Osteoporosis International

, Volume 16, Issue 12, pp 1917–1923

Familial aggregation of bone mineral density and bone mineral content in a Chinese population

  • Yan Feng
  • Yi-Hsiang Hsu
  • Henry Terwedow
  • Changzhong Chen
  • Xin Xu
  • Tianhua Niu
  • Tonghua Zang
  • Di Wu
  • Genfu Tang
  • Zhiping Li
  • Xiumei Hong
  • Binyan Wang
  • Joseph D. Brain
  • Steven R. Cummings
  • Clifford Rosen
  • Mary L. Bouxsein
  • Xiping Xu
Original Article

DOI: 10.1007/s00198-005-1962-9

Cite this article as:
Feng, Y., Hsu, YH., Terwedow, H. et al. Osteoporos Int (2005) 16: 1917. doi:10.1007/s00198-005-1962-9

Abstract

Familial aggregation of bone mineral density (BMD) and bone mineral content (BMC) has been shown in twin and familial studies, but most sample sizes were small. We here report a large familial aggregation study in a Chinese population. A total of 13,973 siblings aged 25–64 years from 3,882 families were enrolled from Anhui, China. We assessed the whole-body, hip and lumbar spine BMD and BMC by dual-energy X-ray absorptiometry (DXA). Intra-class correlation coefficients of BMD and BMC between siblings varied among different skeletal sites and between different age groups of male sib-pairs and premenopausal and postmenopausal female sib-pairs, with a range of 0.228 to 0.397. The sibling recurrence risk ratio (λs) of osteoporosis was 2.6 in our population. We also evaluated the joint association of the BMD values of the first siblings and the second siblings with the risk of low BMD (defined as less than the 10th percentile of the same group population) of their younger siblings. If both the first and second siblings’ BMDs were in the lowest tertile, the odd ratios (ORs) of low BMD in their subsequent siblings were 8.32 [95% confidence interval (CI) 5.59–12.39)], 8.71 (95% CI 5.74–13.22) and 5.90 (95% CI 3.57–9.76) for total body, total hip and lumbar spine, respectively. This study demonstrates a significant familial aggregation of BMD and BMC in a large sample of rural Chinese adults.

Keywords

Bone mineral content Bone mineral density Chinese Familial aggregation Siblings 

Introduction

Osteoporosis, which involves a reduction in bone mineral density (BMD) and bone strength, leading to an increased risk of fracture, has become a serious public health issue worldwide [1]. Low BMD is among the most important risk factors for fracture in the elderly [2, 3, 4]. Although both BMD and bone mineral content (BMC) are influenced by environmental factors, such as weight, calcium intake and exercise, twin and familial aggregation studies have demonstrated that genetic factors may explain a large proportion of the variability in BMD and BMC [5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19].

Smith et al. [5] first described the difference between the correlation coefficients of forearm bone mass in adult dizygotic (DZ) (r=0.451) and monozygotic (MZ) (r=0.698) twins and estimated the heritability of bone mass to be 0.36. Since that time, a number of twin studies have confirmed the genetic influences on BMD and BMC and estimated that heritability in different age groups and skeletal sites ranges from 0.42 to 0.92 [6, 7, 8, 9, 18]. The sample sizes of these studies were relatively small (30–71 twin pairs), and the ages of subjects varied, both of which may have partially contributed to the great variability of estimated heritability.

In addition to twin studies, reduced bone mass in relatives of osteoporotic patients was also reported as early as 16 years ago [10, 11]. During the past decade, a number of studies have reported familial aggregation of BMD at different skeletal sites and in different age groups [12, 13, 14, 15, 16, 17, 19]. Although significant familial aggregation was demonstrated, the correlation coefficients of BMD between family members varied among the different studies. This variation among studies may be explained in part, because it has been shown that age, gender and skeletal site influence the heritability of BMD [13, 17]. An additional reason for variability across studies is that, with a few exceptions [13, 19], these familial studies have been conducted using relatively small sample sizes (40–147 families).

With only a few exceptions [18, 19], twin and familial studies have been conducted in Caucasian populations. More data from non-Caucasian populations will contribute to our understanding of the ethnic differences in BMD and its heritability.

Thus, our objective was to evaluate familial aggregation of BMD and BMC using a large sample (3,882 families, 13,973 siblings) of Chinese individuals. We also evaluated the impact of gender, age and menopausal status on sibling-pair (sib-pair) resemblance of BMD and BMC.

Materials and methods

Study site

This study was conducted in Anqing, Anhui Province, China. Anqing stretches for approximately 80 km along the north bank of the Yangtze River. It has three urban areas and eight rural counties, with a total area of 15,300 km2. The current total population is 6.1 million (10% urban and 90% rural). Available records indicate that the Anqing district was settled on 2,000 years ago. The population has a very stable base and is extremely homogeneous with respect to ethnicity, dietary habits, lifestyle, and environmental factors.

Identification of eligible sib-pair families

The present study enrolled a total of 13,973 siblings from 3,882 families. The inclusionary criterion was a minimum of three participating siblings of 25 to 64 years of age, of whom two members must be 40–64 years old. Individuals with a history of the following conditions were excluded from further study: type 1 diabetes mellitus, renal failure, chronic infections such as tuberculosis malignancy, rickets or other metabolic bone diseases, chronic glucocorticoid use, and thyrotoxicosis. Also excluded were those women who could not rule out their being pregnant.

Phenotype collection

The study was initiated in October 2003, and is on-going, by a team of locally hired and well-trained interviewers, along with help from our collaborator, Anhui Medical University. Eligible participants were selected, and the following procedures were carried out: (1) a questionnaire was used to collect the subjects’ date of birth, disease history, dietary habits, occupation, physical activity, history of smoking and alcohol consumption, and, for the women, menstruation and reproductive history; (2) BMC (in grams) and BMD (in grams per square centimeter) of whole-body, total hip and lumbar spine were measured by dual-energy X-ray absorptiometry (GE-Lunar, USA). The whole-body and hip scans were done for each subject, but the lumbar spine scan was added partway through the study, so, therefore, we obtained these data from only 5,755 subjects; and (3) anthropometry measurement and physical examination were carried out. Height was measured to the nearest 0.1 cm on a portable stadiometer and weight to the nearest 0.1 kg. Body mass index (BMI) was calculated as weight/height2 (in kilograms per square meter). The protocol was approved by the Institutional Review Boards of the Harvard School of Public Health and Anhui Medical University.

Statistical analyses

We used a family size-weighted estimator to calculate the intra-class correlation coefficient (ICC) of BMD and BMC values between siblings. The study population was stratified into four groups: men <45 years old, men ≥45 years old, premenopausal women, and postmenopausal women, and the ICC was estimated within the subgroup siblings. The statistical significance of ICC was tested by F test, and the difference between groups was tested by comparison of the 95% confidence interval (CI) of two ICCs. Meanwhile, to control for other confounders of BMD/BMC, we set up a regression model that included age, gender, height, weight, BMI, smoking status in men, and years since menopause in women, to calculate each individual’s residual BMD and BMC in the four above subgroups separately. The residual BMD or BMC was used in the estimation of the ICC and the subsequent analysis.

Using logistic regression analysis, we assessed the odds ratios (ORs) of low BMD among younger siblings in relation to their first and second siblings’ BMD status. In this case, low BMD was defined as an individual’s residual BMD being in the lowest 10th percentile of the same gender group. In the present study, we used the 25–35 year-old men (n=382) and women (n=514) as reference peak bone density groups to calculate a T-score for BMD. Thus, the lowest 10th percentile roughly equaled a T score of <−1.3 for total body, total hip and lumbar spine BMD. The first and second siblings’ BMD status were classified as high, middle or low tertile in the same gender group. The general estimated equation (GEE) model was used for the correction of multiple sib-pairs from the same family. In estimation of the sibling recurrence risk ratio (λs), osteoporosis was defined as the individual’s total hip BMD T-score below −2.5.

All the above analyses were performed with SAS 8.2 (SAS Institute, Cary, N.C., USA). All P values were two-tailed, with statistical significance defined as P<0.05.

Results

The present study included a total of 13,973 subjects from 3,882 families with complete data on total body and hip BMD/BMC and major covariates. We obtained lumbar spine BMD/BMC data for 5,755 of the subjects. General demographic and lifestyle characteristics are shown in Table 1. In general, our study population is quite lean, with a mean BMI of only 21.8 kg/m2. Most subjects (86.0%) reported farming as their occupation. Regular exercise was very rare in this population (<1%), and the majority of people (99%) did not consume milk or take calcium supplements regularly. Thus, these variables were not considered in the analyses. The frequency of current smokers and alcohol drinkers was high in men (70.1% and 46.7%, respectively), but was very low in women (3.0% and 2.9%, respectively).
Table 1

Characteristics of the study population stratified by gender, age and menopausal status (mean ± SD or %)

Variables

Men

Women

Age <45 years

Age ≥45 years

Premenopausal

Postmenopausal

(n=2,711)

(n=4,428)

(n=4,585)

(n=2,249)

Age (years)

39.3±3.7

52.2±4.8

41.5±5.3

52.6±4.7

Height (cm)

164.5±5.8

163.0±5.8

153.9±5.0

152.6±5.1

Weight (kg)

58.6±8.1

56.8±8.0

53.1±7.4

51.1±7.7

BMI (kg/m2)

21.6±2.6

21.3±2.5

22.4±2.8

21.9±2.9

BMD (g/cm2)

  Total body

1.145±0.074

1.129±0.080

1.108±0.069

1.021±0.086

  Total hip

1.001±0.115

0.956±0.125

0.972±0.112

0.863±0.120

  Lumbar spine (L2–L4)a

1.100±0.130

1.040±0.146

1.099±0.123

0.931±0.138

Years since menopause

--

--

--

6.1±5.3

Number of births

--

--

2.2±1.2

3.3±1.4

Smoking statusb

  Current

68.5%

71.1%

1.9%

5.2%

  Former

6.9%

11.1%

0.1%

0.6%

  Never or occasional

24.6%

17.8%

98.0%

94.2%

Alcohol drinkingc

  Current

44.1%

48.4%

2.5%

3.7%

  Former

2.4%

4.7%

0.3%

0.8%

  Never or occasional

53.5%

46.9%

97.2%

95.5%

Education

  Under middle school

42.7%

73.6%

88.0%

97.3%

  Middle school and above

57.3%

26.4%

12.0%

2.7%

Occupation

  Farmer

79.4%

84.2%

89.6%

90.1%

  Worker

3.4%

2.2%

1.3%

0.6%

  House wife/husband

0.3%

1.0%

4.5%

7.1%

Physical activity

  Slight

13.7%

12.5%

14.0%

16.4%

  Moderate

55.2%

50.1%

53.0%

52.2%

  Heavy

31.1%

37.4%

33.0%

31.4%

aThe numbers of subjects in <45-year-old men, ≥45-year-old men, premenopausal women and postmenopausal women were 1,398, 1,594, 2,017 and 746, respectively

bDefined as “current” if the subject had smoked more than ten packs in total and had smoked at lease one during the past year; “former” if ≥10 packs in total but none during the past year; “never or occasional” if <10 packs in total

cDefined as “current” if the subject had consumed more than one drink per week during the past year; “former” if subject used to have ≥1 drink/week but <1drink/week during the past year; “never or occasional” if <1 drink/week

The distribution of total body, total hip and lumbar spine BMD in men <45 years old, men ≥45 years old, and premenopausal and postmenopausal women are also presented in Table 1. Consistent with previous reports, our data showed that (1) men had higher BMD than the corresponding age-group of women, (2) at most skeletal sites older men (≥45 years ) and postmenopausal women had lower BMD than younger men and premenopausal women, respectively. Table 2 shows the intra-class correlation coefficients (ICCs) for BMD and BMC between sib-pairs. The ICCs of BMD, which ranged from 0.251 to 0.380, and those of BMC, which ranged from 0.228 to 0.397, were all statistically significant and varied by skeletal site and subgroup.
Table 2

Intra-family correlation coefficient of residual BMD and BMC between siblings. BMD/BMC was adjusted by age, weight, height and BMI, plus smoking status for men and years since menopause for postmenopausal women, respectively

Skeletal site

Men

Women

Total

Age <45 years

Age ≥45 years

Total

Premenopausal

Postmenopausal

(na=2,349)

(na=643)

(na=1,281)

(na=2,189)

(na=1,357)

(na=435)

BMD

  Total body

0.341

0.372

0.329

0.304

0.317

0.290

  Total hip

0.299

0.373

0.278

0.273

0.298

0.251

  Lumbar spine (L2–L4)b

0.301

0.380

0.308

0.273

0.274

0.294

BMC

  Total body

0.393

0.379

0.387

0.376

0.397

0.300

  Total hip

0.266

0.344

0.228

0.257

0.269

0.250

  Lumbar spine (L2–L4)b

0.300

0.338

0.301

0.239

0.245

0.276

aNumber of families

bThe numbers of families in total men, <45-year-old men, ≥45-year-old men, total women, premenopausal women and postmenopausal women were: 984, 372, 441, 894, 609 and 136, respectively.

The correlation of BMD was significantly higher in male siblings than in female siblings, a finding that was seen at all three skeletal sites. In subgroups, younger male sib-pairs had higher ICCs than their older counterparts, and premenopausal women sib-pairs had higher ICCs than postmenopausal women (except for lumbar spinal BMD). Similar to BMD, the ICCs for BMC were also significantly higher in male siblings than those in female siblings at all three skeletal sites. Also similar to trends seen with BMD, younger male sib-pairs generally had higher ICCs for BMC than their older counterparts, and premenopausal women sib-pairs generally had higher ICCs than postmenopausal women. For comparison, the ICCs of height and weight, calculated using the same approach, were 0.404 and 0.364 in male siblings, 0.322 and 0.278 in female siblings, respectively.

The sibling recurrence risk ratio (λs), which is defined as the ratio of the prevalence of disease in the siblings of affected subjects versus the population prevalence, is widely used in the estimation of familial aggregation of disease. The λs value can be used to estimate the power that affected-sib-pair methods have to detect linkage [20]. In the present study, the prevalence of osteoporosis, defined as total hip BMD T-score below −2.5, was 2.08% and 5.43%, in the total population and the siblings of probands, respectively. Therefore, the estimated λs of osteoporosis was 2.6.

Finally, we evaluated the joint association of the BMD values of the first and second siblings with the risk of low BMD (defined as less than the 10th percentile of the same gender group) of their younger siblings. Each individual’s BMD was ranked by his/her residual BMD, which was adjusted by age, weight, height, BMI, plus smoking status in men and years since menopause in women. Thus, we studied nine groups defined by the first and second siblings’ residual BMD tertiles (Fig. 1). If we treat the case where both the first and second siblings are in the highest tertile (high–high group) as the reference, then, if either the first or the second sibling were in the lowest tertile, the risk of low BMD in their subsequent siblings increased significantly. If both the first and second siblings were in the lowest tertile (low–low group), the ORs of low BMD of their subsequent siblings were 8.32 (95% CI 5.59–12.39), 8.71 (95% CI 5.74–13.22) and 5.90 (95% CI 3.57–9.76) for total body, total hip and lumbar spine, respectively.
Fig. 1

OR of lowest 10th percentile of total body (a), total hip (b) and lumbar spine (c) BMD in the subsequent siblings by tertiles of the first and second siblings’ corresponding BMD. BMD rank was derived from each subject’s residual BMD, which was adjusted for age, weight, height and BMI for men and women, plus smoking status for men and years since menopause for postmenopausal women. *P<0.05, **P<0.01

Discussion

We report the largest familial aggregation study of BMD and BMC thus far. This large sample size allows us to compare the correlation of BMD and BMC between different age groups of male sib-pairs, and premenopausal and postmenopausal female sib-pairs. We demonstrate significant familial resemblance in BMD and BMC for total body, total hip and lumbar spine.

Because each family had more than two siblings, we used the intra-class correlation analysis to maximize the information available. Two methods have been widely used in the literature in the ICC estimation [21]. One is the uniform weight estimator proposed by Smith [22], which has a very high efficiency only if the ICC is high (i.e., greater than 0.30) [21]. The other method is the family size-weighted estimator, known as Fisher’s estimator, which is highly efficient when the ICC is small (i.e., less than 0.30) [21]. Therefore, we chose the latter for our analyses. The ICCs for total body, total hip and lumbar spine BMD between sib-pairs in the present study are consistent with those in previous reports in the literatures [7, 12, 14]. For instance, Pocock et al. [7] reported that the correlation coefficients between DZ twins for hip and spine BMD ranged from 0.33 to 0.47. Krall et al. [12] showed that the correlation of BMD Z-score of total body and lumbar spine between midparent–offspring ranged from 0.34 to 0.54.

In the present study, we found that the correlation coefficients of BMD between sib-pairs varied with gender, age group and skeletal site. In both men and women, younger sib-pairs had higher ICCs for BMD than older sib-pairs. Wu et al. [23] studied BMD in a large, female Chinese population aged from 10 years to 90 years and found that the peak BMD (PBMD) at various skeletal sites occurred within the age range of 30–44 years. Another study, of Chinese men in Taiwan, showed that the PBMD values occurred in the 20–30 year age group [24]. In another independent study, which investigated the BMD distribution in 4,118 pairs of twins aged from 5 years to 65 years and from the same area as the present study, the PBMDs of total body, total hip and lumbar spine (L2–L4) were within the age range of 35–40 years, 20–25 years and 20–25 years, respectively (unpublished data). The higher correlation of BMD between young siblings may indicate that there is a stronger genetic component for PBMD than for the bone loss that occurs later in the aging process, or that the younger siblings share more similar environmental factors than older siblings, or even that there is gene–environment interaction. Our study population is from a Chinese rural area and most people marry when they are quite young (in the present study, 81.3% men and 96.2% women were married in the 25–30 years age group) and lived separately from their siblings. Thus, the similarity of environmental exposure between younger sib-pairs (aged 25–45 years) may be slightly, but not much, greater than that of older sib-pairs. If we had had parents available, we could have conducted segregation analyses and variance component analyses to further evaluate the genetic and environmental components in the familial resemblance of BMD. Two twin studies of BMD, which were conducted in 10–26 year-old females and 25–80 years old females separately, also found that genetic variances were age-related and highest in early adulthood [9, 25]. Meanwhile, they also found evidence for environmental effects shared by twins on BMD and concluded that the genetic and environmental etiology of BMD is more complex than previously thought [25].

In the present study, the correlation of BMD between siblings varied by skeletal site as well. This agrees with many studies in mice showing that the genetic control of bone mass and morphology is highly site-specific [26, 27]. A previously reported familial segregation analysis also demonstrated that the BMD at different skeletal sites is likely determined by not only shared (pleiotropic) genetic and environmental effects but also by site-specific genetic factors [28].

The familial aggregation of BMD found in this study may result from genetic and/or environmental factors. A number of factors, including age, gender, menopausal status, weight, BMI, body composition and smoking status have been reported to influence BMD values [24, 29, 30, 31, 32]. One unique feature of our study is that we used residual BMD or BMC in all of the analyses, such that the BMD values were adjusted by age, height, weight, BMI, smoking status in men, and years since menopause in women. This approach may have decreased the environmental component in the familial aggregation study to some extent. However, due to the complex etiology of BMD, additional studies are still needed to identify the exact contributions of genetic and environmental factors, as well as their possible interactions.

Because there is a significant correlation of BMD between siblings, one can predict an individual’s risk of low BMD according to his/her older siblings’ BMD, an approach that may be useful in clinical practice. We found that the prevalence of osteoporosis in the affected subjects’ siblings was 2.6 times that of the general population. On the other hand, even if osteoporosis had not occurred in a family, our data also clearly showed that the lower the first and/or second siblings’ BMDs, the higher the risk of low BMD in their subsequent siblings. If both the first and second siblings’ BMDs were distributed in the lowest tertile, their younger siblings will have an eight-times higher risk of having their total body or total hip BMDs in the lowest 10th percentile. These data indicate not only a high correlation of BMD among siblings but also a significant predictive value of low BMD and osteoporosis among siblings.

Our study population, which is from an underdeveloped rural area of China, offers a unique opportunity to study genetic and environmental determinants of BMD. In addition to characteristics such as a stable resident population, homogeneity, and a large family size, the rural environment, abundance of physical activity, and lack of use of medication and calcium supplements contrast sharply with the typical urban and suburban settings in the USA. A major weakness of our study is that we only have siblings without parents. No husband–wife and/or parent–child pairs were available for the analysis, and thus the heritability of BMD could not be estimated. Other limitations of the study include its cross-sectional design and the potential information bias with regard to dietary and environmental/occupational variables, given that the data were collected by interview. Nevertheless, the main phenotypes, such as BMD, BMC, height and weight, were measured objectively, and standard protocol and quality-control procedures were employed systematically and consistently throughout the project.

In summary, our data suggest a strong familial aggregation of BMD and BMC in this rural Chinese population. We also found that the intra-class correlations for BMD and BMC between siblings differ by age, gender and skeletal site. This study provides useful information for future genetic studies of BMD, as well as aiding healthcare providers by allowing identification of patients with a high risk of developing osteoporosis according to their family history.

Acknowledgements

This research was supported by a grant from the National Institutes of Arthritis and Musculoskeletal and Skin Disease (R01 AR045651).

Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2005

Authors and Affiliations

  • Yan Feng
    • 1
  • Yi-Hsiang Hsu
    • 1
  • Henry Terwedow
    • 1
  • Changzhong Chen
    • 1
  • Xin Xu
    • 1
    • 1
    • 2
  • Tianhua Niu
    • 1
  • Tonghua Zang
    • 2
  • Di Wu
    • 2
  • Genfu Tang
    • 2
  • Zhiping Li
    • 2
  • Xiumei Hong
    • 2
  • Binyan Wang
    • 1
  • Joseph D. Brain
    • 1
    • 3
  • Steven R. Cummings
    • 4
  • Clifford Rosen
    • 5
  • Mary L. Bouxsein
    • 6
  • Xiping Xu
    • 1
    • 1
    • 2
  1. 1.Program for Population GeneticsHarvard School of Public HealthBostonUSA
  2. 2.Anhui Medical University Institute of MedicineAnhuiChina
  3. 3.Department of Enironmental HealthHarvard School of Public HealthBostonUSA
  4. 4.San Francisco Coordinating CenterUCSFSan FranciscoUSA
  5. 5.Maine Center for Osteoporosis Research and EducationSt. Joseph HospitalBangorUSA
  6. 6.Beth Israel Deaconess Medical CenterBostonUSA

Personalised recommendations