The present study was part of the Dubbo Osteoporosis Epidemiology Study (DOES), of which the details of study design and protocols have been described previously 282930. Briefly, DOES is a community-based, epidemiological study, where participants were recruited from the Dubbo city, a semi-urban city 400 km northwest of Sydney, Australia, with a population of approximately 32 000 people, 98.6% Caucasian, of whom 1 581 men and 2 095 women were aged 60 years or above in 1989. Dubbo's relative isolation in terms of medical care allows virtually complete ascertainment of all fractures. From 1993, blood samples were collected from individuals who volunteer to donate at initial visits or follow-up visits. Our present study reports data on a random sample of 446 women and 265 men aged 60 years or above in 1989, who had consented to donate blood for DNA analysis.

After obtaining written informed consent, participants were interviewed by a trained research nurse at initial and subsequent visits at approximately 2-year intervals. A structured questionnaire was used to collect data including age, anthropometric variables, life style and clinical data 31. Bone mineral density (g/cm²) was measured at the lumbar spine and femoral neck by dual-energy X-ray absorptiometry (DXA) using a Lunar DPX-L densitometer (GE Lunar, WI. USA). The measurement was performed at the same visit or within one month of the blood sample collection in 99% of cases. In the cases in which the blood was taken at a time other than the scheduled visit, the BMD used was that closest to the time of blood collection.

The incidence of fracture was ascertained during the study period (1989–2004) by radiologist's report and confirmed by personal interview. Only low trauma fractures were included (i.e. a fall from the standing height or less). Fractures due to motor vehicle accidents or pathological conditions were excluded. Fractures were categorised into 4 groups: hip, symptomatic vertebral, Colles' and other fractures.

Genomic DNA was extracted from peripheral blood leukocytes. The ADRB3 polymorphism was determined by polymerase chain reaction (PCR) performed with 20 ng of genomic DNA with upstream primer 5' CCA GTG GGC TGC CGA GGG 3' and downstream primer 5' GCC AGT GGC GCC CAA CGG 3' 32. The amplification products were digested with the restriction enzyme BstOI (Promega, Madison, WI, USA) and genotyped as Trp homozygotes (Trp/Trp), heterozygotes (Trp/Arg), or Arg homozygotes (Arg/Arg). This Trp64Arg polymorphism has been commonly used in genetic epidemiological studies.

Alleles were counted and the genotype frequencies of the polymorphism was checked for Hardy-Weinberg equilibrium (HWE) by the likelihood Chi-square statistic 33. The baseline characteristics of the subjects in the fracture group and the non-fracture group were compared using descriptive statistical tests with appropriate techniques (e.g. Student's t-test for normally distributed variables, such as BMI and BMD) and Chi-square (for categorical data), depends on the distribution of the variables. The association between each polymorphism and BMD or BMI was analyzed by the analysis of covariance (ANCOVA) model taking into account the effects of potential covariates such as age, weight and life-style factors.

In order to quantify the risk of sustaining a fracture in relation to ADRB3 genotypes, subjects were grouped into those with Arg allele and those without Arg allele. The association was analyzed by the logistic regression model taking into account the effects of potential confounders.

In order to determine the association between BMI and ADRB3 polymorphisms, a Bayesian approach was used 3435. In the Bayesian analysis, the currently observed data are combined with known data to derive a "posterior distribution" with credible interval. In the present study, a prior mean difference (M₀) in BMI between subjects with and without Arg allele was obtained from a meta-analysis 26 with a 95% confidence interval (95%CI) being (L₀, U₀), from which the standard deviation (SD₀) for the prior distribution was calculated as follow: SD₀ = (U₀- L₀)/3.92, assuming the difference follows the normal distribution. In order to construct a posterior distribution with mean (M_p) and standard deviation (SD_p), the prior distribution was combined with the present study's data, which has a mean of M_d and standard deviation of SD_d. The posterior mean and SD are: M_p =(SD_p)² [(M₀/SD₀²)+(M_d/SD_d ²)] and SD_p = 1/(1/SD0)2+(1/SDd)2 MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacH8akY=wiFfYdH8Gipec8Eeeu0xXdbba9frFj0=OqFfea0dXdd9vqai=hGuQ8kuc9pgc9s8qqaq=dirpe0xb9q8qiLsFr0=vr0=vr0dc8meaabaqaciaacaGaaeqabaqabeGadaaakeaadaGcaaqaaiabcIcaOiabigdaXiabc+caViabdofatjabdseaenaaBaaaleaacqaIWaamaeqaaOGaeiykaKYaaWbaaSqabeaacqaIYaGmaaGccqGHRaWkcqGGOaakcqaIXaqmcqGGVaWlcqWGtbWucqWGebardaWgaaWcbaGaemizaqgabeaakiabcMcaPmaaCaaaleqabaGaeGOmaidaaaqabaaaaa@3E21@. Thus, the posterior mean is a weighted average of two estimates (prior mean and the present study's mean) with the weight being the inverse variances. The analysis was conducted via the SAS system 36.

In total, 446 women and 265 men who had been followed-up for the median of 13.2 years (interquartile range, IQR: 10.4–14.1) and 12.9 years (IQR: 9.0–14.0) for women and men, respectively.

The frequencies of the Trp/Trp, Trp/Arg and Arg/Arg genotypes were 385 (86.3%), 58 (13.0%) and 3 (0.7%) respectively for women, and 224 (84.5), 40 (15.1%) and 1 (0.4%) respectively for men. In both sexes, the allelic frequencies of Trp allele and Arg allele were 0.925 and 0.075 respectively. Because the frequency of Arg allele was low in the population, individuals with Trp/Arg and Arg/Arg genotypes were considered as one group. The overall genotype distribution in the study subjects was in Hardy-Weinberg equilibrium (χ² = 0.007, p = 0.98).

Clinical characteristics of subjects were stratified by the presence or absence of Arg allele (Table 1). The presence of Arg allele in women is associated with a 0.25 SD increase in body weight, but this difference was not statistically significant (p = 0.08). In both women and men, there were no statistically significant differences between Arg allele carriers and non-carriers in terms of anthropometric characteristics and lifestyle factors.

In the present study, there was no significant association between ADRB3 genotypes and BMI. The overall mean difference in BMI for Arg carriers versus non-carriers was 0.54 kg/m² (95% CI: -0.35, 1.44; P = 0.23), or 0.13 SD. In a previous meta-analysis 26, Arg carriers had higher BMI than non-Arg carriers by 0.30 (95% CI, 0.13–0.47) kg/m² as indicated in the prior distribution (Figure 1). After updating the previous data with the present study's data, the posterior distribution showed that BMI in the Arg allele carriers are, on average, 0.31 kg/m² (95% CI, 0.14–0.48) higher than non-Arg carriers.

In multiple regression analysis of BMD measurements at both lumbar spine and femoral neck (Table 2), there were no significant associations between ADRB3 and BMD at both LS and FN sites. However, in both women and men, age and smoking history were negatively associated with femoral neck BMD, and BMI was positively associated with lumbar spine and femoral neck BMD. Although increased physical activity was not associated with BMD measured at femoral neck, it was positively associated with lumbar spine BMD. Collectively these factors (age, BMI, smoking history, calcium intake, physical activity and ADRB3 genotypes) explained about 30% and 12% of variances in femoral neck BMD in women and men respectively.

During the follow up period, 173 women and 49 men had sustained at least one osteoporotic fracture. In women, the most common fracture site was distal radius (Colles' fracture) with an incidence rate of 24% of all the cases, followed by vertebral (23%) and hip (14%). In men, the most common fracture site was vertebrae (41%), followed by hip (16%) and ribs (14%). For both women and men, fracture individuals were older, weighed less, and had lower BMD at both femoral neck and lumbar spine compared to their non-fracture counterparts (Table 3). However, there was no significant difference between the fracture group and non-fracture group in terms of calcium intake, physical activity and smoking history.

The distribution of ADRB3 genotypes in fracture and non-fracture subjects is shown in Table 4. Women with the Arg/Arg and Trp/Arg genotypes had lower incident rate of any fracture than those with the Trp/Trp genotype, but the difference was not statistically significant (32.8% vs. 39.7%, p = 0.30). This non-significant association was observed for all specific fracture sites. However, in men, individuals with the Arg/Arg and Trp/Arg genotypes had higher incident rate of any fracture than those with the Trp/Trp genotype (22.0% vs. 17.9%); the difference was also not statistically significant (p = 0.53). After adjusting for age, BMI and/or BMD, the associations between the ADRB3 polymorphism and fracture risk in women and men were not statistically significant (Table 5). There were no interaction effects between ADRB3 genotypes and age on fracture risk (data not shown).