Bone Mineral Density Reference Values and Prevalence of Osteoporosis in Singaporean Adults – The Yishun Study

Background: While there have been studies in Singapore on the prevalence and economic burden of osteoporotic hip fracture, there is a severe lack of reference data on bone mineral density and prevalence of osteoporosis. The purpose of this study is to establish the reference values for BMD and compare prevalence of osteoporotic conditions using other available reference values so as to better understand the status of bone health in Singaporean adults Methods: We carried out a population-based cross-sectional study using dual-energy x-ray absorptiometry (Hologic Discovery Wi) to measure the bone mineral density of Singaporean adults aged ³ 21 years. 314 women and 223 men were recruited from the large north-eastern residential town of Yishun. We computed T- scores (denoted by T SG ) for each individual in the study. Similar diagnoses were also done based on T-scores provided by the densitometer (T DXA ), NHANES database (T NHANES ), and China (T CHN ), and the differences in prevalence compared. We then compared the concordance between T SG and T DXA in the classication of osteoporosis. Osteoporosis was dened according to criteria by the World Health Organization (WHO). Results: Peak lumbar spine BMD was 1.090±0.168g.cm -2 in women, and 1.041±0.098g.cm -2 for men. Peak whole body BMD was 1.193±0.93g.cm -2 in women at, and 1.224±0.112g.cm -2 for men. Prevalence of osteoporosis based on lumbar spine was 9.3% in women, and 0.7% in men. The percentage difference in prevalence range from 60.5 - 163.6%, when using reference values from T DXA , T NHANES , and T CHN . Among the 28 women who were diagnosed as osteoporotic by T DXA , 16.5% only had osteopenia according to T SG . The kappa statistics was 0.81 for women and 0.85 for men. Conclusion: Our study shows that T-scores provided by DXA manufacturer over-diagnosed osteoporosis in Singaporeans, and the prevalence of osteoporotic conditions is not accurately represented. This over-diagnosis may result in unnecessary treatment in some individuals. aortic calcication, vertebral compression fracture, scoliosis, and osteoarthritis(42). The difference in rate of spinal BMD diminution was found to be 1% and 10% per decade in men and women, respectively(43). However, other studies reported that lumbar spine BMD increases at a rate of 1.5-3.5% per decade in men over 60 years old(44-46). The compound effect of older men physiology and the cohort effect likely explained the lack of signicant change in BMD across age in our study.


Introduction
Osteoporosis is characterised by a systemic loss of bone mineral and micro-architectural deterioration of bone tissue, resulting in increased risk of fracture (1). Osteoporotic fractures result in increased morbidity, disability, and mortality risk, reducing quality of life (2). Common fracture sites include the spine, hip, distal forearm, and proximal humerus. The average risk of osteoporotic fracture for people at 50 years of age have been estimated to be 40% for women, and 13% for men (3). Osteoporosis and associated fractures have become increasingly common in aging populations and is a public health issue (1).
Measurement of bone mineral density (BMD) using dual-energy X-ray absorptiometry (DXA) is the most widely used indicator of bone health and for detection of osteoporosis (4). Most studies to accumulate BMD reference data are from America or Europe with relatively few Asian studies. Manufacturer-supplied BMD reference values from Western populations may not be fully representative of Asian populations and can therefore lead to estimates of the prevalence of osteoporosis that differ greatly from those calculated based on a local reference population (5).
A multi-ethnic country in southeast Asia, Singapore is aging rapidly with proportion of resident population aged 65 years and over having rose from 9.0% in 2010 to 14.4% in 2019(6), and is expected to reach 25% by 2030 (7). Osteoporotic hip fracture incidence rates have risen 1.5-fold for men and 5-fold for women since the 1960s (8). Between 2000 to 2017, there is a 3.5% increase in hip-fracture related hospitalization, translating to an average of 72 additional hip fractures related hospitalisation per 100,000 per year (2). In 2017, the estimated incidence of osteoporotic fracture was 15,267 case, (25.5% hip fracture, 29.1% vertebral fracture, and 45.4% other fractures), and it is estimated to increase by 58% by 2035 (9). The total economic burden associated with osteoporosis and related fractures was estimated to be S$183.5 million in 2017, and forecast to increase to S$289.6 million by 2035 (9). Mortality rate one year post osteoporotic hip fracture was reported to be 20-27%, with most survivors became semi or fully dependent (20%), or experienced reduced mobility status (39-42%) (8,10). Only 26% were cared for by chronic health care facilities suggesting that the main social and nancial burden was borne by the family caregivers (8,10,11).
While there have been studies in Singapore on the prevalence and economic burden of osteoporotic hip fracture (2,8,10), there is a severe lack of reference data on BMD and prevalence of osteoporosis. A study published in 2002, found Singapore men had 10% and 5% lower bone mineral density in the lumbar spine (LS) and femoral neck, respectively, when compared to mean BMD peak of Caucasian reference database (8). A previous study on women aged 20-59 years reported no difference in peak BMD in the LS and femoral neck between the three main ethnic groups of Singapore (12). Previous studies also reported that Chinese women have the lowest femoral neck BMD compared to Malay and Indians (12). However, there is no study on prevalence of osteoporotic conditions or women BMD reference values in Singapore.
World Health Organisation (WHO) de ned osteoporosis as having BMD of 2.5 standard deviation (SD) or more below the young female adult mean, using normative data from NHANES reference database on Caucasian women aged 20-29 (13). However, peak BMD differs among ethnic and gender groups (14)(15)(16).
The purpose of this study is to establish BMD reference values for Singapore and compare the prevalence of osteoporosis by using our BMD data with those calculated from other available reference values. We hypothesise that there are considerable differences in the diagnosis of osteoporosis between DXA-machine provided reference data from other populations with our locally derived reference data.

Study design and participants
Random sampling methodology was employed to obtain a representative sample of approximately 300 male and 300 female participants, lling quotas of 20-40 participants in each sex-and age-group (10-year age-groups between 21-60 years; 5-year age-groups after 60 years). Conventionally, the sample size of 30 per age-group is su cient for normative measures (23). Between October 2017 and February 2019, using a two-stage random sampling method, 50% of all housing blocks were randomly selected, and a random 20% of the units in each block were approached for participant recruitment. Between March and November 2019, 50% of all housing blocks were randomly selected and all units were approached. Up to three eligible participants were recruited from each housing unit using a door-to-door recruitment method. Non-response units were re-contacted a second time at a different time of day on a later date. Older adults above 75 years old were additionally recruited through community sources and from a list of registered participants in four senior activity centres. Exclusion criteria were: individuals with disabilities, injuries, fractures or surgeries that affected function, neuromuscular, neurological and cognitive impairments, or more than ve poorly controlled comorbidities. Pregnant women or those planning for pregnancy were also excluded. The estimated overall response rate was 39.0%. Ethics approval was obtained from the National Healthcare Group Domain Speci c Review Board (2017/00212). All respondents signed informed consent before their participation in the study.

Measurements and data collection
Data collection was conducted by research o cers using validated questionnaires. Participants answered a socioeconomic and lifestyle questionnaire pertaining to education level, housing type, living arrangement, marital status, history of tobacco use (smoking) and alcohol use, a health and medical questionnaire indicating history of medical conditions and comorbidities, a mini nutritional assessment (MNA)(24), a global physical activity questionnaire (GPAQ) (25), and the LASA physical activity questionnaire (26). Menopausal status and hysterectomy were based on self-report.

Anthropometry
Body weight to the nearest 0.1 kg and height to nearest millimetre were measured using a digital balance and stadiometer (Seca, GmbH & Co. KG, Hamburg, Germany). Waist and hip circumferences were measured using a non-elastic, exible measuring tape around the navel and widest part of the hips respectively. Body mass index (BMI) was calculated as weight (kg) divided by height (m) squared.

Bone mineral density
Areal BMD was measured for the whole body using Hologic Discovery Wi (Hologic, Marlborough, MA, USA) in a supine position. The densitometry scan was conducted by experienced radiographers. Longitudinal quality control (QC) check and cross-calibrations were performed regularly. QC was performed daily using whole body and L1-L4 lumbar spine phantom provided by the manufacturer. Crosscalibration was performed weekly to monitor variations between the systems. Air scan test was performed weekly for table top uniformity. All participants were provided with a standard hospital gown, and requested to remove all garments and objects with at would potentially interfere with DXA scan. Lumbar spine BMD was extracted from the whole-body scan. Osteoporosis cut-off value was based on WHO diagnostic criteria using SD scores of BMD related to peak bone mass in healthy young women. Osteoporosis is de ned as having a BMD T-score of -2.5 or less, and osteopenia being de ned as a BMD T-score of between -1 and -2.5. Using young adult mean (YAM) and SD from other studies, T-scores were calculated for post-menopausal women, and men aged 50 and above. T-score derived using YAM from our study is referred to as T SG , T DXA refers to T-score calculated using Hologic densitometry reference value, T NHANES were derived from using NHANES database (27), and T CHN is from China cut-off values (20).

Statistical analysis
All statistical analyses were performed using SPSS Statistics version 22.0 (IBM, Armonk, NY, USA).
Relationship between BMD and age were analysed using polynomial regression model (up to the third degree). Models were tted to the LS and WB BMD as a function of age as shown: BMD = a + b 1 (age)+ b 2 (age) 2 + b 3 (age) 3 , where a is the intercept, and b 1 , b 2 , b 3 are regression parameters, which were estimated from observed data. Linear and quadratic models were tted, and the best-t model was chosen based on the R 2 value, and the signi cance R 2 changes between each degree model. Statistical agreement between Singapore cut-off value and DXA manufacturer reference value was analysed using kappa statistics.

Results
A total of 537 subjects (58.5% women) aged 21 and above, received the DXA scan. A total of 72 subjects were excluded due to poor image scan (i.e. having joint replacement surgery, metal implants, or unable to have a full body DXA scan), existing medical conditions (i.e. endocrine disorders, liver or renal disease, diabetes, or rheumatoid arthritis), and/or on medication or supplements (i.e. calcium supplements, uorides, steroids, or oestrogen) . Of the remaining 465 participants (55.9% women), 34.8% of women were post-menopausal and 30.7% of men were above the age of 50. Ethnic composition is 81% Chinese, 9% Malay, 6% Indians, and 4% others -similar to national population(6). Women' mean weight was 58.1±10.0kg, mean height was 155.4±6.3cm, mean BMI was 24.1±3.9, and the average age of menopause is 51 years. Men were heavier and taller -mean weight was 68.5±11.9kg, 166.5±6.75cm, and mean BMI was 24.6±3.7 (Table 1).
Normative bone mineral density values Table 2 shows height and weight of participants. Data from 413 subjects (223 women and 190 men) were used to compute normative BMD values. BMD, whole body (WB) and posterior anterior lumbar spine (LS) BMD of different age groups are presented in Table 3. Peak LS mean BMD was 1.090±0.168g.cm -2 in the 31-40 age group for pre-menopausal women, and 1.041±0.098g.cm -2 in 21-30 age group for men <50yrs of age. Peak WB mean BMD was 1.193±0.93g.cm -2 in 21-30 age group for pre-menopause women at, and 1.224±0.112g.cm -2 in the 21-30 age group for men <50yrs of age. A cubic polynomial regression model had best t relationship between BMD and age for women ( Figure 1). In men, linear and quadratic regression curve best described the relationship of LS and WB BMD with age respectively. In post-menopausal women, prevalence of osteoporosis using T SG is 9.3%, whereas prevalence was nearly twice at 17.3% with T DXA . In men above 50 years of age, prevalence of osteoporosis was only 0.7% using T SG , and 8-times greater at 5.6% when using T DXA (Table 4).
Cohen's weighted kappa (k w ) statistics were presented in Table 5.

Post-hoc analysis
Variations of WB and LS BMD show a trend of increase across age groups in men, which was different when compared to other studies (17,20,28,29

Discussion
Our results showed consistent differences between prevalence of osteoporosis and osteopenia when using the reference data provided by the densitometry manufacturer and our derived reference data. With higher reference values, the densitometry provided reference data has the tendency to over-estimate osteoporosis in Singapore population. It is well established that measured BMD differs between ethnicities(30). Using our reference values, the prevalence of osteoporosis was found to be 9.3% and 0.7% for post-menopausal women and men above 50 years of age, respectively. Comparatively, the prevalence of osteoporosis is estimated to be about 2-10 times higher at 17.3%, 20.4%, and 18.9% for women and 5.6%, 7.0%, and 5.6% for men, when using reference data from Hologic, NHANES, and China, respectively. Comparison between osteoporosis prevalence derived using reference data from our study, Hologic, and other countries, the percentage differences ranges between 60.5-163.6%.
Similar discrepancy was reported in another study in South East Asia. In a study of 653 Vietnamese men and women, prevalence of osteoporosis was 29% in women and 10% in men, with the local reference values. However, the prevalence were much higher -44% in women and 30% in men, when using the DXAprovided reference values (15). In a large Chinese study, the young adult BMD reference value was found to be 4-5% lower compared to the US reference values (19). This resulted in an arti cial 2-fold increase in the estimated prevalence of osteoporosis -from 13.3% to 26.7% in women, and from 5.9% to 11.1% in men (19).
The LS BMD of women from this study is higher in all age group, except for the 50-59 age group, when compared to Goh et al. (12). Furthermore, Goh (12) reported peak LS BMD occurring at 40-49 age group compared to 31-40 age group in our study. The difference is likely due to cohort effect, for the previous study was conducted about 16 years ago. In this study, a steep BMD decline occurs in women aged 41-50 for both LS and WB BMD. This is best explained by the decrease in oestrogen at menopause, as onset of menopause occur between 36-59 years of age in Singaporean women (31). Similar trends in LS BMD of women -with the decline of BMD in the same age range, were reported in studies of other ethnic groups (15,19,21). Our study showed that Singaporean women have higher LS BMD in the younger age group (21-50yrs) compared to the Caucasians. This may partly be attributed to an earlier onset of menarche in Singaporean women (12 years of age) compared to the Caucasians (13 years of age) (32,33). The YAM value for lumbar spine in women was similar to the NHANES and China values. However the different SD values explained the difference in prevalence. As the WHO de nition of diagnosis of osteoporosis depends on the SD value, a small change in SD value may cause notable variation in Tscore, resulting in signi cant effect on the number of individuals meeting the WHO diagnosis criteria (20).
From the regression model, the relationship between BMD and age in women was best tted with a polynomial equation of the third degree (Figure 1), which is consistent with other studies (15,19). According to the functional relationship, our data shows Singaporean women reaches predicted peak LS BMD at 31 years old -close to Caucasian (30-39 years old) (34,35) and Chinese women (34 years old) (19).
Age group at which peak LS BMD occur in men was similar with the previous study, however the peak BMD value was higher in our study (1.006±0.115g.cm -2 vs 1.049±0.119g.cm -2 )(36). LS BMD across all age group was also found to be lower compared to this study (36). These differences maybe due to better nutrition of the current cohort effect compared to the previous cohort from 18 years ago. The lower peak BMD reported, correspond with the low BMD in the current study 31-40 age group (Table 3). The WB BMD and LS BMD of men in our study do not share similar trend as the women. There is a steep decline in the 31-40 age group, and then increasing trend in older age groups. Similar trends had been reported for WB (17) and LS BMD (35) elsewhere. Singaporean men have a lower LS BMD in the 21-50 years age group but higher in older age groups compared to the Caucasians. Compared to Caucasians, Singaporean men aged 66 years and older have a higher WB BMD. The relationship between LS BMD with age was best tted with a positive linear regression equation, and the relationship between WB BMD and age was best tted with a polynomial equation of the second degree ( Figure 1). Despite the apparent increasing trend in BMD in the older age group, post-hoc analysis revealed no signi cant difference in BMD within age groups. Similar trend was also reported in the NHANES studies, regardless of race and ethnicity (35). YAM was found to be lower in Singapore men compared to NHANES and China. Sedentary lifestyle (78%) in Singaporean adolescent may explain the lower YAM BMD (37). The relationship between age and BMD in older men could be confounded by cohort effect, with the older generation having a healthier lifestyle (38). However, such relationship has been reported in other studies (20,39,40). For older men, BMD have been found to increase with age (41). The apparent increase in BMD have been attributed to degenerativechanges of the LS, such as osteophytes, aortic calci cation, vertebral compression fracture, scoliosis, and osteoarthritis (42). The difference in rate of spinal BMD diminution was found to be 1% and 10% per decade in men and women, respectively (43). However, other studies reported that lumbar spine BMD increases at a rate of 1.5-3.5% per decade in men over 60 years old (44)(45)(46). The compound effect of older men physiology and the cohort effect likely explained the lack of signi cant change in BMD across age in our study.
Ethnicity has been established as a determinant of BMD and the risk of osteoporosis (30). While majority of BMD research have compared the local reference values to the Caucasians, it is important to note that variance in BMD values was also reported among different ethnic groups in Asia. For example, the agespeci c BMD in Chinese was reported to be lower compared to the Japanese and Koreans (20). Differences was also found among different ethnic group in a country based on geographical locations, possibly due to differences in diet, lifestyle and body size (20). As a country in South East Asia, while predominantly ethnic Chinese, diet, lifestyle and culture of Singaporean differs from those of people groups in China. Additionally, there are also ethnic variation within Singapore. Singaporean Chinese women has a 40% higher hip fracture rates compared to local Malay and 90% higher than Indian women (2). Post-hoc secondary analysis of covariance our data showed that there were no signi cant differences in LS BMD among the three major ethnic group for both men (p=0.692) and women (p=0.802). This result is similar to previous study (12). However, signi cant differences was reported in femoral neck BMD among Singapore women (12).Though the BMD differences between different ethnicity in Singapore has not been investigated, various in uencing risk factors for low BMD, such as obesity, vitamin D de ciency, smoking, alcohol consumption and genetic heterogeneity, have been studied. Ethnic-speci c genetic variants and risk factors associated with low BMD warrant future research (2).
Lifestyle is also an established determinant of BMD. As this is a cross-sectional study, inter-generation lifestyle differences may have a cohort effect on the mean BMD of each age group. Singapore's economic development is unique. Over three decades, it has progressed from a labour-intensive industry to a predominantly modern-day service industry. In 2016, 36.5% of Singaporean adults are reported to be physically inactive, and 69.7% adolescent boys and 83.1% adolescent girls between 11-17 years of age are physically inactive (47). It has been reported that 10% increase in peak bone mass in children will reduce the risk of osteoporotic fracture by 50% during adult life (48). Low physical activity in adolescents may increase risk of low bone mass and osteoporosis in middle to older ages.
To our knowledge, this is the rst prevalence study of osteoporotic conditions in Singapore using population-based BMD reference values. While study sample size may not be large, it is randomly selected a priori from a nationally and ethnically representative residential population. Therefore, this population-based dataset adds to the much needed local and South East Asian BMD reference database.
Individuals with clinical conditions deemed to interfere with bone metabolism were also excluded from the data analysis. Ideally, peak bone mineral density should be estimated from a longitudinal study, following up participants from the age of 5 to 40, but such study is not feasible. The DXA lumbar spine BMD was extracted from the whole-body scan, which may affect the accuracy of the BMD obtained. An important limitation is that other skeletal sites, such as femoral neck, Ward's triangle, trochanter, and total hip, were not available. The study was part of a larger study on body composition and physical performance and so only a whole body DXA scan using standard protocol was performed (where the regions of interest are head, arm, forearm, leg, ribs, thoracic spine, lumbar spine and pelvis). Site-speci c scans of femoral neck and hip region was not performed. Fractures at hip or femoral neck are known to be more devastating. Therefore, further local population-based study on these areas will be needed.

Conclusion
Our study showed that the prevalence of osteoporotic conditions of Singaporeans is not accurately represented when using reference data provided by the DXA manufacturer; and contributes to local and Asian population reference databases. There could have been overestimation of osteoporosis and osteopenia by using manufacturer provided reference data, that may possibly lead to over-treatment in certain segment of our population. There is an urgent need to establish a South East Asian BMD reference database for lumbar spine and femur neck to provide an accurate picture of the prevalence of osteoporotic conditions of population in this region. It is also important to better understand determinants of BMD and bone health in Southeast Asia.    % Difference is the percentage of difference between NORMS and other countries.