Osteoporosis International

, Volume 24, Issue 7, pp 1951–1963

Predictors of the rate of BMD loss in older men: findings from the CHAMP study


    • School of Public HealthUniversity of Sydney
  • R. G. Cumming
    • School of Public HealthUniversity of Sydney
  • V. Naganathan
    • Centre for Education and Research on AgeingUniversity of Sydney
  • M. J. Seibel
    • ANZAC Research InstituteUniversity of Sydney
  • F. M. Blyth
    • Centre for Education and Research on AgeingUniversity of Sydney
  • D. G. Le Couteur
    • Centre for Education and Research on AgeingUniversity of Sydney
  • D. J. Handelsman
    • ANZAC Research InstituteUniversity of Sydney
  • H. M. Creasey
    • Centre for Education and Research on AgeingUniversity of Sydney
  • L. M. Waite
    • Centre for Education and Research on AgeingUniversity of Sydney
Original Article

DOI: 10.1007/s00198-012-2226-0

Cite this article as:
Bleicher, K., Cumming, R.G., Naganathan, V. et al. Osteoporos Int (2013) 24: 1951. doi:10.1007/s00198-012-2226-0



Though bone loss tends to accelerate with age there are modifiable factors that may influence the rate of bone loss even in very old men.


The aim of this 2-year longitudinal study was to examine potential predictors of change in total hip bone mineral density (BMD) in older men.


The Concord Health and Ageing in Men Project is a population-based study in Sydney, Australia. For this study, 1,122 men aged 70–97 years had baseline and follow-up measures of total hip BMD measured with dual X-ray absorptiometry. Data about mobility, muscle strength, balance, medication use, cognition, medical history and lifestyle factors were collected using questionnaires and clinical assessments. Serum 25-hydroxyvitamin D [25(OH)D] was also measured. Multivariate linear regression models were used to assess relationships between baseline predictors and change in BMD.


Over a mean of 2.2 years, there was a mean annualised loss of total hip BMD of 0.006 g/cm2/year (0.6 %) and hip BMC of 0.14 g/year (0.3 %). Annual BMD loss accelerated with increasing age, from 0.4 % in men aged between 70 and 75 years, to 1.2 % in men aged 85+ years. In multivariate regression models, predictors of faster BMD loss were anti-androgen, thiazolidinedione and loop-diuretic medications, kidney disease, poor dynamic balance, larger hip bone area, older age and lower serum 25(OH)D. Factors associated with attenuated bone loss were walking for exercise and use of beta-blocker medications. Change in BMD was not associated with baseline BMD, smoking, alcohol consumption, BMI, frailty, or osteoarthritis.


There was considerable variation in the rate of hip bone loss in older men. Walking, better balance and beta blockers may attenuate the acceleration of BMD loss that occurs with age.


BMC lossBone mineral densityEpidemiologyLongitudinal studyMenRisk factors


With worldwide increases in life expectancy, more men will develop osteoporosis and suffer the consequences of osteoporotic fractures [1]. Hip bone mineral density (BMD) is a very strong predictor of fractures in men [25]. By the age of 30, hip BMD in men begins to decrease at a linear rate [6], and in older age, the rate of BMD loss has been reported to accelerate in some [712] but not other studies [1315]. To date, the mechanisms for bone loss remain unclear.

Considerable variability in the rate of bone loss in older men has been reported [11]. Understanding the source of variation and, in particular, identifying modifiable risk factors is important. A number of studies have sought to determine the predictors of BMD loss in older men [8, 9, 1121], but many of these are limited by sample sizes with fewer than 350 older men [10, 14, 15, 18, 19, 21]. Most observational studies have measured only self-reported lifestyle characteristics and anthropometric variables [9, 12, 13, 15] or have investigated just a single predictor variable. Only a few studies have also included objective neuromuscular measures such as muscle strength [1618]. To date, there have been inconsistent reports on the influence of baseline BMD, serum 25(OH)D, lifestyle factors such as smoking, alcohol and physical activity, and anthropomorphic factors such as height, weight and BMI.

The aims of this study were to describe bone loss in a representative group of older Australian men and to examine associations between hip BMD change and an extensive range of neuromuscular functions such as gait, balance and strength, in addition to demographic, anthropomorphic, medical and lifestyle factors. We were particularly interested in identifying the characteristics of men who were losing hip BMD rapidly because hip BMD is the best predictor of hip fracture, and rapid loss of BMD can quickly change fracture risk status.



The Concord Health and Ageing in Men Project (CHAMP) is a longitudinal study of health and ageing in men. During a 29-month period, between 2005 and 2007, all community-dwelling men aged 70 years and over living in three local government areas surrounding Concord Hospital, Sydney, Australia, were invited to participate. Invitation letters were sent to 3,627 men, and contact was made with 3,005. One hundred ninety-four eligible men volunteered for the study prior to receiving their recruitment letter. The state electoral roll was used as the sampling frame. One hundred ninety of the contacted men were not eligible for the study because they were no longer living in the local community. Of the 2,815 eligible men with whom contact was made, 1,511 participated in the study (54 %).

Of the 1,705 men in the study at baseline, 1,366 men attended the follow-up clinic. Of the 339 men who did not attend, 99 were deceased, 109 were too ill and 50 had moved into residential care or out of the area and 81 did not attend for a variety of reasons.

DXA measurements of hip BMD, BMC and bone area

Hip scans were made with dual X-ray absorptiometry (DXA) using the fan beam Discovery-W scanner (Hologic Inc., Bedford, MA, USA). Men removed jewelry and wore light cotton gowns free from metal. All measures of the hip bone mineral content (BMC) (g), hip area and areal BMD (g/cm2) refer to the total hip. Hip area refers to the cross-sectional area of the total hip region utilised by the DXA scanner in the calculation of BMD. The same DXA scanner, operated by the same staff, was used for all scans. The coefficient of variation (CV %) for scans duplicated on 30 men from the study cohort were 1.6 % for the total hip. Quality control scans were conducted daily using the Hologic whole-body phantom and indicated no shifts or drifts.

Potential predictors of BMD change

All information on potential predictors was obtained from the baseline assessments. Potential predictors were identified from published studies of BMD in men and women, including our previous study of predictors of baseline BMD in CHAMP men [22]. Information on demographics, lifestyle and medical and family history was obtained from a questionnaire completed prior to attending the clinic. Birth place was categorised as Australia, Mediterranean (Italy, Greece, or Spain), Asia and Other. Participants from Asia were predominantly from China and Southeast Asia. Medication use was ascertained in the clinic, and participants were asked to bring all prescription and non-prescription medications for verification. The Iowa Drug Information Service drug code numbers were used to code and classify the medication data. Smoking status was classified as never, past (ex-smoker), or current. Alcohol consumption was reported in drinks per week. Weekly consumption of alcohol was calculated from self-report of the average frequency of drinking per week × average number of drinks containing 10 g of alcohol consumed per day. Alcohol use was further categorised and drinkers were compared to non-drinker groups. Cut points were determined by the Australian guidelines in use at the time of the study. These guidelines define <28 drinks per week as low-risk drinking and 29+ drinks per week as risky drinking. An additional category of 1–14 drinks was created [23].

The questionnaire asked men about their use of walking aids, perception of their health and functional limitations imposed by their health, number of hours spent sitting each day and the number of weeks spent in hospital or resting in a chair due to illness during the preceding 12 months. Walking for daily exercise was self-reported and measured in kilometers/day. General physical activity levels were determined using the Physical Activity Scale for the Elderly questionnaire [24]. Frailty was determined using the Fried Frailty criteria [25].

A fracture history was obtained at the baseline clinic interview. Three observers (KB, VN and RC) independently rated fractures using the description of the mechanism of injury. Fractures were classified as “low trauma fractures” if the mechanism of injury was described as a fall from standing height or less; a fall on steps, or kerbs (but not down steps); or a minimal trauma incident other than a fall, for example turning over in bed or coughing. There was a high level of agreement between observers (all kappas > 0.84). Maternal and paternal fractures were categorised as yes/no and unknown.

Grip strength (kg) of both hands was measured twice using a hand-held dynamometer (Jamar). The average strength was used in the analyses. Isometric quadriceps’ strength (leg strength) (kg) was measured three times on each leg using a 100-kg spring gauge with the knee at 90° flexion while sitting on a high stool. The best result was used. Leg strength was categorised into quartiles for multivariate analyses. The chair-stand test (yes/no) assessed the ability to stand five times from sitting, without using arms. Usual gait speed was determined over a 6-m path. The narrow walk test was determined by the ability to walk 6 m within a 20-cm wide path. Postural sway (cm2), an indicator of standing balance, was measured with a sway meter [26] standing on a firm floor. A higher score indicates poorer balance.

Weight was measured in indoor clothing without shoes on a digital scale (Tanita BWB-600). Height was measured using a Harpenden stadiometer (Dyfed, UK). Body mass index (BMI) was calculated using weight (kg)/height (m2). BMI was examined as a categorical variable using the cut points <21, 22 to 25, 26 to 29 and 30+ (kg/m2) as well as a continuous variable. Height and weight change since age 25 years were calculated from self-reported information minus current measured weight. Height and weight losses since age 25 were further coded as “height loss of at least 50 mm” and “weight loss greater than 10 kg.”

All participants were screened for cognitive impairment using the Mini-Mental State Examination (MMSE) and the Informant Questionnaire on Cognitive Decline [27]. Those who scored 26 or below on the MMSE or more than 3.6 on the IQCODE underwent further clinical assessment by a geriatrician.

Serum 25-hydroxyvitamin D [25(OH) D] was measured at baseline using a Diasorin radioimmunoassay kit (Stillwater, MN, USA). Sensitivity was measured at 4 nmol/l; intraassay precision was 7.6 % and interassay precision was 9.0 %, with a laboratory reference range of 39–140 nmol/l. Serum 25(OH)D did not have a linear relationship with BMD change, and thus, serum 25(OH)D was categorised into quintiles. The cut points in nanomoles per liter for serum 25(OH)D were as follows: quintile 1, ≤36; quintile 2, 37 to ≤48; quintile 3, 48 to ≤59; quintile 4, 59 to ≤72; and quintile 5, ≥73.

Statistical analysis

All statistical analyses were conducted using SAS Version 9.1 (SAS Institute Inc., Cary, NC, USA.). Student’s t tests and Mantel Haenszel chi-squared (χ2) tests were used to evaluate differences between men who attended follow-up and those who did not. Annualised changes of BMD, BMC and hip bone area were calculated as the follow-up measure minus the baseline measure divided by the time between baseline and follow-up. Annualised percentage changes were calculated for each participant as (annual change/baseline measure) × 100. A negative numeral reflects a loss. For descriptive purposes, and to provide a comparison to previous studies [11, 28], BMD change was categorised according to whether participants had “accelerated” loss of BMD, “expected” loss of BMD, or “maintenance” of BMD. “Accelerated loss” describes BMD loss of at least one standard deviation (1 SD) below the annualised mean BMD loss (1 SD = 0.01303 g cm−2 year−1, the mean loss = −0.00564 g/cm2; thus, the cut point was −0.01867 g/cm2). “Expected loss” describes BMD loss between no change and 1 SD below the mean change. “Maintained BMD” describes BMD change of ≥0 (no change or increased BMD).

The data were initially analysed using unadjusted linear regression models and then age-adjusted models. Variables were selected for entry into a stepwise multivariate regression analysis if they (a) had a p ≤ 0.25 in age-adjusted models or (b) had been found to be associated with change in BMD in previous studies. For example, smoking, alcohol consumption, and weight were further analysed even though they were not associated with BMD change in age-adjusted analyses. The final model was produced by backwards step-wise elimination until all variables were significant at p ≤ 0.05. Interaction terms between final variables were assessed within the multivariate model. Collinearity was assessed using the variance inflation factor. Interaction effects between potential predictors and age were examined by including interaction terms in multivariate models.

The relationship between baseline BMD and follow-up BMD was assessed in a model that included all the final predictor variables of BMD change in order to determine the importance of baseline BMD to BMD 2 years later.


Of the 1,366 men who attended the 2-year clinic, 40 men did not have both baseline and follow-up scans, 9 men had a different hip scanned because of a new hip replacement, 183 men were taking bone medications and 12 men reported having Paget’s disease. Men who were not included in this follow-up analysis tended to be older, frailer, less active and have lower BMD (Table 1).
Table 1

Comparison of baseline characteristics [mean ± SD or prevalence (%)] of men included in the follow-up BMD analysis compared to those who were not included


Men not in this analysis (n = 583)

Men in this follow up analysis (n = 1,122)

p valuea

Mean ± SD or prevalence (%)

Mean ± SD or prevalence (%)

Age (year ± SD)

78.4 ± 6

76.2 ± 5.1


BMI (kg/m2 ± SD)

27.5 ± 4.4

28.0 ± 3.8


Weight (kg)




Leg strength (kg ± SD)

28.9 ± 7.4

31.7 ± 8.1



103.8 ± 60.4

124.7 ± 55.7


Total hip BMD (g/cm2 ± SD)

0.9103 ± 0.15

0.9445 ± 0.13


Femoral neck BMD (g/cm2 ± SD)

0.7379 ± 0.14

0.7671 ± 0.12


Lean mass (kg)

47.4 ± 6.9

49.3 ± 6.1


Fat mass (kg)

21 ± 7.8

21.5 ± 6.9


Alcohol consumption


173 (29.7 %)

242 (21.6 %)


1–14 drinks/week

324 (55.6 %)

675 (60.2 %)


15–28 drinks/week

54 (9.3 %)

132 (11.8 %)


29+ drinks/week

21 (3.6 %)

40 (3.6 %)




210 (36.0 %)

419 (37.3 %)



328 (56.3 %)

627 (56.9 %)



39 (6.7 %)

64 (5.8 %)


Poor health

43 (7.4 %)

37 (3.3 %)


Serum 25(OH)Db

25(OH)D quintile 1

136 (23.3 %)

187 (16.7 %)


25(OH)D quintile 2

141 (24.2 %)

214 (19.1 %)


25(OH)D quintile 3

96 (16.5 %)

228 (20.3 %)


25(OH)D quintile 4

85 (14.6 %)

234 (20.9 %)


25(OH)D quintile 5

102 (17.5 %)

239 (21.3 %)



217 (37.2 %)

411 (36.6 %)



101 (17.3 %)

57 (5.1 %)


Loop diuretics

62 ( 10.6 % )

60 (5.3 %)



13(2.2 %)

13(0.8 %)


Kidney disease

32 (5.5 %)

26 (2.3 %)



55 (9.4 %)

38 (3.4 %)


PASE Physical Activity Scale for the Elderly

ap derived from χ2 tests for categorical variables and t tests for continuous variables

bSerum 25(OH)D(nmol/l): quintile(Quin) using cut points of original CHAMP cohort: Quin 1, ≤36; Quin 2, >37 to ≤ 48; Quin 3, >48 to ≤ 59; Quin 4, >59 to ≤72; Quin 5, 73

The mean baseline age of the 1,122 men included in this study was 76.2 (SD, 5.1; median, 75.0) years with a range of 70–97 years. From a mean baseline, hip BMD of 0.957 g/cm2 (SD, 0.12), there was a mean annualised percentage loss of hip BMD of 0.6 %/year (0.0056 g cm−2 year−1) and of hip BMC of 0.1 %/year (0.117 g/year). Hip area increased by 0.3 %/year (0.14 cm2/year).

Figure 1 illustrates the range of annualised percentage change of hip BMD. A substantial proportion of men (34 %) lost more than 1 % of their baseline BMD per year with 14 % of men losing more than 2 % of baseline BMD/year. One hundred fifty-nine men (14.2 %) had accelerated BMD loss (loss of at least 1 SD), 576 (51.3 %) had expected BMD loss and 387 (34.5 %) maintained BMD. Men in the accelerated BMD loss category lost a mean of 3 % of baseline BMD. Men with accelerated BMD loss were equally spread across the quartiles of baseline BMD (χ2, p = 0.4).
Fig. 1

Distribution of annual percentage change in BMD at the total hip (%/year)

Figure 2 depicts the change in hip BMD, BMC and cross-sectional area. The mean rate of hip BMD and hip BMC loss accelerated with increasing age (p < 000.1; Fig. 2). In contrast, the rate of increase in mean hip bone area was constant (0.3 cm2; SD, 1.3). The acceleration of BMD and BMC loss appeared to plateau after age 90; however, there were only 16 men aged 90 years and over, and there were large confidence intervals for all measures of change in this age group (Fig. 2).
Fig. 2

Annual percentage change in total hip BMD, BMC and cross-sectional area by age group. Vertical bars denote 95 % CIs

Age-adjusted results

Table 2 depicts the age-adjusted association between potential predictors and annualised percentage BMD change. Unadjusted associations (data not shown) were similar to age-adjusted associations. Age-adjusted predictors (p ≤ 0.1) of greater BMD loss, in decreasing order of effect size, were the following: anti-androgen treatment, thiazolidinedione medication use, inability to walk a narrow 6 m path, kidney disease, self-reported poor health, loop diuretic and thyroxine medication use, low serum 25(OH)D, weak quadriceps, older age, poor dynamic balance, higher baseline hip area and height loss since age 25. Age-adjusted predictors associated with reduced rate of BMD loss were androgen and beta-blocker medication use and walking daily for exercise.
Table 2

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


Mean (SD) or prevalence (%)


Age adjusted beta/unit/year (95 % CI)

p value


Age unadjusted

76.2 (5.1)


−0.24 (−0.32,−0.16)


Asian birth

4.70 %


0.33 (−0.06,0.71)



80.2 (12.3)


0.001 (−0.07,0.07)


BMI (kg/cm2)

28.0 (11.8)


−0.02 (−0.1,0.07)


Height loss since age 25 (cm)

3.5 (4.3)


0.01 (0,0.02)


Weight change since age 25

10.3 (11.9)


0.03 (−0.01,0.06)


Baseline total hip DXA

BMD (g/cm2)

0.957 (0.13)


0.06 (−0.02,0.14)


BMC (g)

41.5 (7)


−0.02 (−0.1,0.06)


Bone cross-sectional area (cm2)

43.3 (4)


−0.11 (−0.19,-0.03)



Alcohol: 1–14 drinks/week

62.10 %

no alcohol

0.17 (−0.03,0.37)


Alcohol: 14–28 drinks/week

12.10 %


0.1 (−0.19,0.39)


Alcohol: 28+ drinks/week

3.80 %


−0.05 (−0.51,0.41)


Smoker current

5.80 %


−0.04 (−0.54,0.46)


Smoker past

56.50 %


−0.16 (−0.64,0.33)


Fracture history

Father hip fracture

1.30 %


−0.48 (−1.19,0.22)


Mother hip fracture

7.80 %


−0.05 (−0.35,0.25)


Low trauma fracture since age 50

8.30 %


−0.21 (−0.51,0.08)




40.30 %


−0.07 (−0.25,0.12)



5.60 %


−0.05 (−0.43,0.33)


Self report of poor/very poor health

3.30 %


−0.52 (−0.97,-0.07)


Medical history


10.20 %


−0.01 (−0.28,0.26)



17.10 %


0.09 (−0.13,0.3)



0.60 %


0.18 (−0.85,1.21)



13.40 %


−0.02 (−0.26,0.22)



46.70 %


0.11 (−0.05,0.27)


Kidney disease

2.30 %


−0.53 (−1.07,0.01)



2.50 %


−0.23 (−0.76,0.3)


Kidney stones

11.90 %


0.08 (−0.18,0.33)


Liver disease

1.60 %


−0.2 (−0.85,0.44)



18.20 %


−0.13 (−0.34,0.08)


Parkinson’s disease

1.50 %


−0.08 (−0.74,0.58)


Rheumatoid arthritis

9.80 %


0.19 (−0.09,0.46)



7.60 %


−0.12 (−0.43,0.19)




1.2 %


0.95 (−0.16,2.05)



0.8 %


−1.25 (−2.15,−0.34)


Beta blockers

15.0 %


0.27 (0.05,0.50)


Corticosteroid tablets

3.8 %


0.28 (−0.15,0.71)



0.6 %





5.3 %




Diuretic thiazide

7.8 %


−0.17 (−0.47,0.13)



2.6 %


−0.28 (−0.79,0.23)



2.7 %


0.39 (−0.12,0.89)



44.6 %


0 (−0.17,0.16)



2.7 %


−0.43 (−0.93,0.07)




3.9 %


−0.18 (−0.63,0.28)


Neuromuscular function and activity levels

Poor dynamic balancec

42.1 %


−0.22 (−0.39,-0.05)


Unable to do 5 chair standsd

4.4 %


−0.52 (−0.92,−0.12)


Unable to pass narrow walk teste

1.0 %


−0.97 (−1.79,-0.16)


Quadriceps leg strengthf

Leg strength (kg) (<25 kg)

14.6 %


−0.25 (−0.49,0.01)


Leg strength(kg) (25 to < 30 kg)

16.1 %


−0.07 (−0.31,0.16)


Leg strength (kg)(30 to <35 kg)

21.6 %


−0.02 (−0.23,0.19)


Walk daily for exercise

Walk daily > 0 to ≤1 km

13.5 %


0.1 (−0.21,0.31)


Walk daily >1 to ≤2 km

13.8 %

no walking

0.21 (−0.02,0.44)


Walk daily >2 to ≤4 km

19.9 %


0.27 (0.04,0.50)


Walk daily > 4 km

9.9 %


0.22 (−0.08,0.5)


PASE (units)



−0.01 (−0.09,0.07)


Serum 25(OH)D (continuous nmol/l)



0.002 (−0.0001,0.003)


Serum 25(OH)D quintilesg

25(OH)D quintile 1

21.0 %

Quintile 5

−0.37 (−0.63,-0.10)


25(OH)D quintile 2

16.7 %


−0.05 (−0.30,0.20)


25(OH)D quintile 3

20.2 %


−0.22 (−0.47,0.03)


25(OH)D quintile 4

21.7 %


0.01 (−0.23,0.26)


PASE Physical Activity Scale for the Elderly

aUnit/referent: for continuous variables, unit is approximately 1 SD; for dichotomous variables, the referent group does not have the characteristic

bDementia vs other (ie normal and MCI)

cPoor dynamic balance = if race track test score was >median [18]

dFive chair stands-without using arms

eNarrow walk test: walk two laps without touching boundary (6 m long × 20 cm wide)

fLeg strength − static quadriceps force (kg) − groups represent quartiles

gSerum 25(OH)D(nmol/l):Quintile(Q) cut points: Q1, ≤36; Q2, >37 to ≤48; Q3, >48 to ≤59; Q4, >59 to ≤72; Q5, 73+. The F - test for difference between 25(OH)D groups: p = 0.02

Multivariate models

After excluding men who did not have complete data on all covariates, the fully adjusted model included 975 men. The most common reason for men not to be included in our final multivariate analyses was their inability to do the balance tests (7 %). In addition, 3.5 % of men did not provide information on walking for exercise, 1 % did not provide medical history and 1.5 % did not have serum 25(OH)D measures.

Table 3 shows the final multivariate model. In the order of effect size, anti-androgen treatment, thiazolidinedione and loop-diuretic medications use, kidney disease, poor dynamic balance, lower serum 25(OH)D, larger hip bone area and older age were associated with an increased rate of BMD loss. A reduced rate of BMD loss was associated with beta-blocker medication use and walking more than a kilometre per day for exercise. The protective association of walking appeared to be dose related with a larger effect associated with a greater distance walked. However, distances >4 km/day appeared to offer no additional benefit. Heavier body weight had a very small protective relationship with slower BMD loss. Although baseline BMD was not a predictor of the rate of bone loss, it was the most important predictor of follow-up BMD and explained 84 % of variability of follow-up BMD (data not shown).
Table 3

Multivariate predictors of annualised percentage change in BMD at the total hip per unit change, adjusted for multiple variables



Partial regression coefficients per unit change (95%CI)b

p value

Anti-androgen treatment


−1.39 (−2.32,−0.46)


Thiazolidinedione use


−1.07 (−2.07,−0.07)


Loop diuretic use


−0.56 (−0.97,−0.16)


Kidney disease


−0.98 (−1.57,−0.39)


Poor dynamic balance C


−0.29 (−0.46,−0.11)


DXA Hip area (cm2)


−0.18 (−0.06,−0.02)


Age (years)


−0.13 (−0.04,−0.01)


Beta-blocker use


0.35 (0.12,0.58)


Walk daily for exercise

Walk daily >0 to ≤1 km


0.09 (−0.18,0.36)


Walk daily >1 to ≤2 km


0.18 (−0.06,0.41)


Walk daily >2 to ≤4 km


0.29 (0.06,0.52)


Walk daily 4+ km


0.19 (−0.1,0.49)


Serum 25(OH)D quintilesd

25(OH)D quintile 1




25(OH)D quintile 2




25(OH)D quintile 3




25(OH)D quintile 4




Weight (kg)


0.08 (0.0,0.02)


CI 95 % confidence interval

aUnit ∼ standard deviation. If the referent is not listed then the referent group did not have the characteristic

bPoor dynamic balance = if race track test score was > median (16 mm2)

cSerum 25(OH)D Quintile(Q) uses cut points of original CHAMP cohort, i.e. Q1, ≤36; Q2, >37 to ≤48; Q3, >48 to ≤59; Q4, >59 to ≤72; Q5, 73+ (nmol/l); test for difference between groups p = 0.06

Table 4 details the prevalence or mean of the predictor variables in the final regression model according to whether BMD change was accelerated, expected or maintained.
Table 4

Prevalence (%) or mean ± SD of factors significantly associated with total hip BMD change in multivariable adjusted models: presented by category of rate of change


Accelerated BMD loss n = 159 (14.2 %)

Expected BMD loss n = 576 (51.3 %)

Maintained BMD n = 387 (34.5 %)

Total sample n = 1,122

Age(years )


± 0.4


± 0.2


± 0.2


Age groups

70–74 years


(32.1 %)


(43.6 %)


(49.4 %)


75–79 years


(32.1 %)


(31.1 %)


(32.3 %)


80–84 years


(20.8 %)


(19.1 %)


(13.2 %)


85+ years


(15.1 %)


(6.3 %)


(5.2 %)


Weight (kg)


± 13.7


± 11.6


± 12.7


Hip bone area (cm2)


± 4.9


± 4.2


± 4.3


Kidney disease


(3.8 %)


(2.6 %)


(1.3 %)


Anti-androgens use


(2.5 %)


(0.9 %)




Beta-blocker use


(12.6 %)


(13.0 %)


(19.4 %)


Thiazolidinedione use


(1.9 %)


(0.7 %)




Loop-diuretics use


(10.1 %)


(5.7 %)


(2.8 %)


History of low trauma fractureb


(11.3 %)


(7.8 %)


(7.5 %)


Poor dynamic balance


(47.8 %)


(39.1 %)


(35.4 %)


Walk for exercise groups (km/day)

No walking, n = 398


(42.1 %)


(35.9 %)


(32.0 %)


>0 to ≤1 km, n = 145


(17.0 %)


(12.8 %)


(11.4 %)


>1 to ≤ 2 km, n = 249


(19.5 %)


(17.5 %)


(30.2 %)


>2 to ≤ 4 km, n = 213


(10.7 %)


(20.3 %)


(20.4 %)


4+ km, n = 105


(7.5 %)


(8.5 %)


(11.4 %)


Low serum 25(OH)Db


(22.6 %)


(17.5 %)


(12.9 %)


Accelerated BMD loss: loss of at least one standard deviation below the annualised mean BMD change. There was a mean loss of 0.00564 g cm−2 year−1 thus accelerated BMC loss = loss of ≥ 0.01867 g cm−2 year−1. Expected BMD loss: between no change and a loss of 0.01867 g cm−2 year−1 (1SD below the mean change). Maintained BMD: no change or increased BMD

aLow trauma fracture after age 50 years

bLow serum 25(OH)D ≤ 36 nmol/l

Factors not independently associated with the rate of change in hip BMD in the multivariate model included baseline BMD, BMC, country of birth, BMI, general physical activity, sports participation, physical and functional limitations, alcohol consumption and smoking (whether categorised as current, past or assessed by dosage). There were no interaction effects.


This study presents data on longitudinal changes in hip BMD in older men and demonstrates an accelerating loss of hip BMD at the oldest ages. Although in most men the magnitude of BMD loss was modest, we found that a substantial proportion (34 %) of men lost between 1 and 7 % of baseline BMD/year. This annual loss may contribute significantly to increased fracture risk, particularly for men who already have low BMD. Factors associated with accelerated bone loss included use of anti-androgen, thiazolidinedione and loop diuretic medications, kidney disease, poor balance and low serum 25(OH)D. Walking as an exercise and the use of beta blockers were the only protective factors.


Anti-androgen therapy had the strongest association with BMD loss. BMD loss associated with anti-androgen therapy has been reported to range from 1.9 % [29] to 4.6 % [30] in the first year of anti-androgen therapy. Furthermore, men on anti-androgen therapy are at a heightened risk of fractures [31, 32]. This increased fracture risk is neither restricted to the elderly nor is the relationship diminished by the excess mortality in patients with advanced prostate cancer [33, 34]. As survival of patients with prostate cancer increases, the effect on skeletal morbidity of anti-androgens needs to be considered by prescribing doctors.

This is the first study to show that men using thiazolidinediones had significantly greater BMD loss after accounting for confounders. The rate of bone loss was not associated with other types of diabetic medication. Thiazolidinedione treatment is known to affect bone health in women, increasing fracture risk [35], bone turnover [36] and BMD loss [37]. However, the influence of thiazolidinedione use on bone in men has been conflicting [37, 38]. Schwartz et al. reported significantly greater BMD loss in women but not in men. In a case control study that did not adjust for confounders, Yatura et al. found greater hip and spine BMD loss in diabetic men using thiazolidinediones. There is also evidence of a greater fracture risk [3941] in men taking thiazolidinediones. Our findings of an additional 0.9 % BMD loss in men using thiazolidinediones is similar to the additional hip BMD loss (1 %) reported by Matura et al. [38].

To our knowledge, this is the first longitudinal study to have analysed the association between beta blockers and BMD change in men. Men on beta blockers were more likely to maintain BMD and overall had less BMD loss. The strength of the association was not weakened by adjustment for a large range of covariates, including the use of statins, nitrates and thiazide diuretics. Beta blockers have been associated with higher cross-sectional BMD in age and weight adjusted models [22, 42, 43], though in these studies the cross-sectional association was lost after models were adjusted for related medications and lifestyle factors. The underlying mechanism whereby beta blockers may have a beneficial effect on bone health is not clear. However, results from animal studies suggested that the central nervous system might play a role in regulation of skeletal remodelling activity. Evidence of a protective effect against fractures is inconsistent. The Dubbo Osteoporosis Epidemiology Study reported a decreased risk of fractures (OR, 0.5) in men on selective beta blockers [43]. However, two meta-analyses have drawn opposing interpretations. Weins et al. [44] concluded that beta blocker use is associated with a significant decreased fracture risk, while Reid [45] cautioned that there is not yet adequate evidence to reach any conclusion.

Men taking loop diuretics lost almost three times as much BMD as men who did not take them. Loop diuretics increase calcium excretion, plasma parathyroid hormone and osteocalcin and may negatively affect the activity of bone forming cells [46]. Our results are similar to findings from MrOS prospective study, which reported a 2.5-fold increase in bone loss in men taking loop diuretics after adjusting for confounders [47].


Men who reported having been diagnosed with kidney disease had substantially greater BMD loss, independent of age, low serum 25(OH)D or physical activity levels. Increased bone loss [48, 49] and an increased risk of hip and vertebral fractures have been reported in some [50], but not all [49, 51] studies of renal impairment and osteoporosis. It is likely that men in the later stages of renal failure would not have participated in CHAMP, thus underestimating the rate of BMD loss. On the other hand, kidney disease was self-reported, and it is likely that men in the early stages may not have been diagnosed.


Our findings of accelerating BMD loss with age support other longitudinal reports of BMD loss in older men [711, 14, 16, 21, 37, 52], except in men aged over 90 years in whom BMD loss appeared to attenuate somewhat. There were, however, few men in this age group, and there were wide confidence intervals. Additionally, these very old men were likely to represent the healthiest men of this age in the local population (many men in this age group live in nursing homes, making them ineligible for this study). Thus, it is likely that BMD loss continues to accelerate through life, adding to the increasing fracture risk as men age. The pattern of increasing BMD loss with age is similar to BMD loss in older women. Ensrud et al. [53] reported that BMD loss in the total hip increased from 0.5 %/year in women aged 70–74 up to 1.6 %/year from age 85 onwards.

Activity and function

We found a beneficial relationship between walking and decreased BMD loss. We are unaware of any previous studies reporting this relationship in men. Meta-analyses of mixed exercise modes in men and women, which included walking, have reported small positive effects of walking at the lumbar spine [54] and femur in women [55]. However, a controlled exercise intervention trial in middle-aged men, which included walking, found no effect on BMD over 4 years [56]. This intervention was in a much younger and healthier cohort than CHAMP. The protective effect size observed in CHAMP of 0.3 % per year (in those who walked 2–4 km/day though small, is comparable to the size of effects on femoral neck BMD reported in a meta-analysis of walking in post menopausal women [57]. In CHAMP, the positive association between walking and BMD change was not altered by accounting for serum 25(OH)D, participation in sport or other measures of physical activity, number of hours spent sitting each day or number of weeks in hospital or resting in a chair due to illness during the preceding 12 months. Only a small group reported walking more than 4 km/day, and they did not have significantly slower BMD loss than non-walkers. It is possible that this reflects the large confidence intervals because of the small number of men. Alternatively, distance was self-reported and may have been misreported.

Poor balance has been associated with bone markers, indicating decreased bone formation in the elderly [58] and with an increased risk of falls and fractures [59]. Our previous analysis of CHAMP data [60] found that better balance was associated with higher cross-sectional total hip BMD. However, an association between poor balance and greater BMD loss has not previously been reported. The dynamic balance test requires participants to have good hip control in order to move in both anterior–posterior and lateral directions. It is possible that men who do poorly in this test have poorer hip musculature. Although poorer dynamic balance was highly correlated with weaker quadriceps, lower grip strength, inability to walk a narrow path, inability to stand from sitting and slower sit to stand speed, none of these factors were significant in the final multivariate models nor did they attenuate the relationship between poorer balance and BMD change.

Body size and bone size

In final multivariate models, larger hip area was associated with greater BMD loss, while weight was protective. Most [9, 11, 15], but not all [17], studies have found greater weight [11, 15], higher BMI [11, 13, 18, 61], or greater height [11] were protective against absolute or proportional BMD loss. BMI was not associated with BMD change in our study (p = 0.5).

DXA measured hip-bone area is rarely included in epidemiological analyses. However, Prentice et al. [62] recommended that bone area always be included and warned that variables such as muscle strength may spuriously be associated with BMD, when in fact their relationship is derived from their relationship with body size. In our study, the relationship between quadriceps weakness and BMD was attenuated by weight but was not affected by hip bone area. The reason for the association between faster BMD loss and larger hip area is unclear. Potentially, there could be faster bone turnover associated with the larger surface area. Alternatively, it may be a chance finding. Cawthon et al. [11] reported an association between femoral neck area and BMD loss in the MrOS cohort, but, in contrast to our findings, smaller femoral neck area was associated with greater femoral neck BMD loss.

Serum 25(OH)D

The relationship between serum 25(OH)D and the rate of BMD loss did not appear to have a linear relationship. Men in the lowest quintile of serum 25(OH)D (≤36 nmol/l) appeared to have faster BMD loss than men in the highest quintile of 25(OH)D. It was inconclusive whether BMD loss was any different between other groups (p = 0.07). The association between low serum 25(OH)D and BMD in the literature is inconsistent. Two systematic reviews concluded that there was insufficient evidence of an effect of vitamin D supplements on BMD [63, 64]; however, few randomised control trials (RCTs) have accounted for baseline serum 25(OH)D. In contrast to findings from RCTs, observational studies support an association between circulating serum 25(OH)D and increased loss of BMD at the femoral neck. The concentration of circulating serum 25(OH)D below which increased bone loss has been observed, ranges from 30 to 80 nmol/l [6467].

Strengths and limitations

The current report is one of the most comprehensive studies to describe the predictors of BMD change in a population-based cohort of older men. The men in CHAMP are representative of men in the study area in terms of age [68] and have similar health characteristics to older Australian men [69]. Furthermore, CHAMP includes a large group of men over 80 years of age, and the follow-up rate of 85 % of living men is high. Repeated measures of BMD were carried out on the same DXA scanner, and the follow-up time was similar for all men (98 % were followed up between 2 and 2.5 years). In addition, we accounted for a wide range of potential correlates including validated objective clinical measures.

Men who were not included in this analysis because they had died, withdrawn or started on bone medications were older and in poorer health. Thus, our estimates of rates of bone loss may have been conservative. A relatively large proportion of men, (11 %), were excluded because they had taken or were taking osteoporosis medication. Men in the study were primarily white, and all were community dwelling. Our findings may not apply to institutionalised men or other populations. Short-term follow-up studies are subject to relatively large DXA measurement error [70]; however, our rates of bone loss are within the range of those reported in studies with 4–10 years of follow-up [79, 11, 21, 52]. Self-reported information may be affected by poor recall, particularly for exposures a long way in the past. Associations that are marginally significant may need to be replicated in other studies, and non-significant relationships should be viewed with caution particularly if the exposures were uncommon.


In summary, bone loss tends to accelerate as men get older. However, there is considerable variation in the rate of bone loss at all ages, and men with accelerated BMD loss were spread evenly across all the quartiles of baseline BMD. Conversely, a third of the older men in this study maintained their baseline BMD. Walking, greater body weight and beta-blocker medication may attenuate the rate of loss of hip BMD. Low serum 25(OH)D, poor balance and the use of anti-androgen, loop diuretic and thiazolidinedione medications put men at high risk of accelerated bone loss. Fracture risk needs to be considered when these medications are prescribed. Baseline BMD remains an important determinant of follow-up BMD and highlights the importance of achieving a high peak BMD and minimising bone loss over a lifetime.


The CHAMP Study is funded by the Australian National Health and Medical Research Council (NHMRC project grant no. 301916) and the Ageing and Alzheimer’s Research Foundation (AARF). Many thanks to the scientists, Lynley Robinson and Beverly White, for assessing the scans and to the participants who have graciously given their time for this study.

Ethics approval

The CHAMP study was approved by the Concord Hospital Human Research Ethics Committee. Written and informed consent was given by all participants prior to their inclusion in the study.

Conflicts of interest


Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2012