Study subjects were postmenopausal Korean women who have voluntarily undergone a routine health examination repeatedly with an interval of 6 to 8 years in a health promotion center of a tertiary care hospital located in Seoul after taking an initial health examination between January 2001 and December 2003. Among the initially identified 1,421 women, 463 women were excluded due to the following reasons: taking lipid-lowering drugs (220 women), hormone supplements (116 women), or both lipid-lowering drugs and hormone supplements (4 women) at final follow-up health examination; having higher free T4 level (>2.0 ng/dL, cut-off level which is known to affect bone metabolism)²⁵⁾ (26 women); information on smoking, drinking, and sports activity was missing (97 women). Thus, a total of 958 women were finally included.

Blood sample was drawn after 12-hour overnight fasting on the same day BMD was measured. Total cholesterol (TC), HDL-C, and LDL-C were assayed enzymatically in fresh serum in a laboratory. BMD (g/cm²) was measured at left femur neck and at lumbar spines (2nd, 3rd, and 4th) using dual-energy X-ray absorptiometry (Lunar Radiation Co., Madison, WI, USA). For analysis of lumbar spine, we used the averaged BMD values measured at 2nd, 3rd, and 4th level of lumbar spine. Height (cm) and weight (kg) were measured to the nearest 0.1 decimal point using a digital scale and stadiometer in light clothing and shoes taken off. Body mass index (BMI) was calculated as the weight divided by the height squared (kg/m²).

Information about health behaviors (smoking, alcohol consumption, and physical activity) was assessed using a self-administered questionnaire. Study subjects were categorized into three groups by smoking habit (never smoker, ex-smoker, and current smoker), four groups by the level of physical activity ('do not exercise at all,' 'exercise less than once a week,' 'exercise 1-4 times a week,' and 'exercise more than 5 times a week'), and four groups by the levels of alcohol consumption ('never drinks,' 'drinks less than a bottle in a month,' 'drinks less than 2 bottles a month,' and 'drinks more than two bottles a month'). This study was approved by the institutional review board of Samsung Medical Center.

Changes in the level of lipid profiles and BMD were calculated by subtracting the value at initial measurement from the value at final follow-up measurement. Because the distribution of lipid profiles and BMD measured at initial and final examination and the distribution of change in lipid profiles and BMD during follow-up did not follow normal distribution, we estimated the association between lipid profiles and BMD using Spearman correlation analysis with or without an adjustment for covariates such as age, BMI, smoking, alcohol consumption, and physical activity. At first, we evaluated the association between the lipid profiles and BMD measured at initial examination with a consideration of age and BMI measured at initial examination, and information on smoking, alcohol consumption, and physical activity measured at follow-up examination as covariates. Then, we evaluated the association between the lipid profiles and BMD measured at final follow-up examination with a consideration of age and BMI measured at follow-up examination, and information on smoking, alcohol consumption, and physical activity measured at follow-up examination as covariates. Finally, we evaluate the association between the changes in lipid profiles and the changes in BMD, with a consideration of age at initial examination, duration of follow-up, change in BMI during follow-up, information on smoking, alcohol consumption, and physical activity measured at follow-up examination as covariates. All the analyses were done using SAS ver. 9.1.3 (SAS Institute Inc., Cary, NC, USA). A two-sided test was done with the level of significance set at 'less than 0.05'.

The mean (SD) interval between the time at initial and time at final examination was 7.1 (0.8) years. Mean (SD) age of study subjects at the initial examination was 58.6 (5.8) years (Table 1). During follow-up, mean (SD) height (cm) changed from 156.3 (4.9) to 156.0 (5.0) and mean BMI (kg/m²) shifted from 23.8 (2.7) to 23.8 (2.8). Total cholesterol (mg/L) level changed from 209.4 (33.3) to 204.3 (34.3), LDL-C (mg/L) from 140.3 (31.6) to 126.8 (30.2), HDL-C (mg/L) from 58.7 (14.1) to 57.0 (30.2), which was not a significant change. The BMD (g/cm²) at lumbar spine changed from 1.12 to 1.01 and the BMD at femur neck changed from 0.86 to 0.80, which was also not significant.

The lumbar spine BMD measured initially had no significant correlation with total cholesterol, LDL-C, HDL-C. This finding did not change after the adjustment of age, smoking, drinking, exercise, and BMI (Table 2). The femur neck BMD measured initially had no significant correlation with total cholesterol and LDL-C, whereas it had a weak positive correlation with HDL-C (correlation coefficient, 0.092; 95% confidence interval, 0.019 to 0.163). The significant correlation between femur neck BMD and HDL-C did not change after the adjustment of covariates (partial correlation coefficient, 0.077; 95% confidence interval, 0.005 to 0.149). There was no significant correlation between the lipid profiles and BMDs (at lumbar spine and femur neck) measured at final follow-up examination regardless of the adjustment for covariates (Table 3).

The change in BMD at lumbar spine had no significant correlation with the change in total cholesterol, LDL-C, and HDL-C level regardless of the adjustment for age, smoking, drinking, exercise, and BMI (Table 4). When covariates were not considered in analysis, change in femur neck BMD had weak but significant negative correlation with total cholesterol (correlation coefficient, -0.089; 95% confidence interval, -0.161 to 0.017) and LDL-C (correlation coefficient, -0.087; 95% confidence interval, -0.158 to -0.014). However, the correlation did not persist after the adjustment for covariates.