Skip to main content

Table 4 Description of included studies investigating the association between physical activity and osteoporosis employing an observational design

From: Evidence on physical activity and osteoporosis prevention for people aged 65+ years: a systematic review to inform the WHO guidelines on physical activity and sedentary behaviour

Reference

Overall Risk of Bias

Study design Included / Analysed

Participants (n, age mean (SD), % women, setting, health status)

Exposure

Outcomes

Results

aFoley 2010

(Tasmania Older

Adult Cohort study)

Overall risk of bias: High [81]

Prospective cohort study

(2.6 years follow-up)

875

Setting: Community, Australia

Health status: Healthy

n= 875 (included)

Age: 62.7 (7.3) (included)

[NB: quartile 1, 2 and 3 are not of interest in this study as mid-point age <65 years]

Female: 49%

Mid-point of age quartile 4:

Age: 74.8

Ambulatory activity (steps per day) was assessed using pedometer for 1 week at both baseline and follow-up and participants were divided in quartiles

Classification: total physical activity

1. Lumbar spine areal BMD

2. Hip areal BMD

Adjusted point estimates (95% CIs)

1. Lumbar spine areal BMD

Not reported for sample >65 years

2. Hip areal BMD at follow-up

Age quartile 4, Female

Q1. 0.434 (0.372 to 0.497)

Q2. 0.441 (0.378 to 0.503)

Q3. 0.446 (0.383 to 0.509)

Q4. 0.466 (0.401 to 0.532)

Age, quartile 4, Male

Q1. 0.554 (0.490 to 0.618)

Q2. 0.566 (0.501 to 0.631)

Q3. 0.572 (0.507 to 0.637)

Q4. 0.584 (0.518 to 0.651)

aMuir 2013

(Canadian multicentre osteoporosis study)

Overall risk of bias: Low [84]

Retrospective (previous 12 months)

Setting: Community; Canada

Health status: Mixed (included participants with diagnosis of osteoporosis)

n= 1169

Age: 79.84 (4.43)

Female: 100%

Physical activity was quantified based on the level of activity (moderate or strenuous or vigorous) and the reported frequency and duration of said activity over the course of the previous 12 months.

Classification: total physical activity

1. Lumbar Spine (L1-4) BMD

2. Femoral neck BMD

3. Total hip BMD

4. Ward’s triangle BMD

5. Trochanter BMD

Multiple regression analysis of the relative effects of moderate activity on BMD. Coefficient (95% CI)

1. Lumbar Spine (L1-4) BMD

-0.006 (-0.013 to 0.000)

2. Femoral neck BMD

0.004 (0.000 to 0.008)¥

3. Total hip BMD

0.006 (0.001 to 0.011)¥

4. Ward’s triangle BMD

0.004 (-0.001 to 0.009)

5. Trochanter BMD

0.005 (0.006 to 0.074)¥

aNakamura 2012

(Muramatsu Study)

Overall risk of bias: High [85]

Cohort

(6 years follow-up)

774/382

Setting: Community; Japan

Health status: Healthy

n= 382

Age: 73.3 (3.7)

Female: 100%

Physical activity was assessed via questionnaire based on whether participants regularly engaged in light or moderate physical activity (yes/no) activities:

A. Light physical activity (includes croquet, taking walks and traditional Japanese dancing): yes or no

B. Moderate physical activity (includes farm work and gardening): yes or no

Classification: total physical activity (light and moderate)

1. Forearm BMD

p-value association between baseline physical activity levels BMD changes

1. Forearm BMD

A. Light: p=0.5122 or

B. Moderate: p=0.0711

[Quantitative estimates not reported]

aRodriguez-Gomez 2019

(Toledo Study for Healthy Aging)

Overall risk of bias: High [86]

Cohort

(4 years follow-up)

227/192

Setting: Community; United States

Health status: Healthy

n= 192

Age: 80.5 (4.3)

Female: 52.6%

Physical activity was assessed by accelerometry during waking hours for seven consecutive days, except while bathing or swimming activities

A. Sedentary behaviours - mean % of waking hours: 55 (baseline); 59 (follow-up)

B. Light physical activity - mean % of waking hours: 41.9 (baseline); 38.8 (follow-up)

C. Moderate to vigorous physical activity – mean % of waking hours: 3.1 (baseline); 2.2 (follow-up)

Classification: total physical activity

1. Whole body BMC

2. Pelvic BMC

3. Arms (mean) BMC

4. Legs (mean) BMC

5. Lumbar spine (L1-4) BMC

6. Femoral regions BMC (proximal femur – mean, femoral neck, trochanter, ward’s triangle)

7. Whole body BMD

8. Pelvic BMD

9. Arms (mean) BMD10. Legs (mean) BMD

11. Lumbar spine (L1-4) BMD

12. Femoral regions BMD (proximal femur – mean, femoral neck, trochanter, ward’s triangle)

NB: Only significant findings were reported here]

Multiple regression coefficient (y) of change in the composition of movement behaviours and changes in BMC or BMD:

4. Legs (mean) BMC

A. NS

B. NS

C. y=1.767, p= 0.04¥

5. Lumbar spine (L1-4) BMC

A. NS

B. NS

C. y=0.050, p= 0.03¥

10. Legs (mean) BMD

A. NS

B. NS

C. y=0.005, p= 0.04¥

aShephard 2017

(Nakanojo Study)

Overall risk of bias: Low [87]

Cohort

(5 years follow-up)

615/496

Setting: Community; Japan

Health status: Healthy

Men

Age: 71.2 (3.9)

n= 212

Women

Age: 71.3 (4.2)

n= 284

Female: 57.3%

Physical activity was measured using pedometer for 5 years and analysed as daily step count and the daily duration of exercise at an intensity >3 METs

A. Physical activity level Quartile 1

Step count (steps/day) at baseline: 3888 (1117) [men]; 3824 (1298) [women]

Duration of activity > 3 METs (min/day): 4.0 (1.8) [men]; 4.0 (2.4) [women]

B. Physical activity level Quartile 2

Step count (steps/day) at baseline: 5994 (943) [men]; 5931 (924) [women]

Duration of activity > 3 METs (min/day): 10.9 (2.4) [men]; 10.1 (2.3) [women]

C. Physical activity level Quartile 3

Step count (steps/day) at baseline: 7521 (833) [men]; 7626 (691) [women]

Duration of activity > 3 METs (min/day): 19.3 (2.9) [men]; 18.4 (2.6) [women]

D. Physical activity level Quartile 4

Step count (steps/day) at baseline: 10892 (1433) [men]; 10199 (1398) [women]

Duration of activity > 3 METs (min/day): 31.8 (5.6) [men]; 30.3 (4.3) [women]

Classification: total physical activity

1. Osteosonic Index (OSI) from the ultrasonic measurement of the calcaneus (Fracture threshold – yes/no)

Multivariate Cox proportional hazard ratio (risk of the OSI falling below the fracture threshold) and 95% CI

Step count (steps/day)

Men

Q1. 2.63 (1.35 to 4.41)¥

Q2. 1.75 (1.03 to 3.95)¥

Q3. 1.01 (0.55 to 3.37)

Q4. 1

Women

Q1. 3.33 (2.10 to 5.21)¥

Q2. 2.51 (1.25 to 4.03)¥

Q3. 1.12 (0.47 to 2.16)

Q4. 1

Duration of activity >3 METs (min/day)

Men

Q1. 2.77 (1.46 to 5.59)¥

Q2. 1.91 (1.02 to 3.99)¥

Q3. 1.00 (0.48 to 2.27)

Q4. 1

Women

Q1. 3.94 (2.35 to 6.73)¥

Q2. 1.87 (1.00 to 3.60)¥

Q3. 0.99 (0.40 to 2.06)

Q4. 1

aSvejme 2014

Overall risk of bias: Low [88]

Cohort

(25 years follow-up)

Setting: Community; Sweden

Health status: Healthy

A. Active women

n= 91

B. Inactive women

n= 21

Age: women recruited at age 48 and followed up for 25 years

Female: 100%

Physical activity measured using questionnaires at four defined time periods: at menopause, 5 and 10 years after menopause, and at age 72.

A. Active women (>30 min/day) – mean (95% CI) number of hours of physical activity per week:

Baseline: 9.0 (7.8 to 10.2)

Average post-menopausal physical activity: 8.7 (7.6 to 9.8)

Physical activity at age 72: 8.2 (6.9 to 9.4)

B. Inactive women (<30 min/day) – mean (95% CI) number of hours of physical activity per week:

Baseline: 3.0 (1.6 to 4.4)

Average post-menopausal physical activity: 2.0 (1.4 to 2.5)

Physical activity at age 72: 1.2 (0.5 to 1.9)

Classification: total physical activity

1. Forearm BMC (mg/cm)

2. Forearm bone mineral apparent density (mg/cm3)

Mean (95% CI) average annual changes

1. Forearm bone mineral content (mg/cm):

A. -1.2 (-1.3 to -1.1)

B. -1.6 (-1.9 to -1.3)

Mean differences

0.4 (0.1 to 0.6)¥

2. Forearm bone mineral apparent density

A. -1.8 (-1.9 to -1.7)

B. -2.0 (-2.2 to -1.7)

aBleicher 2013

(CHAMP study)

Overall risk of bias: Low [80]

Longitudinal cohort study

(2 years follow-up)

1,705/1,122

Setting: Community; Australia

Health status: Healthy

n= 1,122

Age: 76.2 (5.1); range 70-97

Female: 0%

A. Walking for daily exercise was self-reported and measured in kilometres per day

A. Walk daily > 0 to ≤ 1km

B. Walk daily > 1 to ≤ 2 km

C. Walk daily > 2 to ≤ 4 km

D. Walk daily > 4 km

Classification: planned physical activity (exercise)

B. Physical Activity Scale for the Elderly (units)

Classification: total physical activity

1. Total hip BMD

2. Total hip BMC

Multiple regression coefficient (95% CI)

Reference: no walking

Leisure-time walking

1. Total hip BMD

A. 0.09 (-0.18 to 0.36), p=0.5

B. 0.18 (-0.06 to 0.41), p=0.1

C. 0.29 (0.06 to 0.52), p=0.01¥

D. 0.19 (-0.1 to 0.49), p=0.2

2. Hip BMC: NR

Age adjusted annualised percentage change in total hip BMD per unit change

General physical activity

1. Total hip BMD

-0.01 (-0.09 to 0.07)

2. Hip BMC: NR

Greendale 1995

(Rancho Bernardo study)

Overall risk of bias: Low [49]

Retrospective study

1,703

Setting: Community; United States

Health status: Healthy

n= 1,703

Age: 73

Female: 60%

Lifetime leisure physical activity, calculated based on leisure time physical activity (collected retrospectively via questionnaire) for the past year, age 30 years and age 50 years

Exercise level: classified by the highest level of exercise performed for at least 15 minutes per session at least three times per week.

Participants were divided into levels of physical activity according to the tertiles

A. Low

B. Medium

C. High

Classification: planned physical activity (exercise)

1. Total hip BMD

2. Intertrochanter BMD

3. Femoral neck BMD

4. Greater trochanter BMD

5. Lumbar spine (L1-4) BMD

6. Distal radius BMD

7. Midshaft radius BMD

Adjusted mean (p-value for comparison A vs C)

1. Total hip BMD (p=0.002)¥

A. Low: 0.8241

B. Medium: 0.8367

C. High: 0.8507

2. Intertrochanter BMD (p=0.007)¥

A. Low: 0.9631

B. Medium: 0.9769

C. High: 0.9908

3. Femoral neck BMD (p=0.003)¥

A. Low: 0.6597

B. Medium: 0 6716

C. High: 0.6819

4. Greater trochanter BMD (p = 0.0001)¥

A. Low: 0.5969

B. Medium: 0.6093

C. High: 0.6248

5. Lumbar spine (L1-4) BMD

A. Low: 0.9324

B. Medium: 0.9612

C. High: 0.9479

6. Distal radius BMD: NR

7. Midshaft radius BMD: NR

aGudmundsdottir 2010

Overall risk of bias: High [82]

Cohort

(4 years follow-up)

162

Setting: Community; Iceland

Health status: Healthy

A. Physical activity performed ≤ 3 per week

n= 41 (analysed)

B. Physical activity performed > 3 per week

n= 111 (analysed)

Age: 75

Female: 100%

Physical activity was calculated based on number of leisure time walks per week and number of other exercise session per week (self-reported questionnaire)

Results were presented according to number of times of physical activity performed per week

A. ≤ 3 per week

B. > 3 per week

Classification: planned physical activity (exercise)

1. Femoral neck BMD

2. Total trochanter BMD

3. Total hip BMD

Change score; mean % change (SE)

1. Femoral neck

A. -1.3 (1.1)

B. -0.2 (0.9)

β: NR

2. Total trochanter BMD

A. -1.5 (0.8)

B. -1.2 (0.7)

β= 0.22, non-significant p value

3.Total hip BMD

Mean (SE)

A. -1.4 (0.8)

B. -1.1 (0.7)

β= 0.19, non-significant p value

Huddleston 1980

Overall risk of bias: High [51]

Retrospective observational study

35/35

Setting: Community; United States

Health status: Healthy tennis athletes

n= 35

Age: range 70-79

Female: 0%

Lifetime tennis exposure in athletes with tennis experience ranging from 25 to 72 years

Results were presented for:

A. Playing arm

B. Non-playing arm

C. Comparison with data for a “normal male population”

Classification: planned physical activity (sport – tennis)

1. Radius midshaft BMC

1. Radius BMC

4% to 33% greater for the playing arms as compared with the nonplaying arms

The mean difference between the playing arm and nonplaying arm: 13%

Reference data suggest difference between dominant and nondominant

BMC values of 6% to 9%

aKemmler (2016) Bone

(Erlangen Fitness and Osteoporosis

Prevention Study)

Overall risk of bias: High [83, 93]

Retrospective secondary analysis of the intervention group of a quasi-randomised trial (16 years follow-up)

Setting: Community; Germany

Health status: Osteopenia

A. Exercise group:

n= 55

Age: 55.1 (3.4)

Female: 100%

Exercise group: Supervised group class (aerobic dance exercise, jumping and resistance exercise) + Home training (rope skipping, isometric and dynamic resistance exercise and stretching/ flexibility exercise) five months after study started; 49 to 50 weeks/year throughout the 16 years

Exercise frequency (ExFreq): session/week/16 years

Classification: planned physical activity (exercise)

1. Lumbar spine BMD

2. Total hip BMD

Linear mixed-effect regression analysis. Marginal effect (95% CI)

1. Lumbar spine BMD: 0.035 (0.024 to 0.045)

2. Total hip BMD: 0.015 (0.005 to 0.026)

Minimum effective dose of exercise (training sessions/week)

1. Lumbar spine: 2.11 (2.06 to 2.12)

2. Total hip BMD: 2.22 (2.00 to 2.78)

Rikkonen 2010

(Kuopio Osteoporosis Risk Factor and Prevention study)

Overall risk of bias: Low [66]

Cohort (15 years follow-up)

8560

Setting: Community; Finland

Health status: Healthy

n= 8560 (analysed)

A. Physical activity quartile I

Age: 52.1 (2.9)

B. Physical activity quartile II

Age: 52.0 (2.9)

C. Physical activity quartile III

Age: 52.2 (2.9)

D. Physical activity quartile IV

Age: 52.3 (2.8)

Female: 100%

Leisure-time physical activity (self-reported) collected at 5 years intervals

A. 15-year average PA, hours/week: 0.35 (0.35)

B. 15-year average PA, hours/week: 1.7 (0.39)

C. 15-year average PA, hours/week: 3.2 (0.54)

D. 15-year average PA, hours/week: 7.0 (2.9)

Classification: leisure-time physical activity (exercise, transportation, sport)

1. Femoral neck BMD

2. Trochanter BMD

3. Ward’s triangle BMD

4. Lumbar spine (L2-4) BMD

Beta ± SE (quartile IV vs inactive)

1. Femoral neck BMD

1.752 ± 0.493

2. Trochanter BMD

1.783 ± 0.581

3. Ward’s triangle BMD

2.412 ± 0.723

4. Lumbar spine (L2-4) BMD

0.040 ± 0.649

All results were significant (except for lumbar spine) and suggest a positive effect of physical activity on BMD.

  1. BMC bone mineral content (g unless specified), BMD bone mineral density (g/cm2). Where studies reported effect estimates with differing degrees of adjustment for confounders in different models, we used the estimate from the most adjusted model
  2. aindicate studies that were found in the updated search
  3. ¥indicates statistically significant between-group difference at p < 0.05.