66	Mean Difference (IV, Fixed, 95% CI)	1.50 [‐3.24, 6.24]

13.1. Analysis.

Comparison 13 Non‐weight bearing exercise low force plus calcium versus calcium, Outcome 1 Bone mineral density % change: spine.

13.2. Analysis.

Comparison 13 Non‐weight bearing exercise low force plus calcium versus calcium, Outcome 2 Bone mineral density % change: total hip.

13.3. Analysis.

Comparison 13 Non‐weight bearing exercise low force plus calcium versus calcium, Outcome 3 Bone mineral density % change: femoral neck.

13.4. Analysis.

Comparison 13 Non‐weight bearing exercise low force plus calcium versus calcium, Outcome 4 Bone mineral density % change: trochanter.

13.5. Analysis.

Comparison 13 Non‐weight bearing exercise low force plus calcium versus calcium, Outcome 5 Bone mineral density % change: total body.

Comparison 14. Non‐weight bearing exercise high force plus calcium versus calcium.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral density % change: femoral neck	1	60	Mean Difference (IV, Fixed, 95% CI)	1.15 [‐6.35, 8.65]
2 Bone mineral density % change: trochanter	1	60	Mean Difference (IV, Fixed, 95% CI)	0.01 [‐7.04, 7.06]
3 Bone mineral density % change: total hip	1	60	Mean Difference (IV, Fixed, 95% CI)	1.14 [‐4.04, 6.32]
4 Bone mineral density % change: spine	1	60	Mean Difference (IV, Fixed, 95% CI)	‐0.64 [‐6.33, 5.05]
5 Bone mineral density % change: total body	1	60	Mean Difference (IV, Fixed, 95% CI)	0.09 [‐4.19, 4.37]

14.1. Analysis.

Comparison 14 Non‐weight bearing exercise high force plus calcium versus calcium, Outcome 1 Bone mineral density % change: femoral neck.

14.2. Analysis.

Comparison 14 Non‐weight bearing exercise high force plus calcium versus calcium, Outcome 2 Bone mineral density % change: trochanter.

14.3. Analysis.

Comparison 14 Non‐weight bearing exercise high force plus calcium versus calcium, Outcome 3 Bone mineral density % change: total hip.

14.4. Analysis.

Comparison 14 Non‐weight bearing exercise high force plus calcium versus calcium, Outcome 4 Bone mineral density % change: spine.

14.5. Analysis.

Comparison 14 Non‐weight bearing exercise high force plus calcium versus calcium, Outcome 5 Bone mineral density % change: total body.

Comparison 15. Dynamic weight bearing exercise low force plus calcium/VitD versus calcium/VitD.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Bone mineral density % change: spine	1	35	Mean Difference (IV, Fixed, 95% CI)	1.42 [‐1.28, 4.12]
2 Bone mineral density % change:wrist	1	35	Mean Difference (IV, Fixed, 95% CI)	1.64 [‐4.81, 8.09]

15.1. Analysis.

Comparison 15 Dynamic weight bearing exercise low force plus calcium/VitD versus calcium/VitD, Outcome 1 Bone mineral density % change: spine.

15.2. Analysis.

Comparison 15 Dynamic weight bearing exercise low force plus calcium/VitD versus calcium/VitD, Outcome 2 Bone mineral density % change:wrist.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Bemben 2000.

Methods	Type of study: RCT
Participants	Number of participants randomised = 35 Losses: 10 (4 high repetitions, 3 high load, 3 control) Age: 41‐60 years Setting: USA Inclusion:1–7 yr postmenopausal and had not performed any resistance training in the previous 6 months Exclusion: 1) diagnosed osteoporosis or a BMD site ≥ 2.5 SD below the mean for the young‐adult reference population; 2) a history of cardiovascular disease; 3) physical or orthopaedic disabilities; 4) a history or current diagnosis of renal disease, chronic digestive or eating disorders, rheumatoid arthritis, or thyroid disease; 5) a history of prolonged bed rest; and 6) current or recent use of medications that affect bone density (i.e. oestrogen, steroid hormones, calcitonin or corticosteroids)
Interventions	Exercise group high load (HL) (NWBHF) (n = 10): 10‐min warm‐up, approximately 45 min of weight lifting, and ended with a 5‐min cool‐down. Quadriceps extension, hamstring flexion, leg press, shoulder press, biceps curl, triceps extension, seated row and latissimus pull. High load low reps (8 reps 80% 1RM) Exercise group high repetition (HR) (NWBLF) (n = 7): 10‐min warm‐up, approximately 45 min of weight lifting, and ended with a 5‐min cool‐down. Quadriceps extension, hamstring flexion, leg press, shoulder press, biceps curl, triceps extension, seated row and latissimus pull. Low load high reps (16 reps 40% 1RM) Control Group (n = 8): usual activity Duration and intensity: 3 sessions per week for 6 months Supervisor: Research assistants Supervision: Group Setting: Gym
Outcomes	% Change BMD spine, hip (total hip, neck of femur, trochanter, Wards triangle), total body
Notes	Compliance/adherence: average attendance for the 6‐month intervention was 93% for HR and 87% for HL Adverse events: none reported Converted absolute data to % change
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Subjects were matched according to the BMD of the spine after baseline testing, then they were randomly assigned, method not described
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	High risk	'As‐treated' analysis done, drop‐outs mentioned but not accounted for in analysis
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Unclear risk	Not reported
Comparability of exercise and control group at entry	Low risk	No significant group differences existed in number of years postmenopausal or in body composition variables
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 6 months, no follow‐up data reported

Bergstrom 2008.

Methods	Type of study: RCT
Participants	Number of participants randomised = 112 Losses: 20 (Exercise: 1 failed to attend DXA, 11 did not start training or trained less than 6 months, Control: 8 undertook other exercise) Age: 59.6 Exercise, 58.9 control Setting: Sweden Inclusion: postmenopausal women 45 to 65 years with forearm fractures and T‐scores from −1.0 to −3.0 (total hip or spine) Exclusion: T‐score lower than −3 at any site, had any disease known to interfere with bone metabolism, were on cortisone therapy or anti‐resorptive medication, including hormone replacement therapy, had a BMI lower than 19.9 or higher than 30.9, or were already training at the level of or above that of the intervention
Interventions	Exercise group (COMB) (n = 48): 3 fast 30‐minute, walks and two sessions of one‐hour training per week. 5‐minute warm‐up, 25 minutes of strengthening exercises for the arms, legs, back and stomach, 25 minutes of aerobic exercise, and 5 minutes of stretching. Individuals chose own level and intensity and encouraged to increase level if possible Control Group (n = 44): usual activity Duration and intensity: 5 sessions per week for 12 months Supervisor: nurses Supervision: group Setting: clinic
Outcomes	% change BMD DEXA spine, total hip
Notes	Compliance/adherence: controlled by study nurse (compliance was 95%) Adverse events: none reported 80% power difference, 3% with 64 in each group Converted absolute data to % change
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Predefined random number table
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Low risk	Per protocol and intention‐to‐treat analysis
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Unclear risk	Not reported
Comparability of exercise and control group at entry	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 1 year, no follow‐up data reported

Bocalini 2009.

Methods	Type of study: RCT
Participants	Number of participants randomised = 35 Losses: 10 (3 exercise, 2 control, plus 5 in exercise did not achieve 90% participation) Age: range 57‐75 years Setting: Brazil Inclusion: women older than 55 years (and able to train 3 x per week for 24 weeks) Exclusion: participation in a regular and structured physical activity for the last 3 months; recent hospitalisation; motor deficiency; symptomatic cardiorespiratory disease; non controlled hypertension or metabolic syndrome; severe renal or hepatic disease; cognitive impairment or debilitating conditions; marked obesity with inability to exercise; recent bone fracture (during the past 2 years); use of any medication that may alter calcium or bone metabolism; other medical contraindications to exercise
Interventions	Exercise group strength training (NWBHF) (n = 15): Eccentric muscle action was emphasised for leg press, chest press, leg curl, latissimus pull down, elbow flexion, elbow extension, leg extension, upper back row, military press, hip abductor, hip adductor and abdominal curls. 10‐minute warm‐up, (running with low impact at 50% MHR), one set 50% 1 RM, progressing to 3 sets 85% 1 RM Control Group (n = 10): usual activity Duration and intensity: 1 hr sessions 3 x per week in non‐consecutive days for 24 weeks Supervisor: fitness instructor and researchers Supervision: probably group Setting: gym
Outcomes	% change BMD DEXA lumbar spine, femoral neck
Notes	Compliance/adherence: all completers required to participate in 90% of programme Adverse events: none reported Converted SE to SD
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation mentioned but insufficient information to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	'As‐treated' analysis done, drop outs mentioned and 5 in exercise group did not achieve 90% participation, thus were excluded
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Participants not blinded
Blinding (assessor)	Low risk	Investigator blind to subject condition
Comparability of exercise and control group at entry	Low risk	No differences were identified between groups, concerning biometric characteristics, muscle strength or bone densitometry parameters
Appropriateness of duration of surveillance	High risk	Outcomes only on immediately postintervention (24 weeks), no follow‐up data reported

Bravo 1996.

Methods	RCT
Participants	Number of participants randomised = 142 Losses: 18 (equally divided across groups) Age: mean 60±6 years Setting: Canada Inclusion: menopausal (> 12 months) community‐dwelling women between the ages of 50 and 70, with low bone mass (spine > 1g/cm2, proximal femur >0.9g/cm2), no contradictions to undertaking physical exercise without supervision Exclusion: not stated
Interventions	Exercise group (DWBLF) (n = 61): warm up, 25 min of rapid walking: 15 min of stepping down and up or aerobic dancing, each had to progressively reach 60‐70% of her heart rate reserve, localised exercise: 10‐15 min of exercises in sitting, standing, prone position, involving the muscles of upper limbs, abdominals and the back, cool down period with relaxation movements, stretching, balancing and coordination exercises Control group (n = 63): to continue their daily routine activities plus education Duration and intensity: 1 hour long exercise classes, 3/week per 12 months Supervisor: exercise leaders Supervision: group Setting: gym
Outcomes	BMD spine, BMD neck of femur
Notes	Compliance/adherence: not reported Adverse events: none reported Converted absolute data to % change.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Random number tables, block randomisation and stratified by age and HRT use
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Low risk	Missing outcome data balanced in numbers across intervention groups, with similar reasons for missing data across groups Intention‐to‐treat analysis performed
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk' but exercise is difficult to blind
Blinding (assessor)	Low risk	Three assessors blind to group allocation
Comparability of exercise and control group at entry	Low risk	No significant differences between groups apart from years postmenopause (longer in control group) and use of oestrogen (more in exercise group)
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 12 months, no follow‐up data reported

Brentano 2008.

Methods	Type of study: RCT
Participants	Number of participants randomised = 28 Losses: not stated Age: not stated Setting: Brazil Inclusion: not stated Exclusion: not stated
Interventions	Exercise group Circuit training (NWBLF) (n = 9): No rest between exercises; progression of loads. Warm‐up: 5 minutes cycloergometer or treadmill, 20–10 repetitions and 45–60% 1RM, performing 2–3 sets for each exercise, leg press, hip abduction, hip adduction, knee extension, chest fly, reverse fly, arm curl, triceps push‐down, sit‐ups and back extension Exrcise group High intensity (NWBHF) (n = 10): 2 min rest between exercises; progression of loads. Warm‐up: 5 minutes cycloergometer or treadmill. 20–6 repetitions and 45–80% 1RM, performing 2–4 sets for each exercise, leg press, hip abduction, hip adduction, knee extension, chest fly, reverse fly, arm curl, triceps push‐down, sit‐ups and back extension Control Group (n = 9): usual activity Duration and intensity: 1 hr 3 x week for 24 weeks Supervisor: not stated Supervision: group Setting: gym
Outcomes	BMD femoral neck, femoral trochanter, Ward's triangle, intertrochanter
Notes	Compliance/adherence: not stated Adverse events: none reported Data presented graphically only
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Subjects were divided into 2 subgroups: taking HRT (n = 14) and not taking HRT (n = 14). Then, the subgroups were randomly divided but insufficient information to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	Unclear risk	Not possible
Blinding (assessor)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Comparability of exercise and control group at entry	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data, no follow‐up data reported

Chan 2004.

Methods	Type of study: RCT
Participants	Number of participants randomised = 132 Losses: 24 (13 Exercise, 11 control) Age: 54 (±3.5) years Setting: Hong Kong Inclusion: ceased menstruation between 1 and 10 years (1) no regular participation in physical exercise (not > 0.5h/wk); (2) no hormone replacement therapy or drug treatment known to affect bone metabolism or cause spontaneous bone loss; (3) no conditions such as hypo‐ or hyperparathyroidism and hypo‐ or hyperthyroidism, or renal or liver disease; (4) no history of fractures; and (5) a body mass index (BMI) above 30kg/m2 Exclusion:
Interventions	Exercise group (DWBLF) (n = 67): Yang Tau Chi Chuan style emphasises slow and smooth movement involving major muscle groups, at a constant speed while practicing Control Group (n = 65): usual activity Duration and intensity: 50 mins 5x week for 12 months Supervisor: not reported Supervision: group Setting: community
Outcomes	% change BMD DXA lumbar spine, neck of femur, total hip, trochanter, distal tibia Fracture rate
Notes	Compliance/adherence: Average attendance rate of the TCC exercise was 4.2±0.9 days per week Adverse events: Fractures occurred during the follow‐up. During the 12‐month study period, a total of 4 fracture cases were documented, including 3 fracture in the control group (1 vertebral fracture, 1 Colles’ fractures, 1 fracture at the fifth metacarpal) and 1 in the TCC group (proximal fibular fracture). All injury cases resulted from overloading during work (the case with vertebral fracture) or falls (the other 3 cases) Sample size of 45 for each group was estimated to achieve a statistical power of 0.8 after excluding a dropout rate of approximately 25% during 12‐month follow‐up. Converted absolute data to % change
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation mentioned but insufficient information to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	'As‐treated' analysis done drop‐outs mentioned
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Unclear risk	Not reported
Comparability of exercise and control group at entry	Low risk	No significant differences in characteristics at baseline
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data 12 months, no follow‐up data reported

Cheng 2002.

Methods	Type of study: RCT
Participants	Number of participants randomised = 80 Losses: 28 (non HRT 8 exercise, 5 control; HRT 10 exercise, 5 control) Age: 50‐57 years Setting: Finland Inclusion: 50 –55‐year‐old women, no serious cardiovascular or locomotor system problems, a body mass index of 33 kg/m2, and not currently or previously (no longer than 6 months and at least 2 years prior to screening) using medications including oestrogen, fluoride, calcitonin, bisphosphonate's, and steroids, last menstruation at least 0.5 years but not more than 5 years ago Exclusion: not reported
Interventions	No HRT Exercise group (DWBHF) (n = 20): 5 circuit‐training periods, each lasting 8 –10 weeks. These periods were interrupted by three high‐impact aerobic dance periods, each of 2 week duration, and a summer pause for 5 weeks. Each session commenced with a 10 min warm‐up period and concluded with stretching activities. During the first two circuit training periods, three rotations were performed of skipping (30 sec), bounding over soft hurdles (13–16 cm), drop jumping (10–15 cm), and hopping (on one leg 10 times, added during the second training period). The following three periods comprised four rotations of bounding (19–25 cm), drop jumping (20–25 cm), hopping (10 times per leg) and leaping (10 times). In addition, all circuit training sessions included 3 or 4 of the following resistance exercises for the upper body: chest fly, latissimus pull down, military press, seated row and biceps curl. The home exercise programme was also designed as a circuit training routine comprising three rotations of skipping (30 sec), hopping (10 times per leg) and drop jumping (15 cm). In addition, exercises to strengthen the abdominal and lower back region were included. Average GRF was 4.3 times body weight (BW) for drop‐landing from a 10 cm height, and 5.2 times BW from 20 and 25 cm heights; bounding over the hurdles 4.9–5.1 BW, skipping, hopping, and leaping 3.8, 3.4, and 4.8 BW, respectively. No HRT Control Group (n = 20): usual activity HRT Exercise group (DWBHF) (n = 20): as exercise group above HRT Control Group (n = 20): usual activity Duration and intensity: 2 x supervised and 4 non supervised sessions per week 12 months Supervisor: not stated Supervision: group/individual Setting: gym/home
Outcomes	BMD DXA proximal femur, tibial shaft Cortical tibia
Notes	Compliance/adherence: average attendance 1 x per week Adverse events: none reported Converted absolute data to % change
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation by drawing lots
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	'As‐treated' analysis done drop‐outs mentioned but not controlled for
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	Unclear risk	Stated double‐blind (may be related to the HRT component) but insufficient information to permit judgement of 'high risk' or 'low risk'
Blinding (assessor)	Unclear risk	Stated double‐blind (may be related to the HRT component) but insufficient information to permit judgement of 'high risk' or 'low risk'
Comparability of exercise and control group at entry	Low risk	No significant differences in physical characteristics at baseline
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 12 months, no follow‐up data reported

Chilibeck 2002.

Methods	Type of study: RCT
Participants	Number of participants randomised = 57 Losses: 9 (4 non bisphosphonate exercise, 3 bisphosphonate exercise, 2 non bisphosphonate control) Age: mean age of groups ranged from 55.9 to 58.8 years Setting: Canada Inclusion: postmenopausal status (cessation of bleeding status for one year) Exclusion: skeletal; disorders, kidney disease or bone related disorders, chronic disease or chronic medication likely to affect metabolism or calcium imbalance. BMD z‐score < ‐2.0, HRT, bisphosphonate therapy in last year, recent participation in exercise programmes, history of cardiac disease or high blood pressure
Interventions	All received 10 μg vitamin D/d and those in non bisphosphonate received 500 mg calcium carbonate/d Non bisphosphonate exercise group (NWBHF) (n = 10): warm up cycling and stretching, 2 sets 8‐10 reps of; bench press, latissimus dorsi pull down, shoulder press, biceps curl, back extension, hip extension, flexion, adduction and abduction, knee flexion, knee extension and leg press. initially 70% 1RM then progressed Non bisphosphonate control Group (n = 12): usual activity Bisphosphonate exercise group (NWBHF) (n = 12): as above Bisphosphonate control group (n = 14): usual activity Duration and intensity: 3 days per week for 12 months Supervisor: not stated Supervision: individual Setting: gym
Outcomes	% change BMD spine, total hip, femoral neck, trochanter, Ward's triangle, whole body % change whole body BMC
Notes	Compliance/adherence: Non bisphosphonate exercise group 77.6%, bisphosphonate exercise group 74.8% of training sessions Adverse events: none reported 9 subjects per group would demonstrate change α of 0.05 with 80% power Converted SE to SD
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation mentioned but insufficient information to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Low risk	Insufficient information although mentions double‐blind
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	'As‐treated' analysis done drop‐outs mentioned but not controlled for
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	Unclear risk	Mentions double‐blind but probably relates to medication status
Blinding (assessor)	Low risk	States double‐blind
Comparability of exercise and control group at entry	Low risk	No significant differences in characteristics at baseline
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 12 months, no follow‐up data reported

Chow 1987.

Methods	RCT
Participants	Number of participants randomised = 58 Losses: 4 controls, 2 (DWBLF), 4 (COMB) Age: mean age 56 years Setting: Canada Inclusion: no history of: fractures, metabolic bone disease, renal, liver or thyroid disorders, gastrectomy, alcoholism, oestrogen or other drugs affecting bone metabolism Exclusion: not reported
Interventions	Exercise group (DWBHF) (n = 19): 5‐10 min of stretching and calisthenic warm up, exercise followed by 30 min of aerobic activities at 80% MaxHR (walking, jogging, dance) Exercise group (COMB) (n = 20) 5‐10 min of stretching and calisthenic warm up, exercise followed by 30 min of aerobic activities at 80% MaxHR (walking, jogging, dance) plus 10‐15 min session of low intensity strength training (isometric and isotonic contractions of limbs and trunk muscles. 10 repetitions for each muscle group Control Group (n = 19): continue daily routine activities, refrain from any regular fitness exercises (telephoned 4 x per year) Duration and intensity: 3 sessions per week for 1 year Supervisor: certified fitness instructor Supervision: group Setting: hospital gym
Outcomes	CaBI.
Notes	Compliance/adherence: overall attendance at exercise class was 70% Adverse events: 1 (DWBLF) knee pain; 2 (COMB) knee pain; 1 (COMB) back pain Power calculation done 15 per group, and all groups of appropriate size Converted absolute data to % change
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Random number generator
Allocation concealment (selection bias)	Low risk	Sequential sealed envelopes
Incomplete outcome data (attrition bias) All outcomes	Low risk	Details supplied. Discussion of reasons for dropout. Comparison with dropout and excluded groups
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not practical for exercise programme
Blinding (assessor)	Low risk	Assessors blind
Comparability of exercise and control group at entry	Low risk	Initial mean values of bone mass and aerobic capacity were within normal ranges for all groups
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 12 months, no follow‐up data reported

Chubak 2006.

Methods	Type of study: RCT
Participants	Number of participants randomised = 173 Losses: 3 exercise group Age:50‐75 years, mean 61 years Setting: USA Inclusion: overweight/obese, postmenopausal women sedentary (< 60 min/wk of moderate‐to vigorous‐intensity exercise), overweight (BMI 25.0 to 30.0 kg/m2, or BMI between 24.0 and < 25.0 kg/m2 and percent body fat > 33%) or obese (BMI > 30.0 kg/m2), no menstrual periods for the previous 12 months Exclusion: using hormone therapy in the past 6 months, being too physically active, having medical conditions contraindicating moderate to vigorous‐intensity exercise, having a clinical diagnosis of diabetes, and currently using tobacco
Interventions	Exercise group (COMB) (n = 87): moderate‐intensity aerobic exercise (60–75% of maximal heart rate), 40% of observed maximal heart rate for 16 min per session and gradually increased to 60–75% of maximal heart rate for 45 min per session by week 8. Treadmill walking and stationary bicycling. Strength training, consisting of two sets of 10 repetitions of leg extension, leg curls, leg press, chest press, and seated dumbbell row Control Group (n = 86): 45‐min stretching sessions once a week Duration and intensity: 45 mins, 5 days per week for 12 months (3 supervised sessions per week months 1–3 and to exercise 2 d/wk at home; months 4–12 at least one of the three supervised sessions weekly and to exercise 4 d/wk either at home or elsewhere Supervisor: not stated Supervision: group and individual Setting: gym and home
Outcomes	BMD Total body
Notes	Compliance/adherence: exercisers averaged 172 min/wk (SD = 89) of exercise Adverse events: none reported Converted absolute data to % change
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Stratifying on body mass index (BMI) (above and below 27.5 kg/m) randomly assigned women to either the exercise or the stretching arm of the trial. Randomisation was performed by random number generation
Allocation concealment (selection bias)	Low risk	Group assignment was placed in a sealed envelope, which was opened by the study coordinator at the time of randomisation
Incomplete outcome data (attrition bias) All outcomes	Low risk	Intention‐to‐treat analysis performed
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	Unclear risk	Not possible
Blinding (assessor)	Low risk	Technicians blinded to group allocation
Comparability of exercise and control group at entry	Low risk	Groups similar with respect to demographic characteristics and known predictors of bone mineral density and other subject characteristics
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 12 months, no follow‐up data reported

Chuin 2009.

Methods	Type of study: RCT
Participants	Number of participants randomised = 34 Losses: not reported Age: 61‐73 years, 66.1 years Setting: Canada Inclusion: healthy, Caucasian, without major incapacity, no medication influencing metabolism, non‐smoker, moderate drinker, NBMI 18‐30 kg/m2, no consumption of antioxidant supplements during last month, postmenopausal, no HRT Exclusion: not reported
Interventions	Antioxidants (600 mg/day vitamin E and 1,000 mg/day vitamin C) Placebo and Exercise group (NWBHF) (n = 8): 15 min warm up, treadmill, cycle and stretching; 45 mins resistance training leg press, bench press, leg extension, shoulder press, sit up, seated row, triceps extensions, biceps curl. 3 sets 8 reps per set at 80% 1RM. Placebo Control Group (n = 7): usual activity Antioxidant exercise group (NWBHF) (n = 8): as exercise above Antioxidant control group (n = 8): usual activity Duration and intensity: 60 mins sessions 3 x week for 6 months Supervisor: not stated Supervision: groups Setting: gym
Outcomes	BMD spine, femoral neck
Notes	Compliance/adherence: not stated but one missed session per month accepted for compliance purposes Adverse events: none reported Converted absolute data to % change
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation mentioned but insufficient information to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Completer analysis. Loss not accounted for
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Comparability of exercise and control group at entry	Low risk	Groups were similar for baseline characteristics, body composition, strength
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data 6 months, no follow‐up data reported

Ebrahim 1997.

Methods	RCT
Participants	Number of participants randomised = 165 Losses: 32 exercise group, 36 control Age: mean ages (66‐70) years Setting: UK Inclusion: women who had sustained an upper arm fracture in the past 2 years Exclusion: not recorded
Interventions	Exercise group (DWBLF) (n = 49): self paced brisk walking Control Group (n = 48): upper limb exercises for fracture Duration and intensity: 40 mins 3 x week for 2 years Supervisor: nurse Supervision: seen every 3 months and phone calls monthly Setting: home
Outcomes	BMD lumbar spine, femoral neck
Notes	Compliance/adherence: All women completing trial reported carrying out regular brisk walking at least 40 mins three times per week Adverse events: 1 exercise related trauma reported. By the end of the trial the brisk walking group had sustained a significant excess of 15.2 falls per 100 person years No power calculation. Very small study (15 total, 5 per exercise group)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer generated randomised allocation
Allocation concealment (selection bias)	Low risk	Sequentially numbered envelopes
Incomplete outcome data (attrition bias) All outcomes	Low risk	Missing outcome data balanced in numbers across intervention groups, with similar reasons for missing data across groups. Data for all participants, including one dropout at 11 months
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk' but exercise is difficult to blind
Blinding (assessor)	High risk	Same nurse saw both groups
Comparability of exercise and control group at entry	Low risk	No significant differences between groups apart from slightly younger women in the exercise group
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 2 years, no follow‐up data reported

Englund 2005.

Methods	Type of study: RCT
Participants	Number of participants randomised = 48 Losses: 3 exercise, 5 control Age: 66–87 years Setting: Sweden Inclusion: not reported Exclusion:dementia, current smoking, current hormone replacement therapy (HRT), and use of a walking aid, cardiovascular disease, or functional disability, of a degree that would contraindicate physical exercise
Interventions	Exercise group (COMB) (n = 24): 10 min of warming‐up, followed by a mix of aerobic (walking and jogging), strengthening (legs, abdominal, and back muscles were trained by means of body resistance only) , balance and coordination exercises for 27 min. The programme then ended with 11 min of cooling down, stretching and relaxation. If participants missed out on a training session they were advised to perform a home exercise programme instead. This programme included brisk walking for 30 min, squats with 3·10 repetitions, and training of hand grip with a piece of T‐foam for 3·15 repetitions. Control Group (n = 24): Duration and intensity: 50 mins twice a week for 12 months, with a 5‐week break during the summer vacation Supervisor: physiotherapist Supervision:not reported Setting:not reported
Outcomes	BMD spine, femoral neck, trochanter, Ward's triangle, arms, total body (g/cm2) BMC total body (g)
Notes	Compliance/adherence: mean percentage of scheduled sessions attended for the exercise group was 67% Adverse events: not recorded A sample size of 24 in each group, a‐level of 0.05 and standard deviation of 10% gave 30% power to detect a 5% difference in change between the two groups Converted absolute data to % change
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Pair‐wise age matched, randomised mentioned but insufficient information to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	'As‐treated' analysis done drop‐outs mentioned but not accounted for in analysis
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Unclear risk	Unclear but the same investigator carried out all analyses
Comparability of exercise and control group at entry	Low risk	There was no significant difference in total BMD between the groups at the beginning of the study but mean age for menopause was significantly higher in the control group
Appropriateness of duration of surveillance	Low risk	Immediately postintervention data 12 months and 5 year follow‐up data reported

Going 2003.

Methods	Type of study: RCT
Participants	Number of participants randomised = 320 (HRT, n = 159; NHRT, n = 161) Losses: Retention rates were 82%, 89%, 78% and 84% for EX/NHRT, NEX/ HRT, EX/NHRT and NEX/NHRT, respectively. The dropout rate for EX (20%) and NEX (13%), NonHRT (19%), HRT (14%) Age: 40‐65 years Setting: USA Inclusion: women who were undergoing hormone replacement therapy for at least 1 year and not more than 5.9 years AND women who had not used HRT during the preceding year. Surgical or natural menopause (3‐10.9 years). BMI < 33, non smoker, no history of osteoporotic fractures, initial lumbar spine and hip BMD > Z ‐3.0. Cancer and cancer treatment free for the last 5 years, excluding skin cancer, no medication affecting BMD, no beta blockers or steroids. Ca intake > 200 mg per day. Less than 120 min physical activity per week. No weightlifting or similar activity. Exclusion: not reported
Interventions	Divided into HRT and non HRT groups. All groups taking Ca supplements as prescribed HRT Exercise group (DWBHF) (n = 86): supervised aerobic, weight‐bearing and weight‐lifting exercise. Leg press, hack squats or Smith squats, lat pull downs, lateral rows, back extensions, right and left arm dumbbell presses, and rotary torso. Two sets of 6 to 8 repetitions 70% ‐ 80% 1‐RM. Weight bearing circuit comprising walk/jog, skipping, hopping, stair climbing/boxstep. Progressive impact regime HRT Control Group (n = 73): usual activity NonHRT Exercise group (DWBHF) (n = 91): supervised aerobic, weight‐bearing and weight‐lifting exercise as above Non HRT control Group (n = 70): usual activity Duration and intensity: 3x per week for 12 months Supervisor: trainer Supervision: group Setting: community
Outcomes	BMD DEXA total body, AP lumbar spine, neck of femur, trochanter
Notes	Compliance/adherence: Attendance at exercise sessions averaged 71.8±19.9% Adverse events: none reported Converted data from average change over one year to change at end of study.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation mentioned but insufficient information to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Low risk	Intention‐to‐treat analysis undertaken
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	Unclear risk	Not possible
Blinding (assessor)	Unclear risk	Not reported
Comparability of exercise and control group at entry	Low risk	Groups similar at baseline for age, oestrogen levels, BMD and physical characteristics. Women not using HRT 1.6 years older
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data 12 months, no follow‐up data reported

Grove 1992.

Methods	Type of study:RCT
Participants	Number of participants randomised = 15 Losses:1 subject in DWBHF group injured at 11 months, all post‐test data from subject collected at this time point Age:46‐64 yrs Setting: USA Inclusion: postmenopausal sedentary Caucasian women Exclusion: women who were active during last year; < 1 year or > 8 years postmenopausal; any renal, thyroid or liver disease; unwillingness to complete study; on medications that would affect calcium metabolism and absorption (except oestrogen)
Interventions	Exercise group (DWBLF) (n = 5): 15 mins warm up (stretching), 20 min exercise(low impact), 20 mins cool down (abdominal exercises). GRF for exercises, slow walk = 1.19 BW, fast walk = 1.49 BW, heel jack no jump = 1.34 BW, Charleston = 1.32 BW Exercise group (DWBHF) (n = 5): 15 mins warm up (stretching), 20 min exercise (high impact), 20 mins cool down (abdominal exercises). GRF for exercises, jumping jack = 3.29 BW, running‐in‐place = 2.47 BW, knee‐elbow with jump = 2.79 BW Control Group (n = 5): usual activity Duration and intensity: 1 hr 3x week for 12 months Supervisor: not stated Supervision: group Setting: gym
Outcomes	Lumbar BMD at baseline, 6 and 12 months
Notes	Compliance/adherence: DWBLF = 80.0±6.6%, DWBHF = 82.6±4.1% Adverse events: 1 subject in DWBHF group injured at 11 months
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Matched by BMD and weight and randomly assigned to groups, method not stated
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Small numbers in each group and only one loss
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Comparability of exercise and control group at entry	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 12 months, no follow‐up data reported

Hatori 1993.

Methods	Type of study:RCT
Participants	Number of participants randomised = 35 Losses: 2 from exercise group (lack of time) Age: 46‐67 years Setting: Japan Inclusion: health postmenopausal women no history of oophrectomy Exclusion: not reported
Interventions	Exercise group: DWBLF (n = 9): stretching of the legs, torso and arms, followed by 30 min of walking on flat grass‐covered ground moderate intensity: 90% of the heart rate DWBLF (n = 12): stretching of the legs, torso and arms, followed by 30 min of walking on flat grass‐covered ground high intensity: 110% of the heart rate Control Group (n = 12): not reported Duration and intensity: 3 times/week during 7 months Supervisor: not stated Supervision: not clear Setting: not stated
Outcomes	% change in BMD Lumbar Spine (DEXA)
Notes	Compliance/adherence: not reported Adverse events: none reported Data for the group working at 110% HR was used in analysis
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation reported but insufficient information about the sequence generation process to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	High risk	'As‐treated' analysis done, drop‐outs mentioned but unclear as to which groups
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk' but exercise is difficult to blind
Blinding (assessor)	Low risk	Assessor had no knowledge of group allocation
Comparability of exercise and control group at entry	Low risk	No significant difference between the groups at entry
Appropriateness of duration of surveillance	High risk	Assessment at 7 months during the exercise programme of 7 month duration

Iwamoto 2001.

Methods	Type of study: RCT
Participants	Number of participants randomised = 35 Losses: not reported Age: 53‐77 years Setting: Japan Inclusion: postmenopausal women with diagnosis of osteoporosis Exclusion: not reported
Interventions	Concurrent calcium lactate 2.0 g and hydroxy vitamin D3 1 μg Exercise group (COMB) (n = 15): brisk walking and two sets a day of gymnastic training, consisting of 15 repetitions of straight leg raising, squatting, and abdominal and back muscle strengthening exercises Control Group (n = 20): usual activity Duration and intensity: daily for 12 months Supervisor: not reported Supervision: individual Setting: home
Outcomes	% change BMD lumbar
Notes	Compliance/adherence: 100% at least five days per week Adverse events: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation mentioned but insufficient information to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	Unclear risk	Not possible
Blinding (assessor)	Unclear risk	Not reported. Insufficient information to permit judgement of 'high risk' or 'low risk'
Comparability of exercise and control group at entry	Low risk	No significant differences in initial lumbar BMD
Appropriateness of duration of surveillance	High risk	Immediately postintervention data at 12 months, and follow‐up after further year

Iwamoto 2005.

Methods	Type of study: RCT
Participants	Number of participants randomised = 50 Losses: none reported Age: 70.6±8.7 control, 71.9±8.1 exercise Setting: Japan Inclusion:55‐88 years, BMD score < 70 or 70‐80%, history of osteoporotic fractures and chronic back pain Exclusion: musculoskeletal diseases considered to cause back pain
Interventions	All participants received 5 mg alendronate Exercise group (DWBHF) (n = 25): whole body vibration plate at 20 Hz Control Group (n = 25): usual activity Duration and intensity: 4 mins, 1 x week for 12 months Supervisor: not reported Supervision: individual Setting: clinic
Outcomes	BMD vertebral fractures (radiographs)
Notes	Compliance/adherence: not stated Adverse events: 2 patients in control group and one in exercise group had falls
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation mentioned but insufficient information to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	'As‐treated' analysis done no drop outs mentioned
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Unclear risk	Not reported
Comparability of exercise and control group at entry	Low risk
Appropriateness of duration of surveillance	Low risk

Karinkanta 2007.

Methods	Type of study: RCT
Participants	Number of participants randomised: 149  Losses: Total = 5 (4 from the training groups and 1 from control). (Drop out rate 3.4%) Age: 70‐79 years: RES = 72.7 (2.5); BAL = 72.9 (2.3); COMB = 72.9 (2.2); CON = 72 (2.1) Sex: female Health status defined by authors: healthy older (> 70) females Setting: Finland Inclusion: willing to participate; age between 70‐79 years; full understanding of study procedures; no history of illness contraindicating exercise or limiting participation in exercise programmes, no history of illness affecting balance or bones; no uncorrected vision problems; no medications known to affect balance or bone metabolism (12 months before enrolment) Exclusion: Involved in intense exercise more than 2x week or t‐score for femoral neck bone mineral density (BMD) lower than ‐2.5
Interventions	Exercise group (DWBHF) = resistance training (n = 37). Progressing towards 75‐80% 1RM 3 sets of 8‐10. Large muscle group ex = sit‐stand with weighted vest, squats, leg press, hip abduct, hip extension, calf raise, rowing with resistance machines.  Different combinations of above were used in 10 week cycle to prevent monotony Exercise group (DWBHF) = balance jumping training (n = 37). Balance agility and impact exercise ‐ 4 different aerobics and step aerobic programmes which were repeated. Progressive difficulty of steps, impact and jumps Exercise group (COMB) = resistance and balance jumping training (n = 38). Reistance and balance training on alternate weeks as above Control group = no training (n = 37) Duration and intensity: 3x weekly for 12 months, 50 mins. Warm up 7‐10 mins; 25‐30 mins exercise; 8‐10 min cool down Supervisor: exercise leaders of UKK institute Supervision: groups but uncertain of number in each as not recorded Setting:  not recorded
Outcomes	BMD DEXA Bone mineral content BMC (g) Cortical density (CoD) mg/cm
Notes	Compliance/adherence: mean training compliance = attendance 67% (RES = 74%; COMB = 67%; BAL = 59%) Adverse events: 14 due to musculoskeletal injuries or symptoms  ‐ 2 falls but they returned to classes.  No difference in monthly reported health problems with exercisers and controls Fractures reported during 1 year follow‐up period: Resistance group 1 hip 1 rib; Balance group 1 shoulder; Combined 1 hip; Control 1 patella Initial study data converted absolute data to % change. Follow‐up study data was not presented in a useable form
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer generated randomisation list drawn up by statistician, blinded to study participants and their characteristics, randomly allocated participants into 4 groups
Allocation concealment (selection bias)	Low risk	Statistician, blinded to study participants and their characteristics, randomly allocated participants into 4 groups
Incomplete outcome data (attrition bias) All outcomes	Low risk	Intention‐to‐treat and per protocol analysis
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'yes' or 'no' all main outcome measures reported on
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	Unclear risk	Not possible
Blinding (assessor)	Unclear risk	Insufficient information to permit judgement of 'yes' or 'no'
Comparability of exercise and control group at entry	Low risk	Statistics reported groups equivalent at baseline
Appropriateness of duration of surveillance	Low risk	Immediately postintervention data at 12 months, and one year postcessation of intervention

Kerr 2001.

Methods	Type of study: RCT
Participants	Number of participants randomised = 126 Losses: Retention at 2 years was 71% (59% in the S group, 69% in the F group, and 83% in the C group), Age: mean 60 (6.5) years Setting: Australia Inclusion: more than 4 years past menopause and physically capable of entering exercise groups but who were not already exercising at a moderate intensity more than 2 h/week Exclusion: hormone replacement or other medications or who had diseases known to affect bone density and those who had cardiovascular, physical, or orthopedic disabilities
Interventions	All subjects given 600 mg calcium per day Exercise group (NWBHF) (n = 24): warm‐up consisting of brisk walking and stretching. This was followed by 30 minutes of resistance weight training exercises and progressively increased the loading, wrist curl, reverse curl, biceps curl, triceps pushdown, hip flexion, hip extension, latissimus dorsi pull down, and calf raise Exercise Group (NWBLF) (n = 30): as above but additional stationary bicycle riding with minimal increase in loading Control Group (n = 36): usual activity Duration and intensity: 1 hr sessions 3 x per week 2 years Supervisor: exercise physiologists Supervision: group Setting: gym
Outcomes	BMD hip (total hip, femoral neck, trochanter, Wards triangle) , lumbar spine, and radial forearm
Notes	Compliance/adherence: Exercise compliance was very high in the first 6 months for both groups (S group, 90±12%; F group, 92± 8%) but declined from this point on. In the last 6 months of compliance was 61±23% for the S group and 67±20% for the F group. The average exercise compliance over 2 years was 74±13% in the S group and 77±14% in the F group Adverse events: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Block randomisation to one of three groups
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	'As‐treated' analysis done drop‐outs mentioned but unclear as to which groups
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Unclear risk	Not reported
Comparability of exercise and control group at entry	Low risk	No difference between the groups at baseline
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data 2 years, no follow‐up data reported

Korpelainen 2006.

Methods	Type of study: RCT
Participants	Number of participants randomised = 160 Losses: 68 women (81.0%) in the exercise group and 65 (85.5%) women in the control group completed the study Age: mean age 73 years Setting: Finland Inclusion: hip BMD value of more than 2 SD below the reference value Exclusion: use of a walking aid device other than a stick, bilateral hip joint replacement, unstable chronic illness, malignancy, medication known to affect bone density, severe cognitive impairment and involvement in other interventions
Interventions	Exercise group (COMB) (n = 84): jumping and balance exercises, including walking, knee bends, leg lifts, heel rises and drops, dancing, stamping, stair climbing and stepping up and down from benches Control Group (n = 76): usual activity Duration and intensity: 1hr sessions, 30 months Supervisor: physiotherapist Supervision:group and individual Setting: clinic and home
Outcomes	BMD Radius and hip (total hip, neck of femur, trochanter) During the 30‐month follow‐up, there were 88 falls in the exercise group and 101 falls in the control group (P = 0.10). The incidence of fall‐related fractures was higher in the control group (n = 16) than in the exercise group (n = 6; P = 0.019). One woman in the control group had two fractures, and all other 20 women had one fracture
Notes	Compliance/adherence: Attendance at the exercise sessions averaged 78% during the first supervised 6‐month period, 74% during the second supervised period and 73% during the last supervised 6 months. The average frequency of performing the home exercise programme was three times per week Adverse events: Three women in the exercise group experienced musculoskeletal problems that required minor modifications in the training regimen. 5% level would require 64 women in each group to give an 80% power to detect a 0.02 g/cm2 difference in the primary outcome (femoral neck, trochanter and total hip BMD with an SD of 0.04 g/cm2) between the groups
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated random numbers
Allocation concealment (selection bias)	Low risk	Randomisation provided by a technical assistant not involved in the conduction of the trial
Incomplete outcome data (attrition bias) All outcomes	Low risk	Data were analysed on an intention‐to‐treat basis, and any missing follow‐up data was replaced with the last known value even if this was the baseline value
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Low risk	Operators were unaware of the women’s trial status
Comparability of exercise and control group at entry	Low risk	No significant difference between the groups at baseline
Appropriateness of duration of surveillance	Low risk	Immediately postintervention data 30 months, with follow‐up data reported mean 7.1 years

Lau 1992.

Methods	RCT
Participants	Number of participants randomised = 50 Losses:10 Age: 62‐92 years Setting: China (Hong Kong) Inclusion: Female residents in hostel for elderly with mental function ≥ 6 on Hodkinson Scale Exclusion: metabolic bone disease; diabetes mellitus; previous hip fracture; blood creatinine level > 125 mUmol/l
Interventions	Calcium supplementation group (n = 12) received 800 mg calcium daily Exercise group and placebo (DWBLF) (n = 11): participants stepped up and down 23 cm high block 100 times then exercised upper trunk while standing for 15 minutes Exercise group and calcium supplementation (DWBLF) (n = 15): participants exercised as above and received 800 mg calcium per day Control Group (n = 12): received placebo tablet daily Duration and intensity: 4 times per week for 10 months. Submaximal exertion effort Supervisor: research nurse Supervision: throughout study Setting: not recorded
Outcomes	% change in BMD hip (neck of femur, Wards triangle) and lumbar spine
Notes	Compliance/adherence: not recorded Adverse events:epigastric discomfort (n = 1), and diarrhoea (n = 1) from calcium supplement
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Random permuted blocks
Allocation concealment (selection bias)	Low risk	List prepared in advance and independent of sequence of entry
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	'As‐treated' analysis done drop‐outs mentioned but unclear as to which groups
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk' but exercise is difficult to blind
Blinding (assessor)	Unclear risk	Serial BMD measures were computerised and largely automatic however the operator referred to copies of the first image in subsequent measurements
Comparability of exercise and control group at entry	Low risk	No significant differences observed in baseline characteristics
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 10 months, no follow‐up data recorded

Lord 1996.

Methods	RCT
Participants	Number of participants randomised = 179 Losses: 32 from exercise group, 19 from control group Age: 60‐85 yrs (mean 71.6; SD 5.3) Setting: Australia Inclusion: women who had participated in a previous falls and fractures study, living independently in the community Exclusion: illness or immobility; hospitalisation; medial conditions of neuromuscular, skeletal or cardiovascular system that precluded participation in exercise programme; non English speaking; participating in exercise classes of equivalent intensity to study intervention
Interventions	Exercise group (DWBLF) (n = 90): exercise: 5 warm up period: 35 min conditioning period (aerobic exercise, activities for balance, hand‐eye and foot‐eye coordination and stretching exercises); stretching period 15 min; relaxation 15 minutes Control Group (n = 89 ): no organised activity Duration and intensity:1 hour exercise classes twice weekly for four 10‐12 week sessions for 12 months Supervisor: 3 trained instructors Supervision: at each class Setting: community exercise class
Outcomes	BMD, Lumbar spine, femoral neck, trochanter
Notes	Compliance/adherence: 59.8 (72.9%) 53 participants attended 50 or more classes Adverse events: none
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Complex randomisation procedure, but unclear how randomisation carried out
Allocation concealment (selection bias)	Low risk	Randomisation conducted prior to recruitment
Incomplete outcome data (attrition bias) All outcomes	Low risk	All initial participants accounted for
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk' but exercise is difficult to blind
Blinding (assessor)	Unclear risk	Not recorded
Comparability of exercise and control group at entry	Low risk	No significant differences between groups at entry
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 12 months, no follow‐up data reported

Maddalozzo 2007.

Methods	Type of study: RCT
Participants	Number of participants randomised = 141 Losses: retention rates 83% NHRT plus exercise, 89% HRT plus exercise; 91% HRT no exercise; and 82% control group Age: 52.1± 3.0 years Setting: USA Inclusion: women who had experienced the menopause within the previous 0–36 months from the time of baseline testing as determined retrospectively from questionnaire reports; (2) no menstrual cycles within the previous 12 months without being pregnant, but not longer than 36 months (based on questionnaire recall phone screening interview); (3) follicle‐stimulating hormone levels ≥ 40 mIU/mL (obtained from the subjects physician); (4) body mass index (19–30 kg m−2), (5) 36 months or less of being diagnosed as being postmenopausal by their general physician; and (6) either taking HRT 0.625 mg conjugated equine oestrogen, (Premarin®) or non HRT use Exclusion: non‐HRT users who had taken HRT for 12 consecutive months prior to applying to the study; (2) hypertension; (3) metabolic disease that may affect bone or muscle metabolism (including diabetes and thyroid disease); (4) statin medications for hypercholesterolaemia), multiple sclerosis; and (4) osteoarthritis or other musculoskeletal disorders that prevented participation
Interventions	Non HRT Exercise group (DWBHF) (n = 35): free weight back squat and free weight dead lift exercises repetitions at a speed of 1–2 sets for the concentric (lifting) and 2–3 sets for the eccentric (lowering) phases. Two warm‐up sets of 10–12 repetitions at 50% of 1RM then 3 working sets at 60–75% of 1 RM (set 1 = 8 reps; set 2 = 10 reps; and set 3 = 12 reps) Non HRT Control Group (n = 34) HRT exercise group (DWBHF) (n = 37): as Non HRT Exercise group HRT Control Group (n = 35) Duration and intensity: 50 mins 2 x week for 52 weeks Supervisor: personal trainer Supervision: individual Setting: gym
Outcomes	BMD DXA lumbar spine (L1–L4), proximal femur (total hip, femoral neck, and greater trochanter) and whole body composition
Notes	Compliance/adherence: non‐HRT plus exercise (84.7±12.8%) and HRT plus exercise group (86.2±11.4%) Adverse events: none reported Desired power ≥ 0.8, alpha = 0.05, and an expected difference between groups of 4% increase in muscle mass and a 1% increase in spine BMD, 25 subjects per group were needed
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Self selected as either HRT or non‐HRT replaced then randomised. Randomisation mentioned but insufficient information to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	'As‐treated' analysis done drop‐outs mentioned but unclear as to which groups
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Unclear risk	Not reported
Comparability of exercise and control group at entry	Low risk	No significant differences were observed at baseline on any variable except for spine BMD between HRT and non‐HRT groups
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 52 weeks, no follow‐up data reported

Martin 1993.

Methods	RCT
Participants	Number of participants randomised = 76 Losses: 21, control n = 5, 30 min group n = 7, 45 min group n = 9 Age:49 ‐ 66 years Setting: Florida USA Inclusion: women at least 12 months postmenopause; non‐smoking; white (mainly European descent); no medical or orthopaedic contraindications to exercise; no use of medication known to interfere with calcium metabolism in preceding 12 months; no actively participation in aerobic or strength training programmes in preceding 12 months; no history of intolerance to dairy products; willing to take calcium and vitamin D supplementation; no bony vertebral abnormalities of lumbar or thoracic spine on x‐ray; willing to accept randomisation Exclusion: not recorded
Interventions	All groups received calcium and vitamin D supplementation Exercise group 1 (DWBLF) (n = 27): 30 minute group. Start and end with 3‐5 min of warm up to 60% of max heart rate. Then treadmill to 7% grade (inclination) and to 70% max heart rate in the first 2‐4 weeks, and after to 85% (gradually) for 30 minutes in total Exercise group 2 (DWBLF) (n = 25): start and end with 3‐5 min of warm up to 60% of max heart rate. Then treadmill to 7% grade (inclination) and to 70% max heart rate in the first 2‐4 weeks, and after to 85% (gradually) for 45 minutes in total Control Group (n = 19): calcium and vit D, no exercise Duration and intensity: 3 times a week for 1 year Supervisor: not recorded Supervision: not recorded Setting: not recorded
Outcomes	BMD Lumbar, Proximal forearm, Distal forearm, Body Mass
Notes	Compliance/adherence: Group 1: 77.5 ‐ 79.2 %, Group 2: 85.2 ‐ 82.4% at 0 ‐6 and 6 ‐ 12 months Adverse events: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation reported but insufficient information about the sequence generation process to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	'As‐treated' analysis done, drop‐outs mentioned but different across the groups
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk' but exercise is difficult to blind
Blinding (assessor)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Comparability of exercise and control group at entry	Low risk	No significant differences observed in baseline characteristics
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 12 months no follow‐up data reported

Metcalfe 2001.

Methods	Type of study: RCT
Participants	Number of participants randomised = 301 (266 completed) Losses: 35 Age: 40‐66 average 55.6 years Setting: USA Inclusion: postmenopausal women, sedentary, non smoking, no history of fracture or osteoporosis Exclusion: not reported
Interventions	Calcium tablets administered to both groups (800 mg per day) Non HRT Exercise group (COMB) (n = 177) warm up (5/10 min), progressive weight bearing (25 min) skipping, jogging, jumping, stair climbing with weighted vests. Resistance exercises with large muscle groups (20 min) 70‐80% 1RM Resistance exercises with small muscle groups (10 min), Abdominal strengthening (5 min), Stretching and balance (5 min) Non HRT Control Group (n = 124): usual activity HRT Exercise group (COMB) (n =) : as exercise above HRT control group (n =): usual activity Duration and intensity: 60‐75 minute session 3 non consecutive days per week for 12 months Supervisor: trainer with BSc MSc in exercise science or related field, certification by nationally recognised fitness and strength training organisation, and specifically trained in BEST programme by physical therapist Supervision: ratio of trainers to participants was 1:5 Setting: community fitness facility
Outcomes	BMD spine and hip using dual energy ‐ray absorptiometry Muscle strength
Notes	Compliance/adherence: 35 of exercise group dropped out. Retention rate 80.2% adherence to programme > 70.4% Adverse events: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation mentioned but insufficient information to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Low risk	Losses accounted for in exercise group
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	High risk	Exercise group appeared to have more support with incentive programmes, social interaction and mentoring from trainers
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Comparability of exercise and control group at entry	Unclear risk	No data provided to compare group demographics
Appropriateness of duration of surveillance	Low risk	Immediately postintervention data at 12 months, 4 year follow‐up data reported

Nelson 1994.

Methods	RCT
Participants	Number of participants randomised = 40 Losses: 1 participant from exercise group (suffered MI on holiday during first month of study) Age: 50‐70 yrs Setting: USA Inclusion: at least 5 years postmenopausal; < 70 years of age; not participating in regular exercise programme (no strength training and < 20 mins of aerobic exercise twice per week); weighing less than 130% of ideal body weight; non‐smoking; no more than 1 crush fracture of spine; no history of other osteoporotic fractures; and had not taken oestrogen or other medications known to affect bone for at least 12 months. Exclusion: not recorded
Interventions	Exercise group (NWBHF)(n = 21): 45 min sessions, 3 sets of eight repetition: high intensity strength training (concentric and eccentric contractions: hip extension, knee extension, lateral pull down, back extension, abdominal flexion using pneumatic resistance machine) Control Group (n = 19): they were asked to maintain their current level of physical activity during the year Duration and intensity: 52 weeks (2 weeks off for vacation), 2 times per week, with at least 1 day of rest between sessions Supervisor: exercise trainer Supervision: no more than two participant to each trainer Setting: not recorded
Outcomes	BMD (lumbar spine), femoral neck, Total Body (BMC)
Notes	Compliance/adherence: attendance averaged 87.5% +/‐ 1.8% Adverse events: 7 participants suffered transient musculoskeletal pain requiring minor modification of training regimen, but completed programme
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation reported but insufficient information about the sequence generation process to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Low risk	'As‐treated' analysis done drop‐outs mentioned but unclear as to which groups
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk
Comparability of exercise and control group at entry	Low risk
Appropriateness of duration of surveillance	High risk

Newstead 2004.

Methods	Type of study: RCT
Participants	Number of participants randomised = 53 Losses: 7 ( 2 exercise, 5 control) Age: 50‐65 years Setting: USA Inclusion: no co‐morbidity e.g. diabetes, CHD, PVD, pulmonary or orthopaedic dysfunctions; not taking alendronate medication etc.; no current exercise programme; no history of osteoporotic fractures; BMI 21‐31; on HRT if postmenopausal for >5 years. BMD T‐score > ‐1.5 SD at hip and lumbar spine Exclusion: not reported
Interventions	Exercise group jumping (DWBHF)(n = 25): progressive multidirectional jumping, increasing jump heights and repetitions (max 200) Control Group (n = 28): usual activity Duration and intensity: 3 sessions per week for 12 months Supervisor: physical therapist Supervision: group 2x week, individual 1 x week Setting: gym
Outcomes	BMD femoral neck, total hip, lumbar spine
Notes	Compliance/adherence: average 82% at month 6 and 75% month 12 Adverse events: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation mentioned but insufficient information to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	'As‐treated' analysis done drop‐outs mentioned but unclear as to which groups
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Unclear risk	Not reported
Comparability of exercise and control group at entry	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data 12 months, no follow‐up data reported

Papaioannou 2003.

Methods	Type of study: RCT
Participants	Number of participants randomised = 74 Losses: 14 lost at 6/12 (n = 60), 3 lost at 12/12 (n = 57) Age: mean age 71.6 (SD = 7.33) exercise group, 72.2 (SD = 7.98) in control. No significant differences in drop outs between groups Setting: Canada Inclusion: postmenopausal women 60yr + with osteoporosis (lumbar BMD >= 2.5 SD below young adult mean) and at least one vertebral fracture Exclusion: vertebral fracture within last 3 months; secondary causes of bone loss; other diagnosis for back pain; resting heart rate > 100 beats per min and uncontrolled hypertension; unable to stand independently for 3 min
Interventions	Exercise group (COMB) (n = 37): exercise programme detailed in manual with diagrams, comprised stretching, strength training upper and lower limbs and aerobics Control Group (n = 37): Duration and intensity: 60 minutes of exercise over the course of the day, 3 days per week, with 1 rest day between for 12 months Supervisor: exercise Therapist. No further details Supervision:exercise group ‐ monthly visits for first 6/12 with programme review. Follow‐up call every 2 weeks to 12 months. Control ‐ telephone contact by exercise therapist every month for 12 months Setting: home
Outcomes	DXA BMD lumbar spine (L2‐4) femoral neck at baseline and 12 months
Notes	Compliance/adherence: defined as completing 3 sessions per week at least 80% of weeks. 62% participant’s adherent Adverse events:  not reported No data reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation mentioned but insufficient information to permit judgement of 'Yes' or 'No'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	No data presented for BMD other than no differences over 12 months between the groups
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Low risk	Research assistant and BMD investigator blinded
Comparability of exercise and control group at entry	Low risk	Baseline variables not significantly different between groups
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 12 months, no follow‐up data reported

Preisinger 1995.

Methods	RCT
Participants	Number of participants randomised = 146 Losses: not reported Age: 45‐75 yrs Setting: Australia Inclusion: caucasian; 45‐75 yrs of age; postmenopausal at least 1 year; did not suffer from malabsorption or other chronic diseases; non smoking; not taking oestrogen, other steroid hormones, anticonvulsants or thiazide diuretics; sedentary lifestyle; and normal blood results from described list. Exclusion: retrospective exclusion of women not attending tests or who commenced drug treatment for osteoporosis during follow‐up period
Interventions	Exercise group (DWBLF) (n = 82): 1a (n = 39): warm up of brisk walking, modest jogging, arm swings and moderate skill exercises, stretching exercises hip and leg muscles, and complex resisted exercises to train movement patterns (diagonal or diagonal spiral movements) using elastic bands and gymnastic balls 1b (n = 43): stopped exercise treatment, performed it irregularly or less than 1 h per week Control Group (n = 64): no therapy Duration and intensity: at 3 times per week for 20 mins. Resisted exercises described as using considerable energy Supervisor: qualified therapist Supervision: 20 times over initial 10 week period, then 5 times every subsequent six months Setting: not recorded
Outcomes	BMD, SPA, Radium, Proximal
Notes	Compliance/adherence: 48% performed exercises regularly for the prescribed time Exercise group split retrospectively based on interviews and review of records at follow‐up visit to 1a, 1b Adverse events: none reported No baseline data for lumbar spine and femoral neck presented to enable % change to be calculated
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation mentioned but insufficient information to permit judgement of 'Yes' or 'No'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'Yes' or 'No'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Insufficient reporting of drop‐outs from beginning of the study
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'Yes' or 'No'
Other bias	Unclear risk	Poorly reported study ‐ difficult to assess potential biases
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Low risk	Assessments were made by same investigator who was unaware of BMD measurements
Comparability of exercise and control group at entry	Low risk	No significant differences between groups at entry
Appropriateness of duration of surveillance	Low risk	Immediately postintervention and six monthly follow‐up over ten year period

Prince 1991.

Methods	RCT
Participants	Number of participants randomised = 120 Losses: 17, exercise group n = 6, exercise calcium group n = 3, exercise oestrogen group n = 8 Age: mean 56±4 yrs Setting: Australia Inclusion: women with low forearm bone density; > 43 years of age; postmenopausal for 1‐10 years; without hypertension or chronic diseases; not taking oestrogen, steroid hormones, anticonvulsants or thiazide diuretic drugs Exclusion: women with bone density not more than 1SD below mean for premenopausal women. Control group was randomly assigned from this group
Interventions	Exercise + placebo group (DWBLF) (n = 41): weekly class consisting of 1 hour low impact aerobics of which 30% of time devoted to arm exercises. Twice weekly 30 min brisk walk Exercise + calcium supplementation (n = 39): a/a Exercise + oestrogen supplementation (n = 40); a/a Control Group (n = 40): no exercise or placebo Duration and intensity: 1 hour class x 1 per week, 30 mins brisk walking x 2 per week for 2 years Supervisor: trained physiotherapist Supervision: during exercise class only Setting: not recorded
Outcomes	BMD forearm measured every 3 months
Notes	Compliance/adherence: 56% exercise only group; 24% exercise + calcium group; 44% exercise = oestrogen group attended a minimum of 10 classes in any 12 week period Adverse events: flushing, breast tenderness, sleeplessness etc.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Shuffling of sealed envelopes
Allocation concealment (selection bias)	Low risk	Only pharmacist knew assignments
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Unclear why so many of the exercise groups dropped out or did not complete sufficient exercise classes
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'Yes' or 'No'
Other bias	High risk	Drug company Upjohn Australia supplied medication and supported research
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Unclear risk	Insufficient information to permit judgement of 'Yes' or 'No'
Comparability of exercise and control group at entry	High risk	Control group had normal BMD, and had been postmenopausal for less time (5.8±2.6 v 4.5±2.1)
Appropriateness of duration of surveillance	Low risk	Only immediately postintervention data at 2 years, no follow‐up data recorded

Prince 1995.

Methods	RCT
Participants	Number of participants randomised = 168 Losses: not recorded Age: 50‐70 yrs Setting: Australia Inclusion: 50‐70 yrs of age, > 10 yrs postmenopausal Exclusion: significant chronic diseases, had received oestrogen, other steroid hormones, anticonvulsants, thiazide diuretic drugs or medications that could influence calcium metabolism
Interventions	Exercise group + calcium (DWBLF) (n = 42): 2 x 1 hour  supervised classes comprising weight bearing exercise (not specified) Calcium group (n = 42): no exercise Milk powder group (n = 42): no exercise Control Group (n = 42): placebo medication only Duration and intensity: 4 hours per week, at 60% of peak maximal heart rate for age for 2 years Supervisor: not recorded Supervision: supervision during exercise class Setting: not stated
Outcomes	BMD lumbar spine, hip, distal tibia/fibula at baseline, 6, 12 18 and 24 months
Notes	Compliance/adherence: 39% of exercise group exercised for 3 hours per week at 60% peak heart rate Adverse events: not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Block randomisation with sealed envelopes prior to study commencing
Allocation concealment (selection bias)	Low risk	Sealed envelopes
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	'As‐treated' analysis done no drop‐outs mentioned and no indication of numbers in each group
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	High risk	Drug company supplied placebo and calcium tablets
Blinding (participant)	High risk	Not possible
Comparability of exercise and control group at entry	Low risk
Appropriateness of duration of surveillance	High risk

Pruitt 1996.

Methods	RCT
Participants	Number of participants randomised = 40 Losses: 14 Age: 65‐82 years Setting: America Inclusion: Healthy caucasian women not currently taking HRT, or those on HRT for 1 year or more Exclusion: evidence of acute or uncontrolled chronic illness or conditions that would prevent participation in exercise class, vertebral compression fractures, disorders affecting bone metabolism
Interventions	Exercise group 1 (NWBHF) (n = 15): supervised exercise session comprising bench press, lateral pull down, military press, biceps curl, knee extension, knee flexion, hip abduction and adduction, leg press, back extension. 1 set 14 reps at 40% 1RM, 2 sets 7 reps at 80% 1RM Exercise group 2  (NWBLF) (n = 13): a/a 3 sets 12 reps at 40% 1RM Control Group (n = 12): no exercises Duration and intensity: 3 times per week for 12 months, lifting time 50 ‐55mins. 1RM tests administered every 2 weeks for first 3 months then every 3 weeks to adjust workload Supervisor: not recorded Supervision: every session Setting: Gym
Outcomes	BMD lumbar spine, hip (total hip, neck of femur, Wards triangle) at baseline and 12 months
Notes	Compliance/adherence: 65% Adverse events: aggravation of pre‐existing back or knee condition (n = 2)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation reported but insufficient information about the sequence generation process to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	'As‐treated' analysis done drop‐outs mentioned but different across the groups
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'. One outlier whose spinal BMD was more than 4SD from group mean was not included in analysis
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Comparability of exercise and control group at entry	Low risk	No significant differences observed in baseline characteristics
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 12 months, no follow‐up data reported

Revel 1993.

Methods	RCT
Participants	Number of participants randomised = 78 Losses: 11 withdrew but were not lost to follow‐up (treatment group n = 6, control group n = 5) Age: 54±3 yrs Setting: France Inclusion: recruited from pension fund membership. Healthy postmenopausal caucasian women, postmenopausal for between 1 ‐ 12 years (mean = 6±3). Not on oestrogen or oestrogen like compounds, corticosteroids, fluoride salts, diphosphonate's or calcitonin Exclusion: not reported
Interventions	Exercise group (NWBLF) (n = 39): performed 60 repetitions of active hip flexion in sitting with 5 kg sandbag on knee. Hip flexion limited to 30 degrees. Could be performed over 2 or 3 sessions Control Group (n = 39): deltoid training, no further details Duration and intensity: 1 year, 60 repetitions daily Supervisor: not recorded Supervision: exercises taught initially no further supervision recorded Setting: not recorded
Outcomes	TBMD  L1/L4
Notes	Compliance/adherence: 55% fully completed the training programme. 5 participants withdrew as they found the study too constraining (treatment group n = 4, control n = 1) Adverse events: treatment group hip pain (n = 1), control group shoulder pain (n = 3, back pain (n = 1) Follow‐up data not reported by group
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation mentioned but insufficient information to permit judgement of 'Yes' or 'No'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'Yes' or 'No'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Outcome reported for all that completed the trial. Authors say they did an ITT analysis on all but 5 of the participants, but none of the tables have figures that match this number
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'Yes' or 'No'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not feasible for exercise programmes
Blinding (assessor)	Low risk	Radiologist who performed scans was blinded to randomisation result
Comparability of exercise and control group at entry	Low risk	No significant differences observed in baseline characteristics
Appropriateness of duration of surveillance	Low risk	Only immediately postintervention data at 12 months, follow‐up data at 2 years postintervention reported

Rubin 2004.

Methods	RCT
Participants	Number of participants randomised = 70 Losses: 6 (1 active, five placebo) withdrew within first 3/12 and were replaced by new subject in same treatment type Age: 47‐64 years Setting: USA Inclusion: 3‐8 years postmenopausal women, normal nutritional status, stable weight, estimated daily Ca intake ≥ 500 mg daily, capable of following protocol, body mass 45 kg‐84 kg Exclusion: any pharmacological intervention for osteopenia within last 6 months, steroid use, current smoking status, consumption of excessive alcohol, evidence of osteomalacia, osteogenesis imperfect, GI disease, history of malignancy, and/or prolonged immobilisation of axial or appendicular skeleton within last 3 years, spondyloarthrosis, thyrotoxicosis, psychomotor disturbances, hyperparathyroidism, renal or hepatic disease, chronic diseases known to affect muscular system, and/or engaged in high impact activities at least 3 x per week
Interventions	Exercise group (DWBHF) (n = 33): vibration plate that vibrated at 30 Hz, 0.2 g peak to peak Control Group (n = 37): placebo device, protocol a/a Duration and intensity: 2 x 10 mins treatments per day separated by a minimum of hrs, 7 days per week for 1 year Supervisor: none Supervision: none Setting: home
Outcomes	BMD by DXA R & L femur, lumbar spine, distal 1/3 radius at baseline, 3, 6 and 12 months
Notes	Compliance/adherence: 37% completing study were at least 80% compliant (10 active, 7 placebo), 72% at least 60% compliant (19 active, 14 placebo) Adverse events:  1 person (placebo group) reported headache Data presented as % change but as a function of compliance > 80% with exercise and only mean values presented
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Confidential randomised number sequence
Allocation concealment (selection bias)	Low risk	Generated by individual statistical consultant
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Losses accounted for and analysis based on the 56 subjects who completed the study and were scanned at end of study
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'Yes' or 'No'
Other bias	High risk	Research funded by inventors of device
Blinding (participant)	Low risk	Each device emitted same low frequency sound, all participants insulated from each other at home
Blinding (assessor)	Low risk	Randomised code broken on completion of study
Comparability of exercise and control group at entry	High risk	Significant differences in body weight and BMI with placebo 5 kg heavier than exercise group (P = 0.03)
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 12 months, no follow‐up data reported

Russo 2003.

Methods	Type of study: RCT
Participants	Number of participants randomised = 33 Losses: 6 at randomisation, 3 at follow‐up (1 control, 3 exercise) Age: mean (SE) exercise 60.7 (6.1) and control 61.4 (7.3) years Setting: USA Inclusion:1 year postmenopausal Exclusion: metabolic bone disorders, conditions contraindicating vibration training
Interventions	Exercise group (DWBHF) (n = 17): vibrating plates lateral oscillations 0.1‐10 g. Progressive frequency up to 28 Hz up to 2 mins duration Control Group (n = 16): usual activity Duration and intensity: 2 x week for 6 months Supervisor: not stated Supervision: individual Setting: gym
Outcomes	BMD Trabecular volumetric bone density (mg/cm3) Cortical volumetric bone density (mg/cm3)
Notes	Compliance/adherence: not reported Adverse events: transient, slight lower leg itching and erythema, was also observed in 6 of 17 treated participants in this study. In no case, however, did this problem persist after the first 3 training sessions or cause interruption of the intervention. Knee pain of moderate intensity, without objective clinical signs, was observed in 2 overweight participants with pre‐existing knee osteoarthritis. The pain subsided in both participants after a few days of rest
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation by random number table
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	'As‐treated' analysis done drop‐outs mentioned
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk
Blinding (participant)	High risk
Comparability of exercise and control group at entry	Low risk
Appropriateness of duration of surveillance	High risk

Sakai 2010.

Methods	Type of study: RCT
Participants	Number of participants randomised = 94 Losses: 16 (3 exercise, 13 control) Age: mean age 68.3 years (61‐85) Setting: Japan Inclusion: not stated Exclusion: not reported
Interventions	Exercise group (SWB) (n = 49): single leg standing Control Group (n = 45): usual activity Duration and intensity: 1 min per leg, 3 x per day for 6 months Supervisor: not stated Supervision: home exercise Setting: home
Outcomes	BMD DEXA neck, trochanter, intertrochanter, Ward's triangle
Notes	Compliance/adherence: not reported Adverse events: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation by envelopes
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	'As‐treated' analysis done drop‐outs mentioned but different across the groups
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Unclear risk	Not reported
Comparability of exercise and control group at entry	Low risk	There were no significant differences in age, body height, body weight, body mass index and hip BMD between the 2 groups at baseline
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data 6 months, no follow‐ up data reported

Sinaki 1989.

Methods	RCT
Participants	Number of participants randomised = 68 Losses: 3 from control group Age: 49‐65 yrs Setting: America Inclusion: postmenopausal, 49‐65 yrs of age, normal diet, without calcium, vitamin D or oestrogen supplementation. With normal ECG, blood results and urine analysis Exclusion: baseline BMD below 5th percentile of normal range
Interventions	Exercise group (NWBLF) (n = 34): back extension exercises performed in prone against resistance using backpack weighted to the equivalent of 30% of maximum isometric strength up to a maximum of 50 lb (22.7 kg) Control Group (n = 34): no active exercise Duration and intensity: 10 repetitions once a day, five days per week for 2 years Supervisor: not recorded Supervision: at outset only Setting: home
Outcomes	BMD lumbar spine at baseline, 6, 12, 18 and 24 months
Notes	Compliance/adherence: not recorded Adverse events: none recorded
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation reported but insufficient information about the sequence generation process to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Low risk	'As‐treated' analysis done, all 3 drop‐outs were in control group
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Low risk	Assessor had no knowledge of group allocation
Comparability of exercise and control group at entry	Unclear risk	No significant differences observed in baseline characteristics apart from total serum calcium and total thyroxine
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 2 years, no follow‐up data reported

Smidt 1992.

Methods	RCT
Participants	Number of participants randomised = 55 Losses: 6; not clear which groups these came from Age: control group = 55.4±8, exercise group 56.6±6.6 Setting: America Inclusion: physician consent, no current medical history (within last 12 months) of low back pain, kidney, cardiac, neuromuscular or musculoskeletal dysfunction. No current involvement in weight training programme for abdominals or back extensors, no obesity that preclude ability to use trunk testing equipment, at least one year postmenopause Exclusion: not recorded
Interventions	Exercise group (NWBHF) (n = 22): 3 sets of 10 repetitions of sit ups, prone trunk extension and double leg flexion (i.e. 90 reps in total) at 70% of maximal strength test, increasing by 2‐5% monthly depending on ability Control Group (n = 27): maintain current lifestyle Duration and intensity: 3 to 3 times per week for 12 months Supervisor: not recorded Supervision: at outset and once per month Setting: home
Outcomes	BMD lumbar spine L2‐4, hip (neck of femur, trochanter) at baseline, 6 and 12 months
Notes	Compliance/adherence: 11 participants performed exercises 3 times per week, 9 exercised 2‐3 times Adverse events: none recorded Bad luck with randomisation: control group turned out to be very physically active
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation mentioned but insufficient information to permit judgement of 'Yes' or 'No'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'Yes' or 'No'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Insufficient information to permit judgement of 'Yes' or 'No'
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'Yes' or 'No'
Other bias	Low risk	The study appears to be free of other source of bias
Blinding (participant)	High risk	Not practical for exercise classes
Blinding (assessor)	Unclear risk	Not reported
Comparability of exercise and control group at entry	Low risk	No significant difference between groups at entry
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 12 months, no follow‐up data reported

Tolomio 2009.

Methods	Type of study: RCT
Participants	Number of participants randomised = 160 Losses: 17. 16 failed to attend first session, 1 did not return following first session, 6 did not attend BMD scanning. 23 lost in exercise group, 12 in control group Age: postmenopausal (no other information recorded) Setting: Italy Inclusion: Postmenopausal women with osteoporosis or osteopenia Exclusion: orthopaedic operations
Interventions	Exercise group (n = 81) (COMB): 15 minutes warm up: walking at reasonable pace, joint movement, balance exercises, stretches. 30 minutes diverse exercises depending on objective i.e. strength, balance or flexibility using weights, balls, theraband and steps. 15 minutes cool down as warm up but lighter Control Group (n = 79): no exercises Duration and intensity: 44 weeks ‐ 60 minutes 3 x per week at gym for 11 weeks; 1x per week at gym; and 2x per week in thermal water (spa) for 17 weeks; then 3x per week home exercises Supervisor: unclear Supervision: not clear for first two stages of trial. Telephone checks with advice for home exercise programme Setting: gym; gym/spa; home
Outcomes	BMD femoral neck and total hip
Notes	Compliance/adherence: not reported Adverse events: none recorded Data converted to % change
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation mentioned but insufficient information to permit judgement of 'Yes' or 'No'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'Yes' or 'No'
Incomplete outcome data (attrition bias) All outcomes	Low risk	All participants accounted for including drop‐outs
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'Yes' or 'No'
Other bias	Unclear risk	Insufficient information to permit judgement of 'Yes' or 'No'
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Unclear risk	Insufficient information to permit judgement of 'Yes' or 'No'
Comparability of exercise and control group at entry	Low risk	No significant difference between groups at entry
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 11 months, no follow‐up data reported

Uusi‐Rasi 2003.

Methods	Type of study: RCT
Participants	Number of participants randomised = 164 Losses: 7 (5 exercise and 2 control) Age: exercise mean 53.3 (2.2), control 53.2 (2.1) years Setting: Finland Inclusion: 1–5 years postmenopause; no previous bone fractures; neither current nor previous use of oestrogen, corticosteroids, bisphosphonates, nor other drugs, nor illness affecting bone metabolism; no contraindication to exercise or alendronate; previous regular exercise less than two times a week; femoral neck BMD 0.650 g/cm2 and an FSH level greater than 30 IU/L Exclusion: less than 1 year or more than 5 years postmenopause; history of chronic illness; evidence of metabolic bone disease or use of bone‐specific medications; concurrent serious medical conditions including sepsis or disseminated cancer; abnormalities of the oesophagus; inability to stand or sit upright for at least 30 min; hypersensitivity to any component of the study drug; and hypocalcaemia
Interventions	Non alendronate Exercise group (DWBHF) (n = 41): placebo plus warm‐up, 20 min of multidirectional jumping exercises, 15 min of callisthenics (stretching and non impact exercises), and 10 min cool down. The programme was progressive peak forces varied between 2.1 and 5.6 times body weight Non alendronate Control Group (n = 41): placebo usual activity Alendronate exercise group (DWBHF) (n = 41): as exercise above Alendronate control group (n = 41): usual activity Duration and intensity: 1 hr 3 x week for 1 year Supervisor: experienced exercise leaders of the UKK Institute Supervision: group Setting: gym
Outcomes	The bone mineral content (BMC, g) and areal bone mineral density (BMD, g/cm2) of the lumbar spine, right proximal femur (femoral neck and trochanter area of the femur), and nondominant distal radius
Notes	Compliance/adherence: Mean (SD) compliance in the exercise group, defined as attendance in the training sessions, was 1.6 ∓0.9 times per week Adverse events: 19 subjects from the exercise group consulted the attending physician (P.K.) due to musculoskeletal injuries or symptoms; 1 subject had an acute severe ankle sprain requiring surgical treatment. The rest were mild overuse symptoms; 1 subject with a mild knee distortion injury; 5 subjects with an overuse problem at the knee joint (3 with chondromalacia patellae and 2 with unspecific knee pain); 4 with an overuse problem at the foot (2 with an insertional tendinopathy of the Achilles tendon and 2 with unspecific foot pain); 2 with low back pain (1 sciatica, 1 unspecific); 2 with hip pain (1 trochanteric bursitis, 1 unspecific); 2 with shoulder pain (both supraspinatus tendinitis); and 2 with unspecific fibromyalgia (tension neck symptoms)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation mentioned but insufficient information to permit judgement of 'high risk' or 'low risk'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Low risk	Intention‐to‐treat analysis
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk' but states double‐blind
Blinding (assessor)	Low risk	Outcomes assessors blinded to treatment group allocation
Comparability of exercise and control group at entry	Low risk	There were no clinically relevant differences between groups
Appropriateness of duration of surveillance	Low risk	Immediately postintervention data 12 months and follow‐up data at 15 months postintervention reported

Verschueren 2004.

Methods	Type of study: RCT
Participants	Number of participants randomised = 70 Losses: not reported Age: 58–74 years Setting: Belgium Inclusion: 60 and 70 years of age, non‐institutionalised, and free from diseases or medications known to affect bone metabolism or muscle strength Exclusion: total body BMD T‐score of less than ‐2.5
Interventions	Exercise group vibrating platform (DWBHF)(n = 25): static and dynamic knee‐extensor exercises on the vibration platform, progressive exercise Exercise group resistance training (NWBHF)(n = 22): warm‐up, resistance training programme for knee extensors on a leg extension and a leg press machine. Training programme was designed (ASCM) for individuals older than 60 years of age. Progressive resistance Control Group (n = 23): usual activity Duration and intensity: 72 training sessions within a 24‐week period. Training frequency was three times a week Supervisor: not stated Supervision: individual and group for resistance training Setting: gym
Outcomes	BMD DEXA total hip, total body
Notes	Compliance/adherence: not reported Adverse events: none reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation by computer‐generated random numbers age‐matched women
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	'As‐treated' analysis done, insufficient information to permit judgement of 'high risk' or 'low risk'
Selective reporting (reporting bias)	Unclear risk	Insufficient information to permit judgement of 'high risk' or 'low risk'
Other bias	Low risk	The study appears to be free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Low risk	Technician unaware of intervention type
Comparability of exercise and control group at entry	Low risk	No significant differences were observed at baseline between the experimental and the control groups in terms of age, weight, body mass, years since menopause, BMD, serum levels of osteocalcin and CTX, isometric and dynamic muscle strength, fat mass or lean body mass
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data, no follow‐up data reported

Von Stengel 2009.

Methods	Type of study:RCT
Participants	Number of participants randomised = 151 Losses: 16, group 1 n = 5, group 2 n = 7, control n = 4. All invited for final measurements, 11 did not attend; group 1 n = 1, group 2 n = 6, control n = 4 Age: 65‐72 years Setting: Germany Inclusion: Over 65, postmenopausal Exclusion: relevant co‐morbidity or drug treatment which could influence bone metabolism
Interventions	Exercise group (COMB) (n = 50): Low impact aerobics, strengthening exercises and balance Exercise group (COMB) (n = 50): Low impact aerobics, strengthening and balance exercise as above and vibration plate. Vibration between 25‐35 Hz, intensity increased at 3 and 6 months Control Group (n = 51): gentle exercise and relaxation class x 1 per week Duration and intensity: 60 minutes 2x per week for 12 months Supervisor: not reported Supervision: not reported Setting: hospital
Outcomes	BMD total hip and spine, rate of falls
Notes	Compliance/adherence: not reported Adverse events: none recorded Selected exercise group with vibration plate for analysis. Data converted to % change
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Randomisation mentioned but insufficient information to permit judgement of 'Yes' or 'No'
Allocation concealment (selection bias)	Unclear risk	Insufficient information to permit judgement of 'Yes' or 'No'
Incomplete outcome data (attrition bias) All outcomes	Low risk	Losses explained and data analysed on intention‐to‐treat
Selective reporting (reporting bias)	Low risk	Reporting as per protocol
Other bias	Low risk	The study appears to free of other sources of bias
Blinding (participant)	High risk	Not possible
Blinding (assessor)	Unclear risk	Not reported
Comparability of exercise and control group at entry	Low risk	No significant differences between groups at entry
Appropriateness of duration of surveillance	High risk	Only immediately postintervention data at 12 months, no follow‐up data reported

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Ay 2005	No BMD DEXA, only US measures
Bebenek 2010	Control group underwent low intensity exercise
Bemben 2010	Subjects were assigned to a group based on their availability to attend the scheduled training sessions
Cao 2009	No BMD DEXA, only US measures
De Matos 2009	Not RCT subjects, selected group
Engelke 2006	Not RCT subjects, selected group
Hans 2002	Not an RCT
Hawkins 2002	Participants allocated to groups based on proximity to laboratory
Heinonen 1996	Pre‐menopausal participants
Kemmler 2003	Not an RCT
Kerr 1996	Participants were their own control, one side of body randomised to a different exercise type
Kerschan‐Schindl 2000	Randomisation not mentioned, described as observational study
Kohrt 1995	Controlled trial, not an RCT
Kontulainen 2004	Pre‐menopausal participants
Kriska 1986	No BMD outcomes
Leichter 1989	Not an RCT or CCT (before/after study)
Lohman 1995	Pre‐menopausal participants
Mayoux‐Benhamou 1995	Duplicate publication, French version
Nelson 1991	No control group
Notelovitz 1991	Surgical menopause, no control group
Pruitt 1992	Not an RCT
Rikli 1990	Controlled trial, not an RCT
Ruan 2008	Not RCT
Shen 2009	No outcome measures for BMD, only bone formation biomarkers
Snow 2000	Original study not an RCT
Song 2010	Participants had osteoarthritis
Tolomio 2008	No BMD DEXA, only US measures
Uusi‐Rasi 2005	Follow‐up of peri‐menopausal women
Villareal 2003	Not an RCT
White 1984	Not an RCT
Xu 2004	Not RCT
Yamazaki 2004	Not RCT. Group assignment according to the wish of the participants

Characteristics of studies awaiting assessment [ordered by study ID]

Ilona 2010.

Methods	Type of study: RCT
Participants	Number of participants randomised = 46 Losses: not recorded Age: 43‐65 years Setting: Romania Inclusion: postmenopausal women diagnosed with osteoporosis or osteopenia Exclusion: concurrent orthopaedic or neurological disorders
Interventions	Exercise group (NWBLF) (n = 23): exercise, diet (diary products and veg), medication (Fosamax, Ca supplements, Vit D) Control Group (n = 23 ): diet (dairy products and veg), medication (Fosamax, Ca supplements, Vit D) Exercise group and controls well matched Duration and intensity: twice per week, 1 hour for 12 months. 8 ‐ 10 repetitions with 1 minute between sets initially, rising to 12‐15 times higher by end of intervention period Warm up 10 min, static stretches, walking, deep breathing, easy running Strength exercise 40 min low load, high repetition exercise for upper limbs performed in sitting, and callisthenic in supine and standing Cool down, 10 min (not reported) Supervisor: physiotherapist, experience not recorded Supervision: close physiotherapist surveillance Setting: not recorded ** paper classifies study as measuring the effect of high impact exercise but impact appears to occur only in warm up phase, clarification on the researchers definition of callisthenic exercise would be helpful
Outcomes	BMD , T score lumbar spine DEXA (L1 ‐4) baseline and 12 months
Notes	Compliance/adherence: not recorded Adverse events: not recorded

Karaarslan 2010.

Methods	Type of study:
Participants	Number of participants randomised = Losses: Age: Setting: Inclusion: Exclusion:
Interventions	Exercise group (n =): Control Group (n = ): Duration and intensity: Supervisor: Supervision: Setting:
Outcomes
Notes	Compliance/adherence: Adverse events: awaiting full paper

Kemmler 2004a.

Methods	Type of study:
Participants	Number of participants randomised = Losses: Age: Setting: Inclusion: Exclusion:
Interventions	Exercise group (n =): Control Group (n = ): Duration and intensity: Supervisor: Supervision: Setting:
Outcomes
Notes	Compliance/adherence: Adverse events: awaiting full paper

Characteristics of ongoing studies [ordered by study ID]

Wayne 2010.

Trial name or title	not known
Methods	Type of study: Pragmatic RCT
Participants	Number of participants randomised = 86 Losses: Age: 45‐70 Setting: America Inclusion: women aged 45‐70; postmenopausal > 12 months; BMD T scores of hip and/or spine between ‐1.0 and 2.5; does not exercise more than 5 days a week on average for more than 60 minutes per day Exclusion: osteoporotic (T‐score < ‐ 2.5) at any site or a fracture in last 2 years not caused by road traffic collision; prior or current use of: oestrogen or calcitonin (within last year); medication that increases risks of fracture (e.g. steroids, anti‐convulsants, anticoagulants, lithium); medications that modify bone metabolism; use of calcium supplements above 1200‐1500 mg; malignancies other than skin cancer; diagnosis of anorexia along with BMI < 17.5; conditions causing secondary osteoporosis; tobacco use in past year; physical or mental disabilities that preclude informed consent in participation; geographical or scheduling limitations that preclude weekly participation in exercise class and study; current regular practice of Tai Chi
Interventions	Exercise group (DWBLF) (n =): Control Group (n = ): Duration and intensity: 9 months. Minimum 2 classes (60 min each session) and 2 additional practice sessions (min 30 min) per week in first month, then a minimum of 1 per week and 3 practice sessions for remaining 8 months Supervisor: classes led by junior instructor supervised by senior instructor Supervision: at class Setting: Tia Chi, school and home
Outcomes	BMD lumbar spine and proximal femur assessed by DEXA Compliance/adherence: Adverse events:
Starting date	Not recorded
Contact information	peter_wayne@hms.harvard.edu
Notes	Registered with Clinical Trials.gov, ID number NCT01039012

Contributions of authors

TEH ‐ designed and reviewed the protocol for the updated review; extracted and entered data and assessed quality; conducted methodological analysis; and wrote and reviewed the manuscript.

BS ‐ designed and reviewed the protocol for the review; extracted data and assessed quality; conducted methodological analysis; and wrote and reviewed the manuscript.

LD ‐ applied the inclusion and exclusion criteria for accepting studies in the review; classified the studies; helped with classifying the interventions; extracted and entered data and assessed quality; and contributed to and reviewed the manuscript.

FD ‐ applied the inclusion and exclusion criteria for accepting studies in the review; classified the studies; helped with classifying the interventions; extracted data and assessed quality; and contributed to and reviewed the manuscript.

AM ‐ applied the inclusion and exclusion criteria for accepting studies in the review; classified the studies; helped with classifying the interventions; extracted data and assessed quality; and contributed to and reviewed the manuscript.

CR ‐ applied the inclusion and exclusion criteria for accepting studies into the review; classified the studies; helped with classifying the interventions; extracted data and assessed quality; and provided critical comments on drafts of the review.

RH ‐ contributed methodological expertise and reviewed early drafts of the document.

LC ‐ wrote the search strategies and identified the literature.

GS ‐ applied the inclusion and exclusion criteria for accepting studies into the review; classified the studies; helped with classifying the interventions; extracted data and assessed quality; and reviewed earlier drafts of the review; provided translation into English.

Sources of support

Internal sources

• University of Ottawa, Canada.

• Clinical Epidemiology Unit, Ottawa Hospital, Civic Campus, Canada.

• Institute of Population Health, Canada.

• Glasgow Caledonian University, UK.

External sources

• The Chartered Society of Physiotherapy, UK.

• National Institute for Health Research, UK.

  Cochrane Incentive Award 2010

Declarations of interest

None known

New search for studies and content updated (conclusions changed)