Descriptive characteristics of the women as well as inflammatory marker values and bone parameters at baseline are presented in . Median values for inflammatory markers were within the range reported in the literature. However, 65 women had low WBC (<4.5 but >2.3 × 109/L) and none of the women had elevated WBC. Fifty women (20.6%) had non-detectable values for IL-1β and four women (1.7%) had non-detectable values for IL-6. Hence, we replaced these nondetectable values with 0.01 pg/mL (lowest detectable value was 0.01 for IL-1β and 0.03 for IL-6) to retain all data in the regression models subsequent to log transformation for regression analysis. Pearson correlation coefficients were used to assess the relationship of inflammatory markers with age, height, weight, TSLM, and 1 year percent change in BMC and BMD for femoral neck, hip, LS, distal tibia, and whole body; correlations are presented in . Only variables that had a significant correlation with bone parameters are shown. IL-6 and TNF-α exhibited significant associations with percent change in BMC and BMD across a majority of bone sites. WBC, a marker indicative of infection and which may be a surrogate marker of inflammation, was significantly associated with percent change in BMC at 2 sites: hip and femoral neck. No significant associations were found among any inflammatory marker and pQCT measures of BMC or BMD.
Descriptive Statistics of Subjects at Baseline
Associations Among 1-Year Changes in Bone Mineral Content and Density and Inflammatory Markers
In the parent project lumbar spine, hip or whole body showed no significant effect of soy isoflavones on BMD except for a modest effect on percent change in femoral next BMD with 120 mg/d soy isoflavones. This positive effect, however, was observed only after controlling for age, whole body fat mass and serum C-Tx, a marker of bone resorption. So, to better understand the possible contribution of inflammatory markers to changes in bone, we developed models using stepwise multiple regression analysis to evaluate their combined contribution to the percent change in BMC or BMD at the hip, LS, femoral neck, trochanter, whole body, and distal tibia (). Variables used in the regression models included: age, BMI, TSLM, CRP, IL-1β, IL-6, TNF-α, and WBC. Site was included in all models as an obligatory variable to account for potential differences across sites. With the exception of the femoral neck, a combination of inflammatory markers contributed to the 12 month percent change in BMC or BMD at the hip, LS, trochanter, and whole body. CRP, an acute phase protein, was a significant contributor to the percent change in BMC or BMD at the LS, whole body, and trochanter. TNF-α and IL-1β were significant contributors to the percent change at the hip and WBC was a significant contributor at the trochanter site. Overall, the combined contribution of the inflammatory markers accounted for 1.1 to 6.1% of the variance (R2) in the regression models for the percent change in BMC or BMD at these specific sites. As expected, based on Pearson correlations, no combination of inflammatory markers was found to contribute significantly to the 1-year changes in BMC or BMD measured by pQCT.
Stepwise Regression Models for Contributions of Inflammatory Markers to 1-Year Change in BMC and BMD at Specific Sites
In summary, using multiple regression modeling, we found that markers of inflammation made small, but important contributions, ranging from 1.1 to 6.1% of the variance, to the 1-year percent change in BMC or BMD across a variety of bone sites assessed by DXA.