Familial aggregation of bone mineral density and bone mineral content in a Chinese population
- First Online:
- Cite this article as:
- Feng, Y., Hsu, YH., Terwedow, H. et al. Osteoporos Int (2005) 16: 1917. doi:10.1007/s00198-005-1962-9
- 60 Downloads
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.
KeywordsBone mineral content Bone mineral density Chinese Familial aggregation Siblings
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 . 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.  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
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.
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.
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.
Characteristics of the study population stratified by gender, age and menopausal status (mean ± SD or %)
Age <45 years
Age ≥45 years
Lumbar spine (L2–L4)a
Years since menopause
Number of births
Never or occasional
Never or occasional
Under middle school
Middle school and above
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
Age <45 years
Age ≥45 years
Lumbar spine (L2–L4)b
Lumbar spine (L2–L4)b
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 . 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.
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 . One is the uniform weight estimator proposed by Smith , which has a very high efficiency only if the ICC is high (i.e., greater than 0.30) . 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) . 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.  reported that the correlation coefficients between DZ twins for hip and spine BMD ranged from 0.33 to 0.47. Krall et al.  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.  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 . 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 .
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 .
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.
This research was supported by a grant from the National Institutes of Arthritis and Musculoskeletal and Skin Disease (R01 AR045651).