SEARCH

SEARCH BY CITATION

Keywords:

  • socioeconomic status;
  • social disadvantage;
  • BMD;
  • population-based adults

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

With few exceptions, an inverse relationship exists between social disadvantage and disease. However, there are conflicting data for the relationship between socioeconomic status (SES) and BMD. The aim of this study was to assess the association between SES and lifestyle exposures in relation to BMD. In a cross-sectional study conducted using 1494 randomly selected population-based adult women, we assessed the association between SES and lifestyle exposures in relation to BMD. BMD was measured at multiple anatomical sites by DXA. SES was determined by cross-referencing residential addresses with Australian Bureau of Statistics 1996 census data for the study region and categorized in quintiles. Lifestyle variables were collected by self-report. Regression models used to assess the relationship between SES and BMD were adjusted for age, height, weight, dietary calcium, smoking, alcohol consumption, physical activity, hormone therapy, and calcium/vitamin D supplements. Unadjusted BMD differed across SES quintiles (p < 0.05). At each skeletal site and SES index, a consistent peak in adjusted BMD was observed in the mid-quintiles. Differences in adjusted BMD were observed between SES quintiles 1 and 4 (3–7%) and between quintiles 5 and 4 (2–7%). At the spine, the maximum difference was observed (7.5%). In a subset of women, serum 25(OH)D explained a proportion of the association between SES and BMD (difference remained up to 4.2%). Observed differences in BMD across SES quintiles, consistent across both SES indices, suggest that low BMD may be evident for both the most disadvantaged and most advantaged.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

To inform future health policy and disease intervention or prevention, socioeconomic status (SES) is now being recognized as important to public health research.(1,2) There exists an inverse relationship between morbidity and mortality and the gradient of SES.(2,3) SES is often used to account for potential confounding but has been much less studied as a significant risk factor for musculoskeletal diseases. Individuals from lower SES have been shown to have lifestyle behaviors that are less protective of the onset of musculoskeletal disease,(1,3–7) and they are less likely to undergo screening for disease.(2,8,9) However, there have also been suggestions that with increasing affluence comes a change toward less healthy lifestyles(10) and that osteoporosis may be a disease of the affluent.(11) These conflicting views are compounded by inconsistent and contradictory findings in the available literature. A descriptive study identified lower BMD in individuals from rural groups with low SES,(12) whereas another study identified an inverse relationship between BMD and SES.(13) The common factor in these studies was the categorization of SES as a binary variable of low or high, limiting an analysis of the continuum of social experience.

Our primary hypothesis was that differences in BMD would exist across quintiles of area-based SES. The role of potential confounders in accounting for these differences needs to be studied. Our secondary hypothesis was that, with too few categories, sensitivity to measure the potential influence of SES to bone health might be lost. Thus, to examine the relationship between BMD and SES, we used multisite measures of BMD and two alternate indices of area-based SES stratified into quintiles.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

Subjects

An age-stratified sample of 1494 women was selected at random from Commonwealth electoral rolls for the Barwon Statistical Division (BSD) in South Eastern Australia, for enrollment in the Geelong Osteoporosis Study (GOS) during 1994–1997. The participation was 77.1%.(14) Registration with the Australian Electoral Commission is compulsory in Australia for all persons ≥18 yr of age; therefore, the electoral roll provides a comprehensive register of most adult residents.

The BSD has one major city, Geelong, and combined with the immediate suburbs, constitutes the third largest non-capital city in Australia. The surrounding areas include coastal resort towns, small acreage properties, larger more traditional farms, and small townships based on agricultural industries.(15)

In a comparison between the BSD and Australia using the Australian Bureau of Statistics (ABS) census data, marital status and age were similar across all regions.(16) BSD residents were 6.5% less likely to have been born overseas, educational attainment was <5.2% lower, adult weekly income was lower by <2.5%, and type of occupation differed by <3% among women 19 yr of age who were currently employed.(16) Thus, the GOS population is representative overall of the broader Australian population.(14,17–21) This population has been used to define the BMD reference range for Australian women.(14)

All participants in the GOS cohort provided informed written consent. Approval for the study was obtained from the Human Research and Ethics Committee, Barwon Health.

BMD

BMD (g/cm2) was measured for the lumbar region (L2–L4) of the spine in the posterior-anterior (PA) projection, the proximal femur (at the femoral neck, Ward's triangle, and trochanter), total body, ultradistal forearm, and mid-forearm, and performed by DXA (LUNAR DPX-L, Madison, WI, USA). Short-term precision in vivo (calculated as the CV of repeated scans) were, respectively, 0.6%, 1.6%, 2.1%, 1.6%, 0.4%, 2.1%, and 1.1%.(19) Calculation for total hip comprised that of femoral neck, Ward's triangle, and trochanter, as previously published.(22)

SES

The residential address of each subject was matched to the corresponding ABS Census Collection District, an area of ∼250 households. ABS software was used to determine the Socio-Economic Indexes For Areas (SEIFA) value from the 1996 census for each subject. SEIFA is a collection of five separate indices, derived from the Australian Census data and constructed from different variables. SEIFA summarizes the characteristics of subjects within an area, thereby providing a single measure to rank the level of disadvantage at the area level, not of the individual subject. Two of the five SEIFA values are equivalized for both advantages and disadvantages(23): (1) the Index of Education and Occupation (IEO) and (2) Index of Economic Resources (IER). The remaining three SEIFA values are the Index of Relative Socio-Economic Disadvantage (IRSD), the Urban Index of Relative Socio-Economic Advantage (UIA), and the Rural Index of Relative Socio-Economic Advantage (RIA). The IRSD incorporate variables indicating disadvantage alone, and both the UIA and RIA incorporate variables indicating advantage alone. The latter three indices do not lend themselves to be used as a measure of advantage/disadvantage in a continuum. Thus, it was decided a priori to apply the two SEIFA indices of IEO and IER to best examine the continuum of SES in relation to BMD.

The IEO measures the proportion of employed individuals within the area, level of educational attainment, and if employed, the type of occupation held.(23) The IER measures area-based annual household income, rental and mortgage payments, dwelling size, and car ownership.

For comparison, SEIFA values were stratified into quintiles of both the BSD and national data determined by use of 1996 Australian Census data. Cross-tabulation identified BSD and national quintiles of IEO and IER to have excellent agreement of 86.7% and 87.8%, respectively. Cut-points of the BSD and the range of SEIFA values in each quintile for the study population are provided in Table 1. Quintile 1 represented the most disadvantaged and quintile 5 represented the most advantaged. A further analysis was conducted by categorizing subjects into binary SES, using the median value for the study region as a cut-point for upper and lower SES. Residential addresses were recorded before BMD assessment, and the coding of SES was conducted while blinded to BMD results.

Table Table 1.. Range of 1996 SEIFA Values for Study Population and Quintile Cut-Points Weighted by Population to the Barwon Statistical Division
Thumbnail image of

Lifestyle and behavioral exposures

Data for lifestyle and behavioral exposures were obtained by self-report. Dietary calcium (mg/d) was determined by a validated, calcium-specific food frequency questionnaire.(18) Smoking was defined as currently smoking at the time of BMD measurement. Physical activity was categorized as hard or very hard versus moderate or other activity for either home- or work-related and recreational or sport-related activities. Alcohol intake was categorized as consuming two or more standard drinks of alcohol per week or less than two standard drinks per week. Use of hormone therapy (HT), calcium or vitamin D supplements, and oral glucocorticoids was defined as current at the time of BMD measurement. Serum vitamin D [25(OH)D] was measured in a subgroup of 861 women after an overnight fast. Blood samples were assayed using an equilibrium radioimmunoassay after extraction with acetonitrile (Incstar, Stillwater, MN, USA). The intra-assay precision was 6% and interassay precision was 15% as previously reported.(21)

Statistical analysis

Univariate analyses were conducted using ANOVA and Kruskal Wallis. Regression models assessing the relationship between each SES index on BMD at each anatomical site were adjusted for age, height, weight, and further adjustment made for dietary calcium, work/home-related physical activity, recreation/sport-related physical activity, smoking, alcohol intake, HT use, calcium or vitamin D supplements, and oral glucocorticoids. Interaction terms were checked for effect modification. The regression models were further applied to a subset (n = 861) of the population for whom serum had been analyzed for 25(OH)D. To account for season, models included a sinusoidal adjustment for serum 25(OH)D, as previously described.(24) Significance was set at p < 0.05, and statistical analyses were performed using MINITAB (release 15) and STATA (release 9; StataCorp, College Station, TX, USA) software.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

There were no differences between participants and nonparticipants living in the region in any of the SES quintiles identified for the IEO (p = 0.18) or IER (p = 0.62).

The baseline characteristics of the participants (n = 1494) in each quintile, across both SEIFA scores, are presented in Table 2. Across all SES indices, study subjects were more likely to be heavier, older, be current smokers, and consume less alcohol if in quintile 1 compared with quintile 5 (all p < 0.05).

Table Table 2.. Subject Characteristics (n = 1494) Across Each Quintile and Index of SES, Presented as Median (Interquartile Range), Mean (±SE), or Frequency (%)
Thumbnail image of

Unadjusted BMD was similar for subjects in quintiles 1 and 5, although quintile 5 had the lowest BMD at the total hip for both IEO and IER (Table 3). After adjusting for age, height, weight, dietary calcium, physical activity, smoking, alcohol intake, HT use, and calcium/vitamin D supplements, the pattern of association between BMD and SES was sustained (Fig. 1). At all anatomical sites, for the SEIFA index of IEO subjects in quintiles 1 and 5 had lower adjusted point estimate BMD than those in the mid-quintiles. Significant differences between the peak in mid-quintiles and quintile 5 were identified for IEO at every anatomical site and between the peak and quintile 1 for every site with the exception of ultradistal forearm (all p < 0.05; Fig. 1). Significant differences in IER were observed at the spine, total body, and mid-forearm between the peak and quintile 1. The greatest difference in adjusted mean BMD between SES quintiles was 7.5%, observed within IEO at the spine. When categorizing our data by national cut-points, patterns of BMD across SES quintiles were similar to those identified with BSD cut-points, and significant differences were detected across multiple skeletal sites in both indices (data not shown).

Table Table 3.. Unadjusted BMD (g/cm2) Across SES Quintiles, Presented as Mean ± SE
Thumbnail image of
thumbnail image

Figure Figure 1. Adjusted BMD (g/cm2) (±SE) at anatomical sites across SES quintiles of Index of Education and Occupation (IEO) and Index of Economic Resources (IER). Significant differences indicated from aquintile 1 (lowest), bquintile 2, cquintile 3, dquintile 4, or equintile 5 (p value).

Download figure to PowerPoint

Using seasonally adjusted serum 25(OH)D, the association between adjusted BMD and SES remained similar. Statistical significance was maintained in both indices at the spine and total hip and also at the total body and ultradistal forearm in the IEO (p < 0.05). Thus, serum 25(OH)D did not fully explain the association between BMD and SES (data not shown).

With SES categorized as a binary variable, greater BMD was observed at spine and total body, total hip, and mid-forearm in the upper SES group, whereas greater BMD at the ultradistal forearm was observed in the lower SES group (IEO; all p < 0.05; data not shown). A binary division of IER identified greater BMD at the total hip and mid-forearm in the upper SES groups, whereas greater BMD at the remaining three sites was observed in the lower SES group (all p < 0.05).

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

Within our cohort, BMD was similar for subjects in quintile 1 and quintile 5, although quintile 5 had the lowest BMD at the total hip for both IEO and IER. The differences were not explained by age, height, weight, dietary calcium, physical activity, smoking, alcohol intake, HT use, or calcium/vitamin D supplements. The pattern of differences remained when a subset analysis was conducted, which also adjusted for seasonally adjusted serum vitamin D [25(OH)D].

Even small deficits of BMD have been shown to increase the fracture risk in individuals, with a 1.5- to 3.0-fold increase in fracture risk for every 1 SD decrease in BMD.(17) It has been documented that a relationship exists between area-based SES and BMD; however, there are conflicting data as to this relationship.(12,13) Thus, an analysis of BMD using quintiles of SES presents a more sensitive measure than previously applied binary divisions. We observed differences in BMD at all anatomical sites across quintiles of SES, which were independent of lifestyle and behavioral exposures. The greatest difference in mean BMD between SES quintiles 1 and 4 was 7.5% (equivalent to 0.5 SD) and 7.4% between quintiles 4 and 5, both observed in the IEO. This represents a potential 1.5-fold increase in fracture risk for subjects in quintiles 1 and 5. We found that the observed relationship between BMD and SES quintiles was distorted when binary divisions of SES were applied. Greater BMD at the three clinical sites of spine, total body, and total hip was observed in the upper SES groups for the IEO. However, for the IER, BMD was greatest at the spine and total body in the lower SES groups, and BMD of the total hip was greatest in the upper SES group. The distorted results identified when fewer categories of SES were applied may explain why two of the most often cited studies of SES and BMD, Del Rio Barquero et al.(12) and Elliot et al.,(13) have produced conflicting results. Our findings suggest that individuals from both ends of the SES spectrum are at increased risk of fracture compared with individuals from the mid-quintiles.

Greater weight has been identified as associated with lower SES,(25,26) an association also reflected within our study population (Table 2). However, our analysis did not identify any effect modification attributable to weight. Another possible theory that may account for higher BMD in some of the population may be the impact of manual labor on BMD. It has been suggested that manual labor, or work requiring weight-bearing activities, may protect against lower BMD.(27,28) Thus, with many higher paid occupations generally being more sedentary, individuals used in white-collar occupations may be less likely to conduct weight-bearing movements as part of their employment. However, on the opposite end of the spectrum, lower SES often experience more sedentary lifestyles,(2,29) perhaps accentuated by unemployment, therefore also lacking weight-bearing movements. This theory may account for higher BMD in the mid-quintiles, in which manual labor, trade professions, and blue-collar occupations are more likely predominant.

Our findings may provide some support for the theory that adult BMD may be affected by childhood determinants or lifetime accumulation of disadvantage.(30–34) However, Pearce et al.(30) identified that, although childhood growth was a significant predictor of bone area for women, early life variables mediated through later life did not account for the variation in BMD. Adult lifestyle and body composition have been suggested to contribute more to the variation in women's BMD than childhood growth or early life variables.(35)

The strengths of our study include a high participation rate, and a 99% white random sample previously identified as representative of the BSD.(14,19–21) Area-based scores have been identified as an accepted proxy for measuring SES for public health purposes.(2,3,23) The two SEIFA indices used in this study were formulated from various aggregate measures, providing a robust approach to examining SES as a risk factor for bone health. However, SEIFA indices are limited by the information collected as part of the Census and thus contain limited information regarding accumulated wealth; no information concerning infrastructure including schools, services or transport; and they do not provide information as to the cost of living within areas. We recognize that it may be possible for relatively advantaged subjects to reside in an area that scored low on the SEIFA and that the variables on which SEIFA are calculated may not be universally applicable. Thus, care should be taken when interpreting our findings in the context of alternate geographical regions and populations of other ethnicity. Our analyses of BMD may have been confounded by the processes involved in reaching peak bone mass and subsequent bone loss or by unrecognized confounders including comorbidities, medication use, or depression.(36,37) Our cross-sectional study did not exclude subjects with known bone disease or diseases affecting bone; however, further studies on the association between SES, BMD, and bone diseases may elucidate this relationship. Furthermore, the findings of our serum vitamin D [25(OH)D] subset analysis need to be studied further in a larger sample.

In conclusion, the novel finding of this analysis was that individuals from upper SES groups had patterns of relatively low BMD, which was sustained after adjusting for many of the known lifestyle risk factors. The pattern of low BMD in individuals from more disadvantaged SES groups also tended to support the widely documented relationship between these individuals and poorer health. Thus, individuals from both ends of the SES continuum may face unique barriers to protective levels of BMD.(10,11,13) The absence of significant differences in BMD between the mid-quintile and quintile 5 when using the IER suggests that annual household income may mediate the relationship. However, SES and health is a complex concept, and the explanation of differences in BMD between SES quintiles warrants further study as an antecedent to potential interventions or preventive public health campaigns.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES

We thank the participants who made this study possible. This study was funded by the National Health and Medical Research Council (NHMRC) of Australia and The Victorian Health Promotion Foundation. S.L.B. was supported by NHMRC PhD Scholarship (519404). A.E.W. is the recipient of an NHMRC Public Health Training Fellowship (317840).

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Acknowledgements
  8. REFERENCES
  • 1
    Begg S, Vos T, Barker B, Stanley L, Lopez AD 2008 Burden of disease and injury in Australia in the new millennium: Measuring health loss from diseases, injuries and risk factors. Med J Aust 188: 3640.
  • 2
    Wilkinson R, Marmot M 1998 Social Determinants of Health: The Solid Facts. WHO European Region, Copenhagen, Denmark.
  • 3
    Turrell G, Stanley L, deLooper M, Oldenburg B 2006 Health Inequalities in Australia: Morbidity, Health Behaviours, Risk Factors and Health Service Use. Queensland University of Technology and the Australian Institute of Health and Welfare, Canberra, Australia.
  • 4
    Subar A, Heimendinger J, Patterson BH, Krebs-Smith SM, Pivonka E, Kessler R 1995 Fruit and vegetable intake in the United States: The baseline survey of the Five A Day for Better Health Program. Am J Health Promot 9: 352360.
  • 5
    Mazess R, Barden HS 1999 Bone density of the spine and femur in adult white females. Calcif Tissue Int 65: 9199.
  • 6
    Peach H, Bath NE 2000 Preventing fractures in large rural centres: Sociodemographic sub-groups at risk of osteoporosis from their lifestyle. Aust J Rural Health 8: 299304.
  • 7
    Hagen K, Zwart JA, Svebak S, Bovim G, Stovner LJ 2005 Low socioeconomic status is associated with chronic musculoskeletal complaints among 46,901 adults in Norway. Scand J Public Health 33: 268275.
  • 8
    Neuner JM, Zhang X, Sparapani R, Laud PW, Nattinger AB 2007 Racial and socioeconomic disparities in bone density testing before and after hip fracture. J Gen Intern Med 22: 12391245.
  • 9
    Demeter S, Leslie WD, Lix L, MacWilliam L, Finlayson GS, Reed M 2007 The effect of socioeconomic status on bone density testing in a public health-care system. Osteoporos Int 18: 153158.
  • 10
    Ho S, Chen Y, Woo JLF 2005 Educational level and osteoporosis risk in postmenopausal Chinese women. Am J Epidemiol 161: 680690.
  • 11
    Mallmin H, Ljunghall S, Persson I, Bergstrom R 1994 Risk factors for fractures of the distal forearm: A population-based case-control study. Osteoporos Int 4: 298304.
  • 12
    Del Rio Barquero L, Romera Baures M, Pavia Segura J, Setoain Quinquer J, Serra Majem L, Garces Ruiz P, Lafuente Navarro C, Domenech Torne FM 1992 Bone mineral density in two different socio-economic population groups. Bone Miner 18: 159168.
  • 13
    Elliot JR, Gilchrist NL, Wells JE 1996 The effect of socioeconomic status on bone density in a male Caucasian population. Bone 18: 371373.
  • 14
    Henry MJ, Pasco JA, Nicholson GC, Seeman E, Kotowicz MA 2000 Prevalence of osteoporosis in Australian women: Geelong Osteoporosis Study. J Clin Densitom 3: 261268.
  • 15
    Sanders KM, Nicholson GC, Ugoni AM, Seeman E, Pasco JA, Kotowicz MA 2002 Fracture rates lower in rural than urban communities: The Geelong Osteoporosis Study. J Epidemiol Community Health 56: 466471.
  • 16
    Henry MJ 2000 Site specific bone mineral density in Australian women Department of Clinical and Biomedical Sciences. PhD dissertation, University of Melbourne, Melbourne, Australia.
  • 17
    Pasco JA, Seeman E, Henry MJ, Merriman EN, Nicholson GC, Kotowicz MA 2006 The population burden of fractures originates in women with osteopenia, not osteoporosis. Osteoporos Int 17: 14041409.
  • 18
    Pasco JA, Sanders KM, Henry MJ, Nicholson GC, Seeman E, Kotowicz MA 2000 Calcium intakes among Australian women: Geelong Osteoporosis Study. Aust N Z J Med 30: 2127.
  • 19
    Pasco JA, Henry MJ, Sanders KM, Kotowicz MA, Seeman E, Nicholson GC 2004 B-adrenergic blockers reduce the risk of fracture partly by increasing bone mineral density: Geelong Osteoporosis Study. J Bone Miner Res 19: 1924.
    Direct Link:
  • 20
    Pasco JA, Henry MJ, Kotowicz MA, Collier GR, Ball MJ, Ugoni AM, Nicholson GC 2000 Serum leptin levels are associated with bone mass in nonobese women. J Clin Endocrinol Metab 86: 18841887.
  • 21
    Pasco JA, Henry MJ, Nicholson GC, Sanders KM, Kotowicz MA 2001 Vitamin D status of women in the Geelong Osteoporosis Study: Association with diet and casual exposure to sunlight. Med J Aust 175: 401405.
  • 22
    Mazess RB, Barden HS 1999 Bone density of the spine and femur in adult white females. Calcif Tissue Int 65: 9199.
  • 23
    Australian Bureau of Statistics 1996 Information Paper: Census of Population and Housing-Socioeconomic Indexes for Areas, Australia V 2033.0. Australian Government Publishing Service, Canberra, Australia.
  • 24
    Pasco JA, Henry MJ, Kotowicz MA, Sanders KM, Seeman E, Pasco JR, Schneider HG, Nicholson GC 2004 Seasonal periodicity of serum vitamin D and parathyroid hormone, bone resorption, and fractures: The Geelong Osteoporosis Study. J Bone Miner Res 19: 752758.
    Direct Link:
  • 25
    Manson JE, Lewis CE, Kotchen JM, Allen C, Johnson KC, Stefanick M, Foreyt J, Klesges R, Tinker L, Noonan E, Perri MG, Hall WD 2001 Ethnic, socioeconomic, and lifestyle correlates of obesity in U.S. women: The Women's Health Initiative. Clin J Womens Health 1: 225234.
  • 26
    Sobal J, Stunkard AJ 1989 Socioeconomic status and obesity: A review of the literature. Psychol Bull 105: 260275.
  • 27
    Qureshi HN, Najib U, Chaudhary UJ, Sallahuddin S, Shah IA, Akram J 2003 Effect of long standing time on bone mineral density in middle-aged males. Asian Ocean J Radiol 8: 8183.
  • 28
    Kondo T, Sakakibara H, Miyao M, Yamada S, Shirota T, Izuhara Y, Kumagai K 1995 Relationship of bone density with lifestyle and labor factors among middle-aged and elderly women in a rural community. Bull Phys Fit Res Inst 88: 98103.
  • 29
    Lindquist CH, Reynolds KD, Goran MI 1999 Sociocultural determinants of physical activity among children. Prev Med 29: 305312.
  • 30
    Pearce M, Birrell FN, Francis RM, Rawlings DJ, Tuck SP, Parker L 2005 Lifecourse study of bone health at age 49–51 years: The Newcastle thousand families cohort study. J Epidemiol Community Health 59: 475480.
  • 31
    Fehily AM, Coles RJ, Evans WD, Elwood PC 1992 Factors affecting bone density in young adults. Am J Clin Nutr 56: 579586.
  • 32
    Arabi A, Nabulsi M, Maalouf J, Choucair M, Khalife H, Vieth R, El-Hajj Fuleihan G 2004 Bone mineral density by age, gender, pubertal stages, and socioeconomic status in healthy Lebanese children and adolescents. Bone 35: 11691179.
  • 33
    Jaglal S, Kreiger N, Darlington GA 1995 Lifetime occupational physical activity and risk of hip fracture in women. Ann Epidemiol 5: 321324.
  • 34
    Korpelainen R, Korpelainen J, Heikkinen J, Vaananen K, Keinanen-Kiukaanniemi S 2006 Lifelong risk factors for osteoporosis and fractures in elderly women with low body mass index-A population-based study. Bone 39: 385391.
  • 35
    Pearce MS, Birrell FN, Francis RM, Rawlings DJ, Tuck SP, Parker L 2005 Lifecourse study of bone health at age 49–51 years: The Newcastle thousand families cohort study. J Epidemiol Community Health 59: 475480.
  • 36
    Jacka FN, Pasco JA, Henry MJ, Kotowicz MA, Dodd S, Nicholson GC, Berk M 2005 Depression and bone mineral density in a community sample of perimenopausal women: Geelong Osteoporosis Study. Menopause 12: 8891.
  • 37
    Koster A, Bosma H, Kempen GI, Penninx BW, Beekman AT, Deeg DJ, van Eijk JT 2006 Socioeconomic differences in incident depression in older adults: The role of psychosocial factors, physical health status, and behavioral factors. J Psychosom Res 61: 619627.