This cross-sectional study was based on data acquired from the KNHANES in 2008, 2009, 2010, and 2011. KNHANES is a population-based nationwide survey conducted by Korea Centers for Disease Control and Prevention (IRB no.: 2008-04EXP-01-C, 2009-01CON-03-2C, 2010-02CON-21-C, and 2011-02CON-06-C). BMD was assessed in data collected from July 2008 to May 2011. A total of 21,071 subjects had available BMD measurements. A total of 4,991 postmenopausal women aged 45 years or older were included in the present analysis. Individuals who had a history of osteoporosis medication use, severe chronic kidney disease (estimated glomerular filtration rate [GFR] < 30), artificial menopause such as hysterectomy, and early menopause (age < 30) were excluded. Subjects were also excluded if they did not have adequate data for BMD or blood samples. Finally, 3,058 postmenopausal women were analyzed in this study.

BMD was measured in the lumbar spine, total femur, and femoral neck using dual-energy X-ray absorptiometry (DXA; QDR 4500A, Hologic Inc., Waltham, MA, USA). DXA instruments were calibrated as described previously [15], and reference values were obtained using this method. DXA calibrations were maintained via an internal referencing system [16]. Coefficients of variation for repeated measurements of < 1.9% for the lumbar spine, < 1.8% for the total femur, and < 2.5% for the femoral neck satisfied the precision criteria. BMD was analyzed using standard techniques by the Korean Society of Osteoporosis and Hologic Discovery software (version 13.1). Osteoporosis was defined as a T score ≤ –2.5 standard deviation according to the World Health Organization criteria [17].

Body height was measured at corrected posture to the nearest 0.1 cm. Body weight was measured in light clothes after zero correction to the nearest 0.1 kg. Body mass index (BMI) was calculated as weight divided by height squared. Blood pressure (BP) was measured using a mercury sphygmomanometer (Baumanometer, W.A. Baum, Copiague, NY, USA). Blood samples were collected and transported to the Neodin Medical Institute in Seoul, Korea. Total cholesterol was measured from fasted blood samples with a Hitachi Automatic Analyzer 7600 (Hitachi, Tokyo, Japan). The levels of serum parathyroid hormone (PTH) were measured by chemiluminescence immunoassay (Diasorin Inc., Stillwater, MN, USA), and serum vitamin D was measured by radioimmunoassay (Diasorin) with a gamma counter (1470 Wizard, PerkinElmer, Turku, Finland). GFR was calculated using the Modification of Diet in Renal Disease formula. The KNHANES consisted of a health interview survey, a health examination survey, and a nutrition survey. Personal medical data, which included the number of household members and area of residence, were collected using a questionnaire. Residential area was classified as urban and rural based on the administrative district. Menopause was defined as natural menopause without a history of hysterectomy or early menopause. Diabetes and hypertension (HTN) were defined in participants who had already been diagnosed. The health interview survey included the International Physical Activity Questionnaire to obtain information about smoking status, drinking status, and physical activity. Smoking status was classified as current smoker, ex-smoker, and non-smoker. Heavy alcoholics were defined as people who consumed ≥ 30 g of alcohol per day. Physical activity was classified as high physical activity for at least 20 minutes for 3 days per week, moderate activity for at least 30 minutes for 5 days per week, and walking activity for 30 minutes for 5 days per week. Dietary intake of food, total energy, protein, fat, and calcium was estimated using the 24 hours recall method and a food frequency questionnaire in the nutrition survey.

All data were presented as mean ± standard error for continuous data and numbers and percentages for categorical data. Statistical analysis was performed using SPSS version 18.0 (SPSS Inc., Chicago, IL, USA). Clinical and demographic variables of subject groups were analyzed by Student t test and chi-square test. Because the KNHANES was designed as a complex sample survey, we estimated BMD according to household size and residential area using a complex samples general linear model. The models were adjusted for age, BMI, number of pregnancies, menopause duration, systolic BP, PTH, vitamin D, HTN, thyroid disease, working hours, physical activity, heavy alcoholics, smoking status, daily food intake, and nutrient support. Subjects were divided into four groups according to household size and residential area, and baseline characteristics were compared using analysis of variance for continuous variables and chi-square tests for categorical variables. The odds ratio (OR) of osteoporosis and fracture was assessed in four groups using the complex samples logistic regression model. A p < 0.05 was considered statistically significant.

The general characteristics of the participants are presented in Table 1. A total of 969 women lived in a rural area (average age, 65.2 years) and the remaining 2,089 women lived in an urban area (average age, 62.4 years). Rural residents had higher systolic BP and vitamin D levels, while BMI, diastolic BP, fasting glucose, total cholesterol, and PTH levels did not differ between the two groups. Although there was no difference in total energy intake between the two groups, daily food, protein, fat, and calcium intakes were lower in rural residents. Urban residents worked fewer hours and included more current smokers. Physical activity and the number of heavy alcoholics did not differ between the two groups. The number of people per household was lower for rural residents, and the proportion of single-person households was 19.4% for rural residents and 13.2% for urban residents. The prevalence of HTN was higher in rural residents, while the prevalence of thyroid disease was higher in urban residents. Anti-hypertensive drug use, the diabetes mellitus prevalence, and cancer history were similar between the two groups. The proportion of people with osteoporosis was significantly higher in rural residents, but the difference in the proportion of people with vertebral fracture was not significant.

Participants living in a rural area had a significantly lower BMD at the lumbar spine than those living in an urban area (Fig. 1A), whereas BMD in the total femur and femoral neck were not significantly different (Fig. 1B and 1C). The lumbar spine BMD of rural residents remained significantly lower than that of other residents after adjustment for multiple confounding factors (Fig. 1A). The prevalence of lumbar spine osteoporosis was significantly higher in rural residents in the unadjusted model (Fig. 2A). There was no significance after adjustment of confounding factors in this association. Although the OR of vertebral fracture was also higher in rural residents than in urban residents, this was not statistically significant (Fig. 2B).

We divided the subjects into the following four groups (Table 2): (1) single-person households in a rural area (single in rural, n = 194); (2) households with two or more people in a rural area (two-more in rural, n = 775); (3) single-person households in an urban area (single in urban, n = 297); and (4) households with two or more people in an urban area (two-more in urban, n = 1,792). The mean age in the two-more in urban group was lower than in the other groups. BMI was lowest in the single in rural groups. Systolic BP was highest in the single urban group and lowest in the two-more in urban group. Diastolic BP, fasting glucose, and total cholesterol were not statistically different among the groups. Vitamin D and PTH levels tended to be highest in the single in rural group. Subjects in the single in rural group reached menopause at a younger age and had more pregnancies. Food, total energy, protein, fat, and calcium intakes were highest in the two-more in urban group and lowest in the single in rural group. Working hours tended to be higher in rural residents and lower in single-person households. The single in urban group contained more smokers, people with HTN, and anti-hypertensive drug users than the other groups. Physical activity, alcohol status, prevalence of diabetes mellitus, and history of cancer were not significantly different among the groups. The prevalence of thyroid disease was significantly lower in the single in rural group. Osteoporosis prevalence was highest in the single in rural group and lowest in the two-more in urban group. The prevalence of vertebral fracture was higher in the single in rural group, although this was not statistically significant.

BMDs at all sites were significantly lower in the single in rural group than in the two-more in urban group (Table 3). After adjustment for age, BMI, number of pregnancies, duration of menopause, systolic BP, PTH, vitamin D levels, HTN, thyroid disease, working hours, physical activity, heavy alcoholics, smoking status, daily food intake, and nutrient support, lumbar spine BMD was still significantly lower in the single in rural group than in the two-more in urban group. Total femur and femoral neck BMDs were not significantly different after adjustment for potential confounding factors.

Compared with the two-more in urban group, the OR of all the other groups for lumbar spine osteoporosis was significantly higher in the unadjusted model (Table 4). After adjustment for multiple confounding factors, the prevalence of lumbar spine osteoporosis did not differ between the two-more in rural and the single in urban groups. However, subjects in the single in rural group had the highest odds of osteoporosis in the lumbar spine (OR, 1.667; 95% CI, 1.083 to 2.565). Although the OR of vertebral fracture was higher in the two-more in rural group than in the two-more in urban group in the unadjusted model, the prevalence of vertebral fracture was correlated with neither household size nor residential area in the adjusted model.