The effects of the Korean reference value on the prevalence of osteoporosis and the prediction of fracture risk

We used the BMD data from the Korea National Health and Nutrition Examination Survey (KNHANES; 2008–2011). This survey is administered nationwide and has been conducted since 1995 in South Korea by the Korea Centers for Disease Control and Prevention. The KNHANES from 2008–2009 was administered to 4,600 households from 200 primary sampling units, while the KNHANES from 2010–2011 was administered to 3,800 households from 192 primary sampling units each year. First, the primary units were selected by the administrative regions; one study unit was extracted from each area after considering the type of house. Second, 20 households from each study unit were extracted using a systematic sampling method, and these households were the basic units in the survey. By using the rolling survey sampling, a sample from each year was representative of the population of that respective year.

The KNHANES was composed of a health interview survey, nutritional survey, and a health examination survey. The health interview and health examination surveys were administered in a mobile examination center, and the nutritional survey was administered by direct visit and interview. The health interview survey was administered via a direct interview when the subjects visited the mobile center; however, alcohol use and current smoking status were surveyed by using a self-reporting questionnaire. The health examination survey was analyzed by using measurements, observation, and laboratory sampling.

The subjects were men and women who received the test for BMD during the KNHANES from 2008–2011 [ 20, 21]. There was no significant difference between subjects who did not receive a BMD scan and those who did; thus, the risk of selection bias was very low. The subjects used for investigating the Korean reference value were 20–29-year-old males and females. The analysis comparing the prevalence and fracture risk probability was conducted to target men and postmenopausal women >50 years. Menopause was determined by the subjects’ responses to the following survey items: (1) those who answered natural menopause or menopause-related procedure to the question, why do you not menstruate?; (2) when the subjects’ period since the last menstruation was >1 year; and (3) those who answered yes or bilateral ovariectomy to menopause status. Women who satisfied all three conditions were considered postmenopausal. Postmenopausal women from 50–64 years and women >65 years were included in the analysis of the prevalence of osteoporosis and fracture risk probability. The BMD data from the NHANES III were used to compare the BMD of white women from the United States NHANES III to the value of men and women from the KNHANES [ 22]. The content and methodology of the KNHANES was approved by the institutional review board of the Korea Centers for Disease Control and Prevention, and all the participants signed informed consent.

The BMD test was performed by a trained radiographer in a mobile examination center using dual-energy radiography absorptiometry (DXA; DISCOVERY-W fan-beam densitometer; Hologic, Bedford, MA, USA) of the lumbar spine (L1–L4) and proximal hip (femoral neck and total femur). Hologic Discovery software, version 3.1 was used for that device. The accuracy was investigated by performing double scans on 30 randomly selected subjects, and the accuracy was maintained within the range of 0.73–1.07% for the lumbar spine, 1.20–2.14% for the femoral neck, and 0.71–1.18% for the total hip. Daily calibration was performed by a phantom served by the company.

In this study, we analyzed the BMD of the femoral neck. The reference value for Koreans was set to the mean BMD and its standard deviation of men and women aged 20–29 years. We used the mean BMD and standard deviation (SD) (0.858 ± 0.120 g/cm ²) published in a previous study as the NHANES III reference [ 23]. Osteoporosis, osteopenia, and normal were defined using the following T-scores according to the WHO criteria: ≤ − 2.5, −1.0 to −2.5, and −1.0, respectively.

The 10-year fracture risk was calculated by the Korea FRAX™ version to assess how the change in the reference affects the relationship between BMD and fracture risk prediction [ 24]. The BMD that corresponds to each T-score −1.5, −2, −2.5, −3.0, −3.5, and −4.0) will vary depending on the reference value. For example, a T-score of −2.5 corresponded to 0.5128 g/cm ² when using a Korean women reference and to 0.5553 g/cm ² when using a NHANES III reference. When these different BMDs were used in FRAX™, the fracture risk probability was calculated as 6% and 6.8%, respectively. For weight and height, the mean values of each age and sex group were used. The difference of fracture risk probability by the references was analyzed by age group and by the number of clinical risk factors for fracture. To simplify the analysis, only a history of fracture and a family history of hip fracture in a parent of two estimates were analyzed.

Statistical analyses were performed using SPSS for Windows, version 21.0 (SPSS Inc., Chicago, IL, USA). Continuous variables were described in the form of mean ± SD; the prevalence was presented as a percentage (%). Firstly, the subjects were partitioned into 10-year age groups, and the mean and SD of BMD of each age group was calculated. The mean BMDs did not satisfy the assumption of equal variance, so the Welch’s test and Tamhane’s T2 post-hoc analysis were used to test the significance of the difference between the mean BMD by age group. To test whether there was a significant difference between the prevalence calculated by the Korean reference and by the NHANES III reference, the McNemar–Bowker test and Cohen’s kappa coefficient were used to evaluate the degree of diagnostic agreement.

The basic characteristics of the subjects are summarized in Table  1. The mean BMD, height, and weight decreased as the subjects’ age increased. In Table  2, the mean and standard deviation of BMD of each age group was compared with the Korean BMD data and the NHANES III BMD data. The mean BMD of Korean men, women, and women in the NHANES III decreased with increasing age. There was a significant difference between the mean BMD of the 20–29-year-old age group and that of another age group in the KNHANES 2008–2011 (Table  3). However, in the NHANES III data, there was no significant difference between those in their 20s and 30s. Table  3 and Figure  1 shows the decrease in the mean BMD with increasing age by the reference groups. The mean BMD was identified from highest to lowest across all age groups as follows: Korean men > NHANES III women > Korean women.

If an agreement was found in the diagnosis of osteoporosis based on each reference value, it was represented by the sex and age group in Tables  4 and 5. For men, the results of the diagnosis were calculated using the Korean reference and the NHANES III women reference, which were fairly consistent. It was estimated by age groups as follows: 50s: k = 0.668, concordance 83.3%; 60s: k = 0.661, concordance 81.1%; 70s: k = 0.654, concordance 81.4%; and 80s: k = 0.633, concordance 78.1%. The concordance rate was slightly decreased as age increased. Despite a high concordance, there were statistically significant differences between the diagnosis and each reference (McNemar-Bowker test, p < 0.001). A significant difference was also found in women; however, the degree of agreement was lower than that of men. It was estimated by age groups as follows: 50s: k = 0.457, concordance 69.5%; 60s: k = 0.446, concordance 71.1%; 70s: k = 0.553, concordance 75%; and 80s: k = 0.565, concordance 82%. While the difference in the diagnosis was increased in aging men, the difference in the diagnosis in women was decreased with an increasing age. However, with regard to osteoporosis, the difference in the prevalence of osteoporosis was increased in both sexes, except in women aged >80 years.

Figure  2 compares the prevalence of osteoporosis and osteopenia according to the references. In men, when the Korean reference value was used, the prevalence of osteoporosis increased. In Figure  2 and Table  4, the difference in osteoporosis in men increased with age and reached 9% for those in their 80s. In women (Figure  2 and Table  5), the prevalence of osteoporosis was higher when the NHANES III white women reference was used. The differences in osteoporosis were 6%, 13%, 19%, and 17% for those in their 50s, 60s, 70s, and 80s, respectively.

Figure  3 and Table  6 show the difference in the 10-year major osteoporotic fracture probability. Figure  4 and Table  7 show the difference in the 10-year femoral neck fracture probability according to the T-score levels and age groups. The results are summarized as follows.

Firstly, the differences in fracture risk were greater in women than in men. For both major osteoporotic fractures and femoral neck fractures, the difference of probability in men was <2%. However, the difference was higher in women. For major osteoporotic fractures, the highest difference identified was 6% for those in their 50s with a T-score of −4.0 and one or two clinical risk factors, and for those in their 70s with T-scores of −2.5 or −3.0 and two clinical risk factors. For femoral neck fractures, the highest difference was 7% for those in their 50s with a T-score of −4.0 and one or two clinical risk factors.

Secondly, the difference showed a tendency to increase as the number of clinical risk factors increased. For major osteoporotic fractures, this was most apparent in men aged 80–89 years with a T-score of −2.5; the difference of 0.3% with no clinical risk factor increased up to 1.7% with two clinical risk factors. In women aged 80–89 years, the difference of 1% with no clinical risk factor increased up to 5% with two clinical risk factors. For femoral neck fractures, the difference in men aged 80–89 years with no clinical risk factor increased from 0.2% to 2% with two clinical risk factors; for women in their 80s, the difference with no clinical risk factor increased from 1.2% to 6% with two clinical risk factors.

Finally, in women with a T-score ≤ −3.5 and none or one clinical risk factors, the difference in the major osteoporotic fracture risk decreased with age. However, in women with a T-score ≤ −3.0 and two clinical risk factors, the difference in the fracture risk increased with age. Table  6 shows data for women with a T-score −4.0 and one clinical risk factor, and the difference in the probability was 6%, 5%, 4%, and 2% for those in their 50s, 60s, 70s, and 80s, respectively. But the difference in women with T-score −2.5 and two risk factors increased from 3% to 5% in 2 risk factors with aging. The tendency was continued for femoral neck fracture in women with T-score ≥ −3.0 and two clinical risk factors, but the change in group with T ≤ −3.5 was slightly different little. The decreased tendency with aging can be seen in women with T-score between −2.5 to −4.0 and one clinical risk factor. Table  7 shows data for the femoral neck fractures in women with a T-score of −2 and two clinical risk factors, and the difference of probability increased from 1.5% for those in their 50s to 4% for those in their 80s. But, the difference in women with T-score −3.5 and one risk factor decreased from 4% to 1.3%.