Introduction

Osteoporosis and its consequence of fracture are recognized as an important global public health problem, because fracture is associated with increased mortality [1], concomitant morbidity, and reduced quality of life [2]. It is projected that approximately half of all hip fractures in the world occur in Asia [3, 4]. Despite the expected magnitude of the problem, the prevalence of and risk factors of osteoporosis in Asian countries are not well documented due to lack of epidemiologic data.

Osteoporosis is often considered as a consequence of industrialization, because the incidence of osteoporotic fractures is higher in industrialized countries than in developing countries [4]. Furthermore, within a country, the incidence of fractures is higher in urban communities than in rural communities [5–8]. Therefore, study of osteoporosis prevalence in emerging developing countries may provide useful information on the evolution of the disease.

Vietnam is a developing country with a population of 84 million. After decades of wars and poverty, the country has been embarking on a market economy with a rapid urbanization being taken place in the entire country. Although osteoporosis is increasingly recognized as a major public health concern in Vietnam, there is currently a lack of documentation of the prevalence of osteoporosis and data on the peak bone mass in the population is not available. In a previous study based on quantitative ultrasound measurements (QUS), it was estimated that 20% of postmenopausal women had low bone mass (subjects whose T-score were ≤ −2.5 SD) [9]. However, QUS is not an ideal means to assess skeletal health; therefore, the prevalence of osteoporosis in the Vietnamese population remains unknown.

The present study was designed to estimate the peak bone mineral density, age of its attainment and the prevalence of osteoporosis in Vietnamese women aged 50+.

Study design and methods

Study design

The study was designed as a cross-sectional investigation with multistage sampling scheme. The study was approved by the ethics committee of Hanoi Medical University. The setting was in Hanoi, the capital of Vietnam. Two districts (Dong Da and Soc Son) in Hanoi were selected for the study site. In each district, 4 communes were randomly selected, and then a full list of subjects in each commune was obtained from the local government authority which served as the sampling frame. The lists were then sorted by age 10-year group. For each age group, a total of 100 women were randomly selected by a computer-generated numbers for screening interview. We assumed that the peak BMD was 1 g/cm2 with standard deviation being 0.1 g/cm2, then the required sample size was about 120–150 women aged between 10 and 49 years. The targeted sample was therefore 400 women. Among the 400 women randomly selected, 373 women (response rate of 93%) agreed to participate in the study. However, after excluding women who did not meet the study entry criteria, only 168 women were ultimately included in the study. A further random sample of women aged between 50 and 65 from the same communes were invited to participate in the study for estimating the prevalence of osteoporosis. In this sampling scheme, women aged 50 and 65 were randomly selected from all communes. Initially 386 women agreed to participate in the study; however, only 160 women met all the inclusion and exclusion criteria. In total, 343 women aged between 10 and 65 were eligible for participating in the study.

Women were excluded from the study if they had conditions that are deemed to affect bone metabolism such as chronic diseases (dysfunctions of liver, gut, kidney, endocrine system, respiratory or cardiovascular system, auto-immune diseases and cancer), or use of medications, or more than one month in bed, or premature menopause or hysterectomy/oophorectomy, or history of fracture.

Measurements

BMD was measured at the lumbar spine (LS), left and right femoral neck (FN) and total hip (TH) in all qualified subjects. The measurements were done with a dual energy X-ray absortiometry (DXA) densitometer (GE Lunar Prodigy advance). The precision error (%CV) in our laboratory was 1.75% for lumbar spine and 1.5% for hips. The machine was standardized by standard phantom every time before measurement. In this analysis, BMD at the lumbar spine was estimated from L2 to L4. Femoral neck and total hip BMD used in the analysis was estimated from the right side.

Anthropometrical parameters including weight, standing height, sitting height, leg length, chest, waist, and hip circumferences were collected. Body weight was measured by using electronic balance with indoor clothing without shoes. Height was determined without shoes on a portable stadiometer with mandible plane parallel to the floor.

Data analysis

In order to estimate peak BMD, a series of polynomial regression models (up to the third degree) were fitted to femoral neck, total hip and lumbar spine BMD as a function of age. In these models, the expected value of BMD at each skeletal site was expressed as follows: \(BMD = \alpha + \beta _{1} {\left( {age} \right)} + \beta _{2} {\left( {age} \right)}^{2} + \beta _{3} {\left( {age} \right)}^{3} \), where α is the intercept, β1, β2, and β3 are regression parameters to be estimated from the observed data. Reduced models (i.e., quadratic and linear models) were also considered, and the “final” model was chosen based on the Akaike Information Criterion (AIC) [10]. Peak BMD and age that reached peak BMD were estimated from the final model. The 95% confidence interval of peak BMD and aged of peak BMD were determined by the bootstrap (resampling) method [11]. In this method of analysis, 1000 samples, each with 300 individuals, were drawn from the original sample with replacement, the cubic equation was fitted to each sample, and the 95% confidence of the 1000 peak BMD and age of peak BMD were then determined. The analysis was performed with R environment [12] and the Design library [13].

From the peak BMD, T-score, which is the number of standard deviations from the peak BMD, were calculated for each individual in the study sample. The prevalence of osteoporosis was estimated for women aged 50+. An individual was classified as having osteoporosis if her femoral neck BMD T-score was equal or lower than −2.5 [14]. Standard error and 95% confidence interval of the prevalence were estimated based on the assumption of binomial distribution.

Results

In total, 328 women aged between 10 and 65 years enrolled in the study, whereas 15 subjects for various reasons were unable to participate. Out of these 328 women, 49% were aged 50 years or older. Mean values for weight and height of the different age groups (Table 1) were similar to the normal range as reported for Vietnamese women [15]. BMD measurement at each site was significantly and independently correlated with height (r = 0.42; p < 0.0001) and weight (r = 0.46; p < 0.0001). Lumbar spine BMD was significantly correlated with total hip BMD (r = 0.75; p < 0.0001), which was in turn correlated with femoral neck BMD (r = 0.94; p < 0.0001).

Table 1 Characteristics of study subjects
Full size table

As expected, cross-sectionally, BMD at the lumbar spine, total hip and femoral neck rapidly increased during the adolescence and then gradually declined after the age of 40 (Figure 1). The equations that best described this trend was a third degree polynomial function of age as follows:

$$\begin{aligned} & {\text{Lumber}}\,{\text{spine}}\,{\text{BMD}} = 0.1236 + 0.0855 \times {\text{Age}} - {\text{0}}{\text{.0022}} \times {\text{Age}}^{2} + 0.0000155 \times {\text{Age}}^{3} \\ & {\text{Total}}\,{\text{hip}}\,{\text{BMD}} = 0.3582 + 0.0532 \times {\text{Age}} - 0.0013 \times {\text{Age}}^{{\text{2}}} + 0.00009 \times {\text{Age}}^{3} \\ & {\text{Femoral}}\,{\text{neck}}\,{\text{BMD = 0}}{\text{.3353 + 0}}{\text{.0508}} \times {\text{Age}} - 0.0013 \times {\text{Age}}^{{\text{2}}} + 0.000009 \times {\text{Age}}^{3} \\ \end{aligned} $$
Fig. 1
figure 1

Relationship between age and lumbar spine BMD (panel A), total hip BMD (B), and femoral neck BMD (C)

Full size image

These equations accounted for 47%, 30% and 36% of the variation at the lumbar spine, total hip, and femoral neck BMD, respectively. The regression coefficients shown above suggest that the increase in BMD during the adolescence was greater at the lumbar spine than at the hip (0.085 vs 0.05 g/cm2); however, the subsequent decrease in BMD was also greater at the lumbar spine than at the hip (0.0022 vs 0.0013 g/cm2). For any given site, BMD among those aged 50–65 years was reduced by between 20% and 28% as compared to the peak BMD level.

Based on the bootstrap analysis, peak BMD and the age when peak BMD reached were estimated (Table 2 ). According to these results, the peak BMD values for the lumbar spine and total hip were 1.16 g/cm2 (SD 0.13) and 1.02 g/cm2 (SD 0.12) respectively. The age at peak BMD was around 29 years. The corresponding value for peak femoral neck BMD was slightly lower: 0.94 g/cm2 (SD 0.11) and the age at peak BMD was around 28 years.

Table 2 Estimate of peak BMD, standard deviation and age of peak BMD
Full size table

Using the estimated peak BMD and standard deviation, femoral neck T-scores were calculated for each women, and the prevalence of osteoporosis (i.e., T-scores ≤ −2.5) [14] was estimated and shown in Table 3. Approximately 23% of women aged 50–65 years had osteoporosis at the femoral neck. The prevalence increased with advancing age such that among those aged between 50 and 59 it was estimated at 21% and increased to 33% in those aged 60–65 years or above. The calculation for osteoporosis at the lumbar spine was 49% for women aged 50–69.

Table 3 Prevalence (%) of osteoporosis by site of BMD measurement
Full size table

There was a discordance in the diagnosis of osteoporosis based on femoral neck and total hip T-scores. When the total hip BMD T-score was used, the prevalence was 26%. However, when either femoral neck or total hip BMD was used, the prevalence increased to 33.7%. Although the proportion of concordance in the diagnosis between the two BMD sites was 80%, the kappa statistic was 0.49 (95% confidence interval: 0.34 – 0.65, data not shown) which was statistically significantly different from 0 (p < 0.0001).

Discussion

BMD is an important predictor of future fracture [16–18] and individuals with osteoporosis are at increased risk [18]. It has been shown that 65% of individuals with osteoporosis will sustain a fracture within the next 10 years [19] or 15 years [20]. The present study suggests that approximately a quarter of Vietnamese women aged 50+ are at high risk of fracture. Thus, it seems that osteoporosis represents an important burden of public health in Vietnam.

It is interesting to note that using the same densitometer (GE Lunar) the peak BMD at the femoral neck in Vietnamese women was comparable to that of Korean [21, 22] and slightly higher than in Japanese [21], Chinese [23] and Indonesian women [24] (Table 4). However, given the sampling variation and the lack of calibration between densitometers the differences among Asian populations seem insignificant. The present estimated prevalence is also equivalent to that of in American Caucasian women [25, 26] even though the latter was based on Hologic DXA. Some of the differences in prevalence could be partially due to the difference in bone densitometry technologies. However, the effect of measurement technology is likely to be small because the prevalence was determined from BMD T-scores, and T-scores are standardized to mean and standard deviation obtained from each densitometer and each population. Notwithstanding, the finding rather suggests that the diagnosis of osteoporosis can be based on a global reference data base.

Table 4 Comparison of young adult bone mineral density
Full size table

However, when the Third National Health and Nutrition Examination Survey (NHANES III) reference data were used, the prevalence of osteoporosis in Vietnamese women was estimated to 60%, 31%, and 21% at the lumbar spine, total hip, and femoral neck, respectively. These estimates were slightly higher than those based on the Vietnamese-specific reference data, because largely of the difference in standard deviations between the two populations.

The present data suggest that Vietnamese women may achieve their peak BMD at a later age (between 28 and 29 years) than Caucasian women among whom the peak BMD was reached between 20–25 years [25, 26]. It is not possible to determine the underlying factors to explain this apparent difference. However, it is well-known that Asian girls tend to have a later menarche (average age of 13 years) than Caucasian girls (12 years). It is possible that nutritional factors and lower levels of estrogen and other sex steroids contribute to the difference. It is also noted that after reaching its peak, lumbar spine BMD starts declining. While this trend was also observed in other populations [27, 28] but not all [29], the reason for this decline is not known. It is possible that this decline reflects a secular trend among segments of the premenopausal population. However, this issue could not be addressed in this study as there is a lack of prospective data (Table 5).

Table 5 Comparison of prevalence of osteoporosis among women aged 50 + years
Full size table

It is likely that the prevalence of osteoporosis was underestimated in this study, because there were no women aged above 65 years in the study. Even so, the prevalence (between 23% and 26%) is slightly higher than in Australian (21%) [30] and American (20%) [31] but lower than in Thai (29%) [32] and Hong Kong women [17]. However, the prevalence of osteoporosis in the present study appears to be substantially higher than Chinese (10%,) [33], Japanese (17%) [34], Korean (10%) [21] and Indonesian women (15%) [24]. It is interesting to note that even though the peak bone mass of Vietnamese women was found comparable to that of Caucasian and other Asian girls, the prevalence of osteoporosis among women aged 50–65 was still higher than in many other countries. One possible explanation could be that of cohort effect: women aged 50+ were born and grew up during many years of war, when the country's economy was in great hardship and many women suffered from malnutrition which accumulatively resulted in low BMD. In contrast, the younger cohort aged between 20 and 30 years was born and raised after the independence in 1975, during which economic and nutritional conditions have significantly been better than that during the war years.

The present results should be interpreted in relation to some strengths and weaknesses. The sample was randomly drawn from the general population after careful screening, which should support its representativity and external validity. The sample size was statistically adequate to estimate a prevalence of osteoporosis as 30% with a marginal error of 3%. The use of DXA technology with reliable measurement of BMD is also a strength of this study. However, the design was cross-sectional and therefore the estimate of peak BMD and age of peak BMD could be subject to uncontrolled bias. Ideally, these values should be determined in a representative cohort which should be followed from 5 years to 30 years of age. However, it is unlikely that such a study is feasible; therefore, a cross-sectional study is an alternative design. The estimate of osteoporosis prevalence in post-menopausal women was based on a rather small sample size, particularly among those aged 60–65 (n = 24), which could result in an increased sampling error. Because the maximum age of participants in this study was 65, the estimate of prevalence was likely underestimate.

In summary, this study has, for the first time, provided an estimation of peak BMD in Vietnamese women. The peak bone mass in this study was found comparable to global reference data but the prevalence of osteoporosis was higher than in some other Asian and Caucasian populations. It seems that osteoporosis is an important public health burden in Vietnam.