Ten-Year Risk of Osteoporotic Fractures in Postmenopausal Chinese Women According to Clinical Risk Factors and BMD T-Scores: A Prospective Study


  • The authors state that they have no conflicts of interest.


Independent risk factors for osteoporotic fracture were identified for a Southern Chinese postmenopausal population. Clinical risk factor assessment with or without BMD measurement was shown to be an effective predictor of 10-yr risk of osteoporotic fracture and provides a more accessible tool for patient evaluation.

Introduction: Asian-specific data on risk factors for osteoporosis remain sparse. However, risk factor assessment, in addition to BMD measurement, is increasingly recognized as a reliable predictor of absolute osteoporotic fracture risk. The purpose of this prospective study was to determine the specific independent risk factors for osteoporotic fracture and to predict the 10-yr risk of osteoporotic fracture in the postmenopausal Southern Chinese population.

Materials and Methods: A total of 1435 community-dwelling, postmenopausal, treatment-naive women were recruited. Baseline demographic characteristics and clinical risk factors were obtained, and BMD at the spine and hip was measured. Subjects were followed for outcomes of incident low trauma fracture. Ten-year risk of osteoporotic fracture was predicted from the risk factor assessment and BMD measurement by Cox proportional hazards models.

Results: The mean age of subjects was 63.4 ± 8.3 yr. After 5.0 ± 2.3 yr (range, 1.0–11.0 yr) of follow-up, 80 nontraumatic new fractures were reported during follow-up. Eight independent clinical risk factors identified at baseline were found to be significant predictors of osteoporotic fracture, with the most important being use of walking aids (RR, 4.2; 95% CI, 2.7–6.7; p < 0.001) and a history of fall (RR, 4.0; 95% CI, 2.5–6.2; p < 0.001). Other predictive factors included being homebound, calcium intake < 400 mg/d, age > 65 yr, history of fracture, and BMI < 19 kg/cm2. Subjects with three to eight clinical risk factors had a predicted 10-year risk of osteoporotic fracture of 25%, which increased to 30% if they also had total hip BMD T-score ≤ −2.5.

Conclusions: Clinical risk factor assessment, with or without BMD measurement, is a reliable predictor of 10-year risk of osteoporotic fracture and may be particularly useful in regions or primary care clinics without access to bone densitometry equipment.


Whereas the burden of many chronic diseases is predicted to increase in the future, primarily caused by an aging global population, osteoporosis is a particular public health concern for Asia as this region is expected to carry the major burden of this disease.(1) The risk and incidence of osteoporosis varies widely between populations. It is well documented that ethnic- and other population-specific data are needed to effectively target osteoporosis in a given population.(2) However, to date, information on clinical risk factors for the prediction of osteoporotic fractures is based on data from untreated postmenopausal white women.(3–6) Knowledge of risk factors for Asian women remains sparse.(7)

Whereas BMD measurement has been the standard used to diagnose and initiate intervention in osteoporosis,(8–10) increasing evidence suggests that BMD alone is inadequate to detect all at-risk individuals.(11–13) This is mainly because osteoporotic fracture risk fluctuates at any given BMD,(11) and subjects defined as having normal BMD values according to current World Health Organization (WHO) classifications(8) have subsequently been found to have increased osteoporotic fracture risk and associated clinical outcomes.(12) Also, the fracture risk burden has been identified to be greatest among women classed as osteopenic (T-score between –1.0 and –2.5),(14) but these women are not necessarily indicated for osteoporosis treatment, based on current recommendations that use T-score ≤ −2.5 as the main diagnostic criterion for osteoporosis. Based on this, a substantial proportion of women with an elevated risk of osteoporotic fracture are being undertreated. Another limitation of using only BMD to predict and diagnose osteoporosis is the lack of accessibility to DXA in some countries (including many developing countries in Asia) and the restriction it places on diagnosis taking place in the primary care setting, as many primary care facilities do not have DXA equipment. Thus, using BMD alone to determine osteoporosis may further contribute to the problem of underdiagnosis and undertreatment of osteoporosis.

Recent publications have suggested that an individual's 10-yr absolute risk of osteoporotic fracture is likely to be a more reliable predictor of osteoporotic fracture than BMD measurement alone.(11,15) Within this context, the usefulness of risk factor assessment is becoming more widely recognized. The purpose of this prospective study—likely to be the first of its kind in Asia—is to determine the specific independent risk factors for predicting the 10-yr risk of osteoporotic fracture in postmenopausal Chinese women in Hong Kong.



The participants were community-dwelling, postmenopausal, Southern Chinese women recruited from public road shows and health fairs held in various districts of Hong Kong. The inclusion criteria for the study has been previously described in detail.(16,17) In brief, women who were at least 12 mo past their last menstrual period, ≥45 yr of age, of Southern Chinese ethnicity, and not already receiving treatment for osteoporosis or drugs known to affect bone and mineral metabolism were eligible for participation. From January 1995 to December 2002, 1435 subjects were assessed at the Osteoporosis Center of Queen Mary Hospital, The University of Hong Kong. Before study entry, all participants gave informed consent, and the study was conducted with strict adherence to the ethical requirements of the Declaration of Helsinki and local Ethics Committees. The study protocol was approved by the Institutional Review Board of the University of Hong Kong and the Hospital Authority Hong Kong West Cluster.

Risk factor assessment

In an interview with a trained research assistant, baseline demographic data on anthropometric measurements, socioeconomic status, education level, and medical and reproductive history were taken. In addition, self- and family history of osteoporosis and low-trauma fractures after the age of 45 yr were obtained using a structured questionnaire. Information on lifestyle habits, including smoking, alcohol consumption, and physical activity, was obtained. Physical activity included all weight-bearing exercise and walking. The time spent on any weight-bearing physical activity was added and scored from 1 to 4 (1 = 0 min/d, 2 = 0–30 min/d, 3 = 30–60 min/d, 4 = >60 min/d). Dietary intake of calcium and phytoestrogen was determined using a semiquantitative food frequency questionnaire.(18) Patients were followed up yearly for outcomes of incident low-trauma fracture by telephone interview using a structured questionnaire. All low-trauma fractures except those of the skull, fingers, and toes were included, but traumatic fractures were excluded from the analyses. The information was subsequently confirmed by participants' medical records. Participants' medical records were readily accessible using the centrally linked, computerized network of the Hong Kong Hospital Authority, which manages outpatient clinics and hospitals attended by the majority (94%) of all Hong Kong patients. For those patients who did not attend Hospital Authority clinics, clinical outcome information was verified by their attending physicians.

BMD measurement

BMD was measured at the L1–L4 lumbar spine, femoral neck, and total hip region using DXA (Hologic QDR-4500). Instruments were calibrated daily. The in vivo precisions of DXA used for lumbar spine, femoral neck, and total hip were 1.2%, 1.5%, and 1.8%, respectively.(18) Thoraco-lumbar spine X-rays were assessed for radiographic evidence of spine fractures at baseline.(19) All DXA measurements were performed by two licensed technologists who had completed training by the equipment manufacturers and were accredited by the International Society for Clinical Densitometry. Bone mass measurements were expressed in grams per square centimeter and T-score, using WHO classifications.(8) BMD Z-scores and T-scores were determined from local reference data.(16)

Statistical methods

The Cox proportional hazards models were used to identify potential independent risk factors for osteoporotic fracture. Results were reported as relative risks (RRs) with 95% CI. The significance level was set at p < 0.05. All statistical analyses were conducted using SPSS for Windows version 13.0 statistical software (SPSS, Chicago, IL, USA). The risk of osteoporotic fracture was optimally expressed as a fixed-term absolute risk; that is, the probability of fracture over a given period of time. Predicted 10-yr fracture risk, as well as the relationship between age, BMD T-score and number of risk factors, was identified.


The average follow-up period of study participants was 5.0 ± 2.3 yr (range, 1.0–11.0 yr). Of the 1435 subjects initially recruited, 115 were lost to follow-up, and 27 had died at the time of analysis—the data for these subjects were analyzed up to their last contact time-point. In addition, 35 subjects who received treatment with antiresorptives or hormonal replacement therapy after recruitment were excluded from analysis. The average age of this population at recruitment was 63.4 ± 8.3 yr (range, 45–91 yr); 58.5% were 45–64 yr of age and 41.5% were ≥65 yr of age. Baseline demographic information and BMD characteristics of the subjects are shown in Table 1.

Table Table 1.. Baseline Demographic and BMD Characteristics of Postmenopausal Chinese Women (n = 1400)
original image

During the mean 5-yr follow-up period, 80 new low-trauma fractures were reported, of which 17 (21.3%) were clinical vertebral fractures; 17 (21.3%) were hip fractures; 10 (12.5%) were proximal humerus fractures; 17 (21.3%) were distal forearm fractures; and 19 (23.8%) were fractures at other peripheral sites.

Predicted 10-yr osteoporotic fracture risk from risk factor assessment

In multivariate analyses, eight independent clinical variables associated with increased risk of osteoporotic fracture (Table 2) were identified. In terms of individual risk factors, the use of walking aids and a history of one or more falls in 12 mo were associated with the highest predicted 10-yr risk of fracture. Other risk factors listed in decreasing order of their impact to fracture risk were being housebound; dietary calcium intake <400 mg/d; age >65 yr; previous history of fracture; BMI <19 kg/cm2; and physical activity <30 min/d. Clinical risk factors of borderline significance (p value between 0.05 and 0.10) included family history of hip fracture, smoking, alcohol consumption, and use of steroids. The percentage of subjects with these risk factors is shown in Table 2.

Table Table 2.. Risk Factors Associated With Osteoporotic Fracture in Postmenopausal Chinese Women
original image

The impact of age on the risk of osteoporotic fracture was analyzed. A 10-yr increase in age was associated with a 2.2-fold increase in fracture risk. Figure 1 shows the interaction of age with other clinical risk factors on fracture risk prediction. Regardless of the risk factor studied, subjects ≥65 yr of age had 1.5- to 2.0-fold higher risk of osteoporotic fractures than subjects younger than 65 yr. Even among those without any other clinical risk of fracture, the 10-yr risk of osteoporotic fractures of subjects ≥65 yr of age was 7.5% compared with 3.8% among subjects <65 yr of age. Interestingly, of all the risk factors, age had the least impact with the variable of low BMD. Among those with total hip BMD T-score ≤−2.5, subjects ≥65 yr of age only had a 1.4-fold increased risk of fracture compared with subjects <65 yr of age (Fig. 1).

Figure Figure 1.

Interaction of age with other clinical risk factors on 10-year risk of osteoporotic fracture.

Predicted 10-yr osteoporotic fracture risk from BMD and number of risk factors

In total, 26.9% (n = 377) of subjects at baseline had a lumbar spine BMD T-score ≤−2.5; 21.9% (n = 307) had a femoral neck BMD T-score ≤−2.5; and 21% (n = 294) had a total hip BMD T-score ≤−2.5. Overall, 36.8% of this population had osteoporosis, with a BMD T-score ≤−2.5 at any site. Only 9.6% of women 45–65 yr of age had osteoporosis at baseline evaluation compared with 35.8% of women ≥65 yr. Similar to observations in white women, each SD reduction in BMD at either the spine or hip was associated with a 1.5- to 2.0-fold increased risk of osteoporotic fracture in this population (Table 2). Documentation of osteoporosis by BMD measurement improved fracture risk prediction. The additional improvement in fracture risk prediction with BMD measurement is shown in Fig. 2. The 10-yr risk of fracture of postmenopausal women who had osteoporosis at the hip but no other clinical risk factors was 6% compared with 3% in women who had no clinical risk factors or osteoporosis at the hip. Women ≥65 yr of age with a total hip BMD T-score ≤−2.5 had a 10-yr risk of fracture of 17% compared with 12% for those with total hip BMD T-score >−2.5 (Fig. 2). Figure 3 shows the 10-yr absolute risk of osteoporotic fracture according to age and total hip BMD T-score based on the data derived from this cohort.

Figure Figure 2.

Comparing the 10-yr fracture risk prediction with clinical risk factors with or without addition of BMD measurement.

Figure Figure 3.

Ten-year risk of osteoporotic fracture in southern Chinese women according to age and BMD T-score.

Figure Figure 4.

Predicted 10-yr risk of osteoporotic fracture according to number of risk factors and total hip BMD T-score.

Although BMD measurement provides additional information for predicting fracture, clinical risk factors are likely to be more important, because subjects with multiple risk factors had higher predicted 10-yr risk of fracture than subjects with low BMD measurement alone. Subjects with three to eight clinical risk factors and a femoral neck BMD T-score ≥−1.0 had a higher 10-yr risk of fracture than subjects with femoral neck BMD T-score ≤−2.5 but no clinical risk factors. Figure 4 shows that subjects with one to two clinical risk factors and total hip BMD >−1 had a 10-yr risk of fracture of 8.8%, whereas subjects with a total hip BMD T-score ≤−2.5 but no clinical risk factors only had a 5.3% risk of fracture in 10 years. In the study, 18% of women had three or more risk factors. The predicted 10-yr osteoporotic fracture risk was found to increase as the number of risk factors increased (Fig. 4), and subjects with three to eight risk factors and low BMD had the highest 10-yr risk of fracture (30%).

Comparison with other ethnic groups

To evaluate the 10-yr fracture risk of this cohort of Chinese postmenopausal women in relation to that of women of other ethnic groups,(20,21) the risk according to age and femoral neck BMD T-score was stratified. The data are shown in Fig. 5.

Figure Figure 5.

(A) Ten-year probability of osteoporotic fracture (hip, clinical spine, forearm) by age and country. Fracture data for other countries were based on references 20 and 21. (B) Ten-year probability of osteoporotic fracture (hip, clinical spine, forearm) by BMD T-score and country. Fracture data for other countries were based on references 20 and 21.


Global epidemiological data indicate that Asia will carry the greatest burden of osteoporosis over coming decades.1 However, osteoporosis is a disease that presents realistic opportunities for prevention and modification, particularly if there is a concerted government- and population-driven effort to implement disease prevention strategies. Central to prevention is the early identification of at-risk individuals. Previous studies of osteoporotic risk in Asian populations have been case-controlled and cross-sectional studies. This study may be the first prospective study of Asian postmenopausal women to verify BMD as a useful predictor of osteoporotic fracture and to study risk factor assessment as a reliable predictor of absolute 10-yr risk.

The study provides ethnic-specific data on independent risk factors that significantly contribute to the absolute 10-yr risk of osteoporotic fracture, which can be reliably used for diagnosis, risk prediction, and more importantly, treatment decision. The independent risk factors identified to be significant in this Chinese population were similar to some of those previously reported from white prospective studies.(3,20)

The risk factors identified in our population can be broadly grouped into those related to falls and low BMD. Interestingly, the three most important independent risk factors for low-trauma fracture in this Asian cohort are the use of walking aids, history of falls, and being housebound. These factors reflect and contribute to poor muscular function and coordination, which in turn contributes to increasing the risk of falls and fracture. Therefore, among postmenopausal Asian women, frailty or poor physical functioning alone should be considered as the most important indicators alerting to a significantly increased fracture risk, even in the absence of other known clinical risk factors and BMD data. In subjects with these clinical risk factors, fall prevention interventions should be the primary emphasis of their treatment protocol. Not only is fall prevention therapy a likely cost-effective approach for fracture prevention in developing Asian countries, but previous studies have shown that the use of risedronate for hip fracture prevention is ineffective in elderly patients selected on the basis of high risk of falling but not low BMD.(22) Information of these risk factors enables the judicial use of expensive anti-osteoporosis agents.

Other risk factors identified in this study are probably related to low BMD. These include walking outdoors for <30 min/d, low calcium intake, previous fractures, age >65, and low BMI. Individuals who spend limited time outdoors are likely to have poor physical function, but it is conceivable that they also have vitamin D insufficiency because of reduced exposure to sunlight. Our previous study found a high prevalence of vitamin D insufficiency in ambulatory Southern Chinese, and it is envisaged that the problem could be more severe in those who are housebound or undertake limited outdoor activities.(23,24) This risk would be further increased if the daily dietary calcium intake was also <400 mg.

In terms of BMD, the relative risk prediction for osteoporotic fracture was better with measurement at the hip than the spine. This observation is in agreement with studies conducted in white populations.(25) In terms of 10-yr risk of osteoporotic fracture based on age and BMD T-score, the data in this study were very similar to the 10-yr probabilities of fracture in Swedish women as reported by Kanis et al.,(20) which was modeled from fracture incidence and future mortality rates. In a previous study of Asian women, it was shown that, among a number of independent clinical risk factors, the two most important predictors of low BMD and osteoporosis were age and body weight.(26) The Osteoporosis Self-Assessment Tool for Asians (OSTA) index, based only on age and body weight, was shown to have sufficient sensitivity to detect individuals at risk for osteoporosis. Indeed, the OSTA index has the same sensitivity in predicting osteoporosis as bone ultrasound.(27) The results of this study confirmed that age and body weight/BMI are two other important independent risk factors contributing to increased absolute risk. Although BMI < 19 kg/m2 was found in only 2.5% of the cohort, this was identified as a significant independent risk factor for fracture. The mean BMI of study participants was 23.8 ± 3.5 kg/m2, which is within the ideal range of BMI range for adult Asian women.(28) However, earlier studies of elderly Hong Kong men and women have reported that there is a trend for decreasing body weight with increasing age among women but not among men.(29) Thus, although this study population had a healthy BMI, maintenance of this BMI is important for long-term outcomes and fracture prevention.

This study found that a previous history of fracture was a strong predictor of increased osteoporotic risk, confirming the results of earlier studies, showing that a previous history of fracture at any site significantly increases the risk of future fracture.(30) In women of all ages with pre-existing vertebral fractures, the risk of subsequent fracture was approximately four times that of women without a fracture history, and this risk increased as the number of prior vertebral fractures increased.(30) Other studies in peri- and postmenopausal women have found that the risk of subsequent fracture in women with prior fracture is double that of women without a fracture history.(30) In addition, the risk of a subsequent fracture occurring within 1 yr of the original event is substantial—up to five times that of women who have not had a previous fracture.(31)

Risk factors reported previously as being independent factors for increased osteoporotic fracture, such as a family history of hip fracture, were studied,(7) but the association was found to be only of marginal significance. This is likely because of a lack of information about family history from many participants who lost their parents in wartime or had lost contact with family members living in mainland China. Current smoking status and alcohol consumption were also not independent risk factors after multivariate adjustment. However, the low proportion of smokers and alcohol drinkers in this study population (3.2%, n = 45), largely representative of the general middle-aged and elderly female population in Hong Kong, means that even if these lifestyle practices were minimized or removed, the effect on reducing osteoporotic fracture risk would be marginal.

The predicted 10-yr risk of osteoporotic fracture in Southern Chinese women in Hong Kong 50–79 yr of age was comparable to white women in Spain, the United Kingdom, and Sweden, but was higher than that for a similar population in mainland China. The very low fracture risk in women in mainland China is believed to relate to a shorter hip axis length, a more active lifestyle, and the practice of squatting, all resulting in lower risks of falling and hip fracture.(32,33) Irrespective of ethnicity, women 80 yr of age have the highest absolute risk of osteoporotic fracture, a finding that is confirmed by this study. However, the results show that the risk among Hong Kong women in the oldest age group (80 yr) is actually considerably higher than that reported for other similar populations. Although western data have shown that the cumulative risk of osteoporotic fracture decreases in subjects >80 yr of age compared with those 70–80 yr, this relative reduction is mainly caused by reduced remaining life expectancy among those older than 80 yr. However, in Hong Kong, with lengthening of the remaining life expectancy of women at the age of 80 over the past two decades, data from the Hospital Authority showed that the mean age of hip fracture patients is 85 yr.(34) Thus, unlike western populations, the 10-yr probability of osteoporotic fracture among Hong Kong Chinese women remains high even with advancing age (≥80 yr). This is also likely to be the case for other populations with increasing life expectancy, particularly beyond the age of 80 yr.

The findings of this study indicate that the use of BMD alone to predict osteoporotic fracture risk may no longer be appropriate, particularly as the range of risk varies markedly at any given BMD(11) and with age.(13) The addition of hip BMD measurement to risk factor assessment undoubtedly improves osteoporotic fracture risk prediction (Fig. 3). However, it is also clear that, in subjects who have a combination of independent risk factors with high predictive value, such as a history of fracture, advanced age, and low body weight/BMI, the probability of fracture predicted from risk factor assessment alone is sufficiently high to warrant the initiation of treatment.

Nevertheless, BMD measurement has a valuable predictive and diagnostic role. The BMD T-score at baseline evaluation provides an index for the severity of bone fragility, and when information on absolute risk of fracture is lacking, it would be difficult to justify the use of long-term treatment to health authorities and patients alike. Furthermore, the experience from use of bisphosphonates showed that risedronate had limited effect in hip fracture prevention in elderly women selected on the basis of a high risk of falling rather than low BMD.(22) It is therefore important to carefully assess individual risk factors, including BMD measurement, before the initiation of long-term drug therapy for maximizing the cost-effectiveness of the treatment. Besides, BMD provides a tangible clinical measure with which treatment effects can be assessed over time. The results of this study support the view that risk factor assessment and BMD measurement have a complementary value for fracture prediction.(3) However, in terms of an accessible, affordable, diagnostic tool in Asia—which currently has widely varying and often limited access to DXA technology, particularly in remote or less socially and economically developed areas—risk factor assessment provides a far more amenable option than BMD measurement. Risk factor assessment also facilitates diagnosis in the primary care setting, where many at-risk individuals initially present. Providing primary care physicians with a reliable and cost-effective means to diagnose osteoporosis (particularly when DXA is not available) would improve the diagnosis and detection of this disease. However, clinical trials are needed to prove that anti-osteoporosis drug treatment would actually prevent fractures in women selected for therapy on the basis of these clinical risk factors alone.

This study has some limitations. Being a community-based study, the population was not randomly selected and may not represent the general population of Hong Kong. However, the prevalence of osteoporosis in this cohort was similar to other published prevalence data on postmenopausal women in Hong Kong.(35) Women who were included in the follow-up may have differed from those lost to follow-up because of nonresponse or death (9.9%). The effect that these nonrespondents may have had on the study results is unknown. There is also the possibility that this study may have underestimated the absolute risk in the general postmenopausal population because, after initial baseline evaluation of BMD, all participants received a clear explanation of their DXA report and were educated on the importance of risk prevention in osteoporosis. The consequences of this intervention were not quantified, but it is thought that it might have contributed to reducing the overall risk of study participants who, after receiving advice, may have implemented lifestyle modifications and other risk prevention strategies during the 5-yr follow-up period. Thus, the actual risk of the general postmenopausal population in Hong Kong that has not received education about osteoporosis prevention or been informed about BMD status is likely to be greater than that reported for the study population. Similar to other studies, the 10-yr fracture risk of this study was predicted using the Cox proportional hazard model based on results generated from a median follow-up period of 5 yr. Actual 10-yr follow-up information for every subject was not available. Because fracture risk increases with age and frailty, the predicted model may well underestimate the actual fracture risk of the population.

In conclusion, this study has identified the major clinical risk factors specific to Southern Chinese postmenopausal women that effectively predict fracture risk, with or without BMD data. This information will allow optimal risk assessment to target those patients to whom treatment would derive the greatest benefit, particularly in the primary care setting and in areas without easy access to instruments for measuring BMD. Information on these clinical risk factors will also enable the triaging of patients for different modalities of intervention. These interventions include fall prevention, lifestyle modification, and pharmacological drug treatment. Availability of clinical risk information and fracture probability will provide a basis for modeling the absolute risk of fracture based on life expectancy and allow determination of interventional threshold for our population.


The authors thank the staff of the Osteoporosis Centre for assistance in this study, and CF Yiu for statistical advice and Cathy Chow for assistance in preparing the manuscript. This study was supported by the Bone Health Fund of the Hong Kong University Foundation, CRCG grant, Matching Grant, Osteoporosis and Endocrine Research Fund of the University of Hong Kong.