Effect of Environmental Factors and Gender on the Heritability of Bone Mineral Density and Bone Size


  • M. Y. M. Ng,

    1. Department of Medicine, The University of Hong Kong, Queen Mary Hospital, Hong Kong, China
    2. Genome Research Centre, The University of Hong Kong, Hong Kong, China
    Search for more papers by this author
  • P. C. Sham,

    1. Genome Research Centre, The University of Hong Kong, Hong Kong, China
    2. Department of Psychiatry, The University of Hong Kong, Hong Kong, China
    3. Institute of Psychiatry, King's College London, United Kingdom
    Search for more papers by this author
  • A. D. Paterson,

    1. Program in Genetics and Genomic Biology and Department Public Health Sciences, Hospital for Sick Children, and Department of Public Health Sciences, University of Toronto, Canada
    Search for more papers by this author
  • V. Chan,

    1. Department of Medicine, The University of Hong Kong, Queen Mary Hospital, Hong Kong, China
    Search for more papers by this author
  • A. W. C. Kung

    Corresponding author
    1. Department of Medicine, The University of Hong Kong, Queen Mary Hospital, Hong Kong, China
    Search for more papers by this author

*Corresponding author: Annie W. C. Kung, Department of Medicine, The University of Hong Kong, Queen Mary Hospital, Hong Kong, China. Telphone: (852)-2855-4769; Fax: (852)-2816-2187; E-mail: awckung@hkucc.hku.hk


Bone mineral density (BMD), a risk factor for osteoporosis, is believed to be under genetic control. The effect of environmental factors and gender on the heritability of BMD and bone size is ill-defined. In this study, heritability estimates (h2) were determined in 3,320 southern Chinese subjects from 1,019 families using the variance components model. The h2 for age, weight and height-adjusted BMD was 0.63–0.71 for females, and 0.74–0.79 for males; and for bone size, 0.44–0.64 for females and 0.32–0.86 for males. Adjustment for lifestyle factors including calcium and phytoestrogen intake, exercise, smoking and alcohol consumption altered the h2 differently in males and females. The proportion of variance in BMD and bone size explained by all covariates varied between skeletal sites, but was consistently greater in females than males. A significant gender difference was observed in the genetic variance of BMD and bone size at the hip but not the spine. In conclusion, a gender difference was observed in the degree of heritability of BMD and bone size at specific skeletal sites. Environmental influences contributed variably at different sites in the two sexes.


Osteoporosis is a significant worldwide public health problem because of the rising number of osteoporotic fractures. By 2050, based on current fracture data and population growth estimation, 50% of the world's hip fractures are predicted to occur in Asia (Cooper et al. 1992). Within Asia, there is great variation in the incidence of fractures: the highest incidence is reported in Hong Kong and Singapore and the lowest in northern China (Cooper et al. 1992). Osteoporosis is characterised by low bone mineral density (BMD) and microarchitectural deterioration, with a consequent increased bone fragility and susceptibility to fracture (Genant et al. 1999). In addition, bone size and geometry, architecture, and bone turnover also contribute to bone strength (NIH, 2001). Nonetheless low BMD remains a strong independent risk factor for subsequent fracture. BMD and bone size are the two bone parameters that can most easily and precisely be clinically defined.

Bone mass in adulthood is a culmination of the peak bone mass acquired during childhood and puberty and the rate of bone loss in late adulthood. Bone mass has a very strong genetic component: twin and family studies reveal that 50 to 90% of the variance in BMD is caused by genetic factors (Pocock et al. 1987; Slemenda et al. 1991; Lutz & Tesar, 1990; Gueguen et al. 1995; Hunter et al. 2001). Whole genome-wide linkage studies (Wilson et al. 2003; Devoto et al. 1998; Kammerer et al. 2003; Karasik et al. 2004; Ralston et al. 2005) have been launched to search for genes underlying BMD variation. To date, a number of chromosomal regions have been identified that show linkage with BMD, although there are inconsistencies between studies. Interestingly, studies that adopt a candidate gene association approach have demonstrated that some of the associated polymorphic sites identified in Caucasian populations are not significantly associated in Oriental populations: the COL1A1 Sp1 (Lambrinoudaki & Kung, 2001) and the VDR BsmI sites (Kung et al. 1998) lack heterozygosity in Asian populations. It is uncertain whether these genetic differences account for the variation in BMD and fracture incidences between Asians and Caucasians. Furthermore, gene-gene and gene-environment interactions may differ among populations, especially for a polygenic disorder such as osteoporosis. We undertook a study to quantify the heritability of BMD and bone size in a homogenous population, and to evaluate the impact of environmental factors on the heritability estimation. It was hoped that this would help understand the interplay of genetic and environmental determinants of osteoporosis.

Methods and Materials

Study Population

Ethnic Southern Chinese subjects were recruited from the community when they attended public road shows, health fairs or health talks on osteoporosis held in different districts in Hong Kong from 1998 to 2003. Family members were also invited to participate. All subjects completed a detailed questionnaire that included demographic information, medical and reproductive histories, lifestyle habits, smoking and drinking habits, and presence of a personal or family history of fractures. Calcium intake and phytoestrogen intake, in particular the isoflavones, were assessed by a semi-quantitative questionnaire (Mei et al. 2001). The standard portion size was specified in commonly used units or portions as judged by a registered dietician. Weight bearing physical activity was assessed as the average number of hours per day spent on all weight-bearing activity, including walking. Cigarette smoking was defined by the duration of smoking and the number of pack-years. Alcohol consumption was assessed by recording the duration of drinking and number of units consumed per week. Subjects were excluded if they had a history of chronic medical illness, premature menopause below age 40 years, malabsorption, major gastrointestinal operations, metabolic bone disease, endocrine disorders including hyperthyroidism, or medications that may affect bone and calcium metabolism. Individuals prescribed bisphosphonates, selective oestrogen receptor modulators, calcitonin and active vitamin D3 metabolites were also excluded.

A total of 3,514 subjects from 1,019 families were recruited. Of these 194 subjects were excluded, leaving 3,320 subjects (534 males and 2,786 females) for analysis. All subjects gave informed consent and the study was approved by the Ethics Committee of the University of Hong Kong and conducted according to the Helsinki Declaration.

BMD and Bone Size Measurement

Bone mineral density (BMD) in g/cm2 and bone size in cm2 were measured at the lumbar (L1-L4) spine and the hip region (femoral neck, total hip and trochanter) using a Hologic QDR 4500 (Waltham, Massachusetts, USA) dual energy X-ray absorptiometer (DXA). The in vivo precision of the machine for BMD at the lumbar spine, femoral neck and total hip in postmenopausal women was 1.2%, 1.5% and 1.5%, respectively.

Heritability and Statistical Analysis

Heritability of raw BMD and bone size at each site was estimated using a general pedigree variance component method that was implemented in SOLAR (Sequential Oligogenic Linkage Analysis Routines) (Almasy & Blangero, 1998). SOLAR uses maximum likelihood methods to estimate variance components for polygenic genetic effects and random individual environmental effects. Heritability (h2) is defined in the narrow sense, as the proportion of total phenotypic variance due to the additive effects of multiple genes, after the effects of covariates have been removed.

The analysis was stratified by gender, and all models included the following covariates: age, age2, height, weight, intake of isoflavones and calcium, weight-bearing physical activity, smoking, drinking, and for females age at menarche and menopause. All covariates were significantly correlated with BMD or bone size at one or more skeletal sites independently in a preliminary gender-specific analysis. Five covariate adjustment strategies were performed sequentially: the first polygenic model analysed the raw BMD or bone size measurements. The second polygenic model simultaneously adjusted for age and age2. The third additionally adjusted for height and weight. The fourth further adjusted for dietary intake of calcium and isoflavone, weight bearing physical activity, smoking and drinking. Weight bearing physical activity per day was divided into two groups, less than or more than 1 hour per day. The final model was further adjusted for age at menarche and age at menopause, and was only applied to the post-menopausal female group. Only significant covariates were included in the polygenic models, and the significance of each covariate was evaluated by testing whether the variation attributable to the specific covariate was significantly greater than zero in the polygenic model. Age and age2 were always included in the models, as age is known to be a major source of BMD or bone size variation (Cummings et al. 1995). A value of p < 0.05 (two-tailed) was considered statistically significant.

A chi-square test of heterogeneity of genetic variance (Neal et al. 1992) between sexes was carried out for the final model with adjustment for all covariates. The test is defined as:


with 1 degree of freedom, where

FVg= absolute genetic variance of females

MVg= absolute genetic variance of males

se(FVg) = standard error of female absolute genetic variance

se(MVg) = standard error of male absolute genetic variance

The trait that had a standard deviation < 0.5 was multiplied by a scaling factor recommend by SOLAR. The scaled traits were then used in analysis. The effect size and standard errors of each covariate were obtained by rescaling them by the suggested factor applied to the associated trait. The h2 estimates were unaffected by the scaling factors.

The relative contribution, i.e. the partial correlation (r2), of the covariates with adjustment of all covariates was determined in the final model. Since age and age2 were two highly correlated covariates, r2 of these two variables were considered together. The r2 (Mardia et al. 1989) was defined as:



inline image= beta-coefficient of covariates of the regression model

inline image= covariance matrix of covariates

δ2Y= variance of the regression model


Study Population

The characteristics of the study sample are summarised in Table 1. There were 534 males and 2,786 females, 33% of women aged over 65 years were widowed and 78% of families contained only female members. Mean age at menarche was 13.5 ± 2 (mean ± SD) years. Overall, 1,307 (47%) women were postmenopausal (mean age of menopause 49.4 ± 3.8 years). 48.6% of the females and 54.9% of the males spent more than 1 hour on weight-bearing physical activity each day. Mean dietary calcium and isoflavone intake per day was approximately 628.5 mg and 24.2 mg for females, and 663.5 mg and 24.3 mg for males, respectively. The percentages of smokers and drinkers were 28.2% and 29.5%, respectively in males and 4.4% and 6.9%, respectively in females. In general, males had a significantly higher BMD and bigger bone size than females (p < 0.001, ANOVA test).

Table 1.  Characteristics (mean ± SD) of study population of 3,320 individuals in 1,019 families
CharacteristicsFemale (n = 2786)Male (n = 534)
  1. a1,307 women were post-menopausal.

Age (years)48.1 ± 15.751.0 ± 16.5
Height (meter)1.55 ± 0.071.67 ± 0.07
Weight (kg)54.2 ± 9.066.2 ± 10.2
Age at menarche (years)13.5 ± 2.0
Age at menopause (years)49.4 ± 3.8a
Daily calcium intake (mg/day)663.2 ± 259.6628.5 ± 251.1
Daily isoflavone intake (mg/day)24.3 ± 22.824.2 ± 21.3
Weight bearing physical activity > 1 hour/day (%)48.654.9
Smokers (%)129 (4.4)157 (28.2)
 Year of smoking19.6 (15.8)24.9 (13.9)
 No. of pack per year 
  (% > 365 packs)21.958.6
Drinkers (%)202 (6.9)164 (29.5)
 Years of drinking13.7 (12.4)18.8 (12.3)
 % > 3 units per week51 (45)40 (42)
BMD (g/cm2)
 Spine0.892 ± 0.1600.937 ± 0.149
 Femoral neck0.685 ± 0.1210.730 ± 0.122
 Trochanter0.592 ± 0.1130.657 ± 0.112
 Total hip0.778 ± 0.1310.869 ± 0.133
Bone size (cm2)
 Spine53.83 ± 5.7263.41 ± 5.90
 Femoral neck4.73 ± 0.345.40 ± 0.32
 Trochanter9.14 ±1.2411.72 ± 1.49
 Total hip29.29 ± 2.8938.22 ± 3.64


The h2 estimates for each polygenic model are presented in Table 2. The estimates were statistically significant at all skeletal sites (p < 0.005), and differed among skeletal sites and gender. The h2 for unadjusted BMD ranged from 0.33 to 0.39 for females and 0.46 to 0.83 for males. The h2 estimates of unadjusted bone size were higher than the BMD, ranging from 0.45 to 0.78 for females and 0.46 to 0.83 for males.

Table 2.  Heritability estimates for bone mineral density (BMD) and bone size
Skeletal sitesBMDBone size
  1. All h2 values are significantly associated with BMD and bone size (p < 0.005).

  2. Results in brackets are standard errors.

1st model: unadjusted for covariates
 Spine0.36 (0.04)0.83 (0.13)0.45 (0.04)0.46 (0.13)
 Femoral neck0.33 (0.04)0.46 (0.13)0.58 (0.04)0.62 (0.15)
 Trochanter0.36 (0.04)0.70 (0.13)0.57 (0.04)0.69 (0.15)
 Total hip0.39 (0.04)0.69 (0.13)0.78 (0.04)0.83 (0.12)
2nd model: adjusted for age, age2
 Spine0.73 (0.04)0.84 (0.13)0.61 (0.04)0.46 (0.13)
 Femoral neck0.64 (0.04)0.77 (0.12)0.59 (0.04)0.63 (0.15)
 Trochanter0.69 (0.04)0.81 (0.12)0.59 (0.04)0.77 (0.14)
 Total hip0.70 (0.04)0.82 (0.12)0.80 (0.04)0.83 (0.12)
3rd model: Further adjusted for weight, height
 Spine0.71 (0.04)0.75 (0.13)0.44 (0.04)0.32 (0.13)
 Femoral neck0.63 (0.04)0.74 (0.12)0.48 (0.05)0.47 (0.17)
 Trochanter0.67 (0.04)0.78 (0.12)0.48 (0.05)0.79 (0.14)
 Total hip0.69 (0.04)0.79 (0.12)0.64 (0.04)0.86 (0.13)

The second polygenic model adjusted for age and age2. Age substantially increased the h2 estimates. The female h2 increased >75%, whereas the male h2 increased by 16 to 68% at the hip. For bone size, adjustment for age only led to a moderate change in h2 for female spine size (from 0.45 to 0.61), with no significant changes in other sites. The age effect was significantly greater in women than in men. The proportion of total phenotypic variance in BMD explained by age and age2 was around 0.3 in women but less than 0.16 in men, and 0.13 and 0.01 in female and male bone size, respectively. In comparison, the effect of age on bone size variation was much smaller, with the proportion of total phenotypic variance in bone size explained by age and age2 around 0.13 in women but less than 0.01 in men.

The third polygenic model further adjusted for weight and height. Adjustment for weight and height slightly altered the h2 for hip BMD, but decreased the h2 for spine BMD, particularly in males (from 0.84 to 0.75, Table 2). Adjustment for weight and height produced substantial changes in the h2 values for bone size at the spine and hip. When environmental and lifestyle factors including intake of calcium and isoflavone, physical activity, smoking and alcohol consumption were added to the model, the main effect was on the trochanter bone size of males, with h2 reduced from 0.79 to 0.67. Further adjustment for age at menarche and age at menopause was performed in post-menopausal females. The h2 for trochanter BMD, after adjustment, increased from 0.67 to 0.78. There was a slight increase of the value for spine BMD by 0.02 and by 0.03 for femoral neck BMD, but no substantial changes were seen for h2 of bone size.

Tables 3 and 4 show the contribution of age, weight and height, and environmental factors, towards the variance in BMD and bone size. Each factor had a variable effect on BMD and bone size, an effect that varied across different skeletal sites. Age, weight and height were the most important contributing factors to the variance in BMD and bone size. The r2 for age and age2 for BMD in females and males was 0.084 to 0.249 and 0.009 to 0.142, respectively with values for bone area of 0.021 to 0.151 and 0.044 to 0.087 for females and males, respectively. The r2 for height for BMD in females was 0.155 to 0.353 and for males was 0.220 to 0.365, while r2 for weight for bone area in females was 0.131 to 0.164 and for males was up to 0.238. Among females, all covariates together accounted for 31.8 to 46.1% of the total variance in BMD and 17.5 to 40.2% of the variance in bone size. In males, the contribution of these factors was less: they accounted for 12.6 to 31.3% of the variance in BMD and 18.8 to 28.0% of the variance in bone size. Some of the variables were correlated with each other, and the variances they accounted for overlap to some extent. Hence, the summation of the r2 of the trait was not exactly the same as the proportion of the total phenotypic variance explained by the covariates.

Table 3.  Determination of bone mineral density by anthropometric and environmental factors*
Spine β (SE)r2Femoral neck β (SE)r2Trochanter β (SE)r2Total Hip β (SE)r2
Age−0.021 (0.004)0.0840.003 (0.003)0.1960.003 (0.003)0.1720.006 (0.00055)0.249
Age2−0.00014 (0.000034) 0.000015 (0.000021) 0.774 (0.081) 0.000093 (0.0000055) 
Height0.172 (0.062)0.0060.087(0.039)0.0030.081 (0.038)0.003
Weight0.006 (0.00038)0.1450.038 (0.00023)0.1440.0035 (0.00023)0.1310.005 (0.00017)0.164
Physical activity0.008 (0.004)0.002
Menopausal age0.002 (0.00082)0.0080.001 (0.00051)0.0040.001 (0.0005)0.003
Menarche age−0.006 (0.0016)0.004−0.003 (0.001)0.004−0.003 (0.00096)0.006−0.004 (0.00082)0.005
Total** 0.318 0.436 0.409 0.461
  1. *Values are expressed as beta-coefficients (Standard Error) with p ≤ 0.05; values expressed in italic are insignificant, with p > 0.05; r2 is the partial correlation; **Total is the proportion of total phenotypic variance explained by the covariates.

Age−0.004 (0.002)0.009−0.006 (0.001)0.1420.004 (0.001)0.037−0.004 (0.002)0.043
Age2−0.000041 (0.000021) −0.00004 (0.000014) 0.000024 (0.000014) 0.000027 (0.000026) 
Height0.212 (0.083)0.012
Weight0.005 (0.00062)0.1270.004 (0.0004)0.0000140.004 (0.00038)0.1320.005 (0.00048)0.238
Years of drinking0.00049 (0.000049)0.004
Years of smoking−0.00074 (0.00055)0.002−0.00068 (0.00033)0.006
Physical activity0.019 (0.008)0.125
Total** 0.126 0.313 0.197 0.261
Table 4.  Determination of bone size by environmental factors*
Spine β (SE)r2Femoral neck β (SE)r2Trochanter β (SE)r2Total Hip β (SE)r2
Age−0.204 (0.031)0.021−0.006 (0.002)0.0200.0061 (0.0076)0.151−0.081 (0.016)0.105
Age2−0.002 (0.00031) −0.000078 (0.000017) 0.00036 (0.000076) −0.001 (0.016) 
Height53.34 (1.60)0.3531.57 (0.090)0.1559.34 (0.393)0.23125.81 (0.918)0.332
Weight0.003 (0.001)0.013 0.031 (0.006)0.010
Calcium intake0.00048 (0.00029)0.0010.00022 (0.00016)0.001
Isoflavone intake0.003 (0.002)0.002
Physical activity0.139 (0.041)0.003
Menarche age0.032 (0.010)0.0040.057 (0.026)0.002
Total** 0.402 0.175 0.205 0.331
  1. *Values are expressed as beta-coefficients (Standard Error) with p ≤ 0.05; values expressed in italic are insignificant with p > 0.05; r2 is the partial correlation; **Total is the proportion of total phenotypic variance explained by the covariate.

Age0.146 (0.078)0.044−0.008 (0.003)0.0690.0077 (−0.00012)0.0870.080 (0.044)0.074
Age20.00077 (0.00078) 0.00016 (0.00013) 0.00019 (0.00020) 0.00019 (0.00045) 
Height52.75 (3.83)0.3591.73 (0.165)0.2339.93 (1.02)0.22031.16 (2.353)0.365
Alcohol consumption−0.412 (0.128)0.019−0.410 (0.237)0.003
Years of drinking0.031 (0.007)0.073
Total** 0.280 0.188 0.196 0.258

The absolute genetic variance between sexes was compared to determine whether there was a gender difference in the genetic variance of BMD and bone size (Table 5). A significant gender difference was observed in the absolute genetic variance in BMD at the total hip (female 0.006, male 0.010) but not the spine, and in bone size at the trochanter (female 0.59, male 1.19) and total hip (female 3.58, male 8.13).

Table 5.  Components of variance for bone mineral density (BMD) and bone size after adjusting for all covariates
  1. Values in bracket are standard errors. BMD data of columns P, G, E, C are in g/cm2, bone size data of columns P, G, E, C are in cm2. h2 is defined as the proportion of additive genetic effect over the phenotypic variation after the adjustment of covariates.

  2. PC - Proportion of the total phenotypic variances explained by the covariates

  3. P - Absolute phenotypic variance

  4. C - Absolute variance explained by covariates = PC × P

  5. G - Absolute genetic variance =h2× (1 − PC) × P

  6. E - Absolute unmeasured environmental variance =(1 −h2) × (1 − PC) × P

  7. *p < 0.05 of the chi-square test of heterogeneity of genetic variance in males and females.

 Spine0.320.73 (0.09)0.0260.0080.013 (0.002)0.0050.130.73 (0.13)0.0220.0030.014 (0.003)0.005
 Femoral neck0.440.65 (0.09)0.0150.0060.005 (0.001)0.0030.310.74 (0.12)0.0150.0050.008 (0.001)0.003
 Trochanter0.410.78 (0.09)0.0130.0050.006 (0.001)0.0020.200.80 (0.12)0.0130.0030.008 (0.001)0.002
 Total hip*0.460.68 (0.04)0.0170.0080.006 (0.000)0.0030.260.80 (0.12)0.0180.0050.010 (0.002)0.003
Bone Size
 Spine0.400.44 (0.05)32.7213.098.6 (1.0)10.990.280.33 (0.13)34.819.7478.3 (3.3)16.79
 Femoral neck0.180.48 (0.05)0.1160.0210.0 (0.0)0.0490.190.47 (0.17)0.1020.0190.0 (0.0)0.044
 Trochanter*0.210.49 (0.05)1.5380.3230.6 (0.1)0.6190.200.67 (0.16)2.2200.4441.2 (0.3)0.586
 Total hip*0.330.64 (0.04)8.3522.7563.6 (0.2)2.0150.260.83 (0.13)13.253.4458.1 (1.3)1.667


Owing to the huge burden of osteoporosis on healthcare resources, there is significant interest in unravelling the genetic control of bone mass and searching for new therapeutic agents. A number of studies have addressed the heritability of BMD, the best surrogate marker currently available for osteoporotic fracture, and applied the information to the search for genetic loci for bone mass determination. Whether BMD or bone size is a polygenic trait or controlled by a major gene with pleiotropic effects is unclear. The first principle component has been applied to analyse the heritability of bone mass and to identify QTLs (Karasik et al. 2003). Nonetheless a limitation of this approach is possible loss of information due to the heterogeneity of gene effects at different skeletal sites. Our present study utilized variance component analysis and the SOLAR programme that is based on the theory of a pure polygenic model of a quantitative trait. The general analysis model assumed that BMD is determined by the mean of the trait, a polygenic background representing genetic effects, a group of non-genetic covariates, and a residual representing the effects of unmeasured and unknown environmental factors that were not analysed in the model. A candidate genes approach has revealed that more than 60 genes play a role in the determination of BMD, each contributing to a small percentage of the total variance (Huang et al. 2003). Furthermore, some of these genes interact with environmental factors to influence BMD, e.g. vitamin D receptor polymorphisms and calcium intake (Krall et al. 1995), oestrogen receptor α and exercise (Suuriniemi et al. 2004).

Compared with previous heritability studies on BMD and bone size, the sample employed in our study represents one of the largest. Most studies that attempted to determine heritability of BMD adjusted only for anthropometric parameters such as age and weight, and failed to take into account other environmental and lifestyle risk factors that have been shown to affect BMD (Cummings et al. 1995; Mitchell et al. 2003). In the current study, BMD and bone size were adjusted for possible covariates that are known to affect BMD and bone size. Our findings suggest that age, anthropometric parameters and non-genetic covariates have significant effects on the heritability of BMD and bone size. These data confirmed our previous reports of the importance of non-genetic factors as the determinants of BMD in pre- and postmenopausal women, as well as men of southern Chinese ethnicity (Koh et al. 2001; Cheung et al. 2004; Ho & Kung, 2004). The roles of calcium and phytoestrogen intake, exercise, smoking and alcohol consumption, age at menarche and menopause on bone mass determination have been extensively studied. The effects of such lifestyle and environmental risk factors are largely dependent on the stage of the subjects' lifespan and the state of bone growth, development and remodelling. Overall, age, weight, height and environmental factors contributed to 32 to 46% of the total variance in BMD among females, and to 13 to 31% of that among males. The traits least affected by non-genetic factors were male spine BMD (13%) and male trochanter BMD (19.7%). The great variation in the contribution of environmental factors to different skeletal sites and bone phenotype highlighted the importance of including these non-genetic factors in analysis.

The present study showed that the effects of the measured covariates explained a much greater proportion of the total phenotypic variance of BMD in females than males. Prior to adjustment of any confounding factors, the h2 value for BMD was much larger at each skeletal site in men than in women, but the difference between h2 values between sexes for bone size was much smaller. However, after adjusting for age and age2, the magnitude of difference substantially decreased. As the proportion of total phenotypic variance explained by age and age2 in BMD and bone size was much greater in women than in men, this raised the h2 estimates in women after the traits were adjusted for age and age2. Additional adjustment, especially for weight and height, narrowed the differences between the sexes in h2 estimates. Reducing the proportion of residual phenotypic variation attributed to random environmental factors may increase the genetic signal-to-noise ratio, and hence increase the residual heritability estimates. Interestingly, the h2 estimates of BMD and bone size decreased after further adjustment of height and weight (model 3). This may have been because height and weight, which are partially genetically mediated, may be determined by the same set of candidate genes as BMD or bone size. Thus, their genetic effects may be removed when BMD and bone size are adjusted for these variables, hence lowering the h2 estimates.

A chi-square test of heterogeneity of genetic variances between the two sexes was performed to test whether the genetic effect was gender-specific. Significant genetic differences between the sexes were found in total hip BMD, and trochanter and total hip-bone size. Thus, our data suggest a site-specific gender difference in the inheritance of BMD and bone size. Whether a gender difference exists in the heritability of BMD has been debated (Orwoll et al. 2001; Naganathan et al. 2002; Karasik et al. 2003). Several researchers (Kammerer et al. 2003; Karasik et al. 2003; Ralston et al. 2005) have observed a significant sex effect on the genetic influences on BMD variation, nonetheless Brown et al. (2004) observed no sex difference. A sex difference has been observed in the heritability of various bone phenotypes in mouse models (Orwoll et al. 2001; Klein et al. 2002). Ralston et al. (2005) reported gender-specific, site-specific and age-specific QTL for BMD with subgroup analysis, while no regions of suggestive or significant linkage were identified when all subjects were analysed together.

Whether the sex difference observed in human studies was related to sample bias is unclear. In this study, like many others, there were far fewer male than female subjects. This resulted in greater variability, as illustrated by a larger standard error of the h2 values. This may have resulted in an underestimation of the difference in the heritability estimates between the sexes. The age of the proband may also affect the h2 estimation. Studies with young probands give much higher h2 values, with the highest being reported in the studies of young children (Ferrari et al. 1998; Liu et al. 2004a). These findings are not surprising as studies involving young children or young adults will likely detect genes that affect peak bone mass, whereas those with probands of an older age group will likely include genetic factors less likely to affect bone loss. The BMD of older individuals is also likely to be influenced more by environmental and lifestyle factors: this will reduce the residual heritability estimates. The presence of osteophytes and aortic calcifications in older subjects might also affect the areal BMD values and increase the variability of lumbar spine BMD, hence altering the relation between parental-children pairs and reducing the heritability estimates.

The present study was limited by the absence of X-rays for any subjects to exclude morphometric vertebral fractures that may affect the spinal bone area and BMD. To reduce the variability, a vertebral body with significant collapse, or vertebral height that deviated from the adjacent vertebrae on DXA scan, was excluded from analysis in the present study. Nonetheless, in keeping with other studies, the h2 values for lumbar spine were lower than for the hip region. It is uncertain whether the lower h2 for lumbar spine is really due to greater unmeasured environmental confounding effects or to problems with BMD determination at the spine.

Few studies have determined heritability estimates for BMD or bone size in the Chinese population. Liu et al. (2004a) reported h2 estimates of 0.62 for spine bone size and 0.59 for total hip bone size. Other studies reported h2 estimates of 0.60–0.63 for spine area and 0.45–0.69 for total hip area (Jian et al. 2004), and 0.81 for spine BMD and 0.90 for total hip BMD (Liu et al. 2004b). The h2 estimates for trochanter and femoral neck BMD and bone size have not been reported previously. In comparison, the h2 estimates of bone size and spine BMD reported in the current study were somewhat lower. The differences may be due to characteristics of the population studied, in particular environmental and lifestyle factors.

In conclusion, this study shows that bone mass and bone size in the southern Chinese population are highly heritable, with heritability estimates being higher in males than females. The covariate age was more closely related to BMD in women than men, as evidenced by the substantial increase in heritability after accounting for age. Further adjustment for height, weight and environmental risk factors eliminates the differences in genetic variance between the sexes at the spine, femoral neck and hip trochanter, but not at the total hip region. In view of these confounding effects, comparison of h2 across different studies should be interpreted with care. Also, future linkage and association analyses should take gender differences into account. Whether biological difference exists in the residual component between genders needs further validation by molecular studies.


The authors thank the Staff of the Osteoporosis Centre, The University of Hong Kong, Queen Mary Hospital for assistance in this project. This study is supported by The Osteoporosis and Endocrine Research Fund and The CRCG Grant of The University of Hong Kong, and The Hong Kong University Foundation-Bone Health Fund and The Matching Grant.