Women's Health Initiative Calcium/Vitamin D Randomized Trial
Women's Health Initiative Calcium/Vitamin D Randomized Trial
Of the 600 women in the CaD trial subsample, 24 were excluded from the analyses: 1 participant did not have valid 25OHD3 data and the remaining 23 had their visit 1 blood drawn after CaD randomization (Figure 1). Among the 576 women with valid data, 291 (50.5%) were receiving CaD and 285 (49.5%) were receiving placebo. The mean (SD) age at baseline was 61.8 (6.7) years. The women, on average, were 14.6 (9.7) years from menopause. The mean (SD) BMI and waist circumference were 30.5 (6.6) kg/m and 90.7 (13.9) cm, respectively. The mean (SD) systolic blood pressure was 130.1 (16.5) mm Hg, and the mean (SD) LDL-C, HDL-C, and TG levels were 124.5 (33.1) mg/dL, 58.6 (15.0) mg/dL, and 149.2 (71.2) mg/dL, respectively. Demographics and comparisons between treatment and placebo groups are presented in Table 1 . No significant differences in baseline characteristics were seen between those on CaD and those on placebo.
Because 25OHD3 is the exposure variable of interest, Table 2 presents the variables associated with 25OHD3. Analyzing baseline characteristics by LDL-C concentrations before randomization (broken into quartiles), we identified several findings across the range of 25OHD3 concentrations. For example, 25OHD3 concentration was inversely associated with being a current smoker (P < 0.001). African-American race was a risk factor for decreased 25OHD3, white race was a risk factor for increased 25OHD3, and Hispanic ethnicity had no influence. Concentrations of 25OHD3 were negatively associated with BMI (P < 0.001) but positively associated with expenditure of energy from recreational physical activity (P < 0.001). Table 3 presents the baseline characteristics of women in the CaD trial subsample by quartiles of LDL-C (mg/dL) at year 1.
Women on CaD had significantly increased mean (95% CI) postintervention 25OHD3 concentrations compared with placebo (24.3 [22.9-25.7] vs 18.2 [17.1-19.3] ng/mL, respectively; P < 0.001; Table 4 ). The mean (95% CI) postintervention 25OHD3 concentration in women on CaD was 1.38 (1.29-1.47) times higher than that in women on placebo after covariate adjustment. Women taking CaD supplementation were more than twice as likely (risk ratio, 2.35; 95% CI, 1.71-3.22; P < 0.001) to have 25OHD3 concentrations of 30 ng/mL or higher: 35.4% in the intervention arm versus 15.1% in the control arm. Similarly, women taking CaD supplementation were 1.58 times more likely (95% CI, 1.38-1.82; P < 0.001) to have 25OHD3 concentrations of 20 ng/mL or higher: 75.6% in the intervention arm versus 47.7% in the control arm. The effect of CaD on 25OHD3 concentration was not modified by covariates such as age, race/ethnicity, BMI, smoking, alcohol consumption, physical activity, HT use, or seasonal temporality ( Table 4 ).
Women randomized to CaD had a 4.46-mg/dL decrease in LDL-C (95% CI, 0.41-8.51) compared with placebo (P=0.03; Figure 3A, path c; Table 5 ). There was a nonsignificant increase in HDL-C (P = 0.82) and a nonsignificant decrease in TG (P = 0.21). When 25OHD3 concentration was included in the model, the effect of CaD on LDL-C was attenuated to a 3.24-mg/dL decrease and was no longer significant (P=0.13). Instead, 25OHD3 concentration was a significant predictor of LDL-C (P = 0.04), where a 38% increase in 25OHD3 was associated with a 1.28 -mg/dL decrease in LDL-C (P=0.04). The multiple imputation analysis yielded similar results where the effect of CaD was attenuated and no longer significant (P = 0.17), whereas 25OHD3 was significantly associated with LDL-C (P=0.01).
Modeling associations at both prerandomization and postrandomization visits and serum 25OHD3 concentrations were significantly associated with all three cholesterol parameters (Figure 4). More specifically, higher 25OHD3 concentrations were associated with higher HDL-C levels (P = 0.003). Lower TG was also associated with higher 25OHD3 (P < 0.001), but it seemed that a certain threshold value of 25OHD3 was needed (approximately 15 ng/mL) before this association became evident. Similarly, higher 25OHD3 was associated with lower LDL-C levels (P = 0.02). The associations between 25OHD3 and lipids were not modified by visit year (P interaction > 0.10) or treatment assignment (P interaction > 0.10).
(Enlarge Image)
Figure 4.
Multivariable-adjusted generalized additive mixed model (GAMM) estimates of the mean LDL-C, HDL-C, and triglyceride (95% CI; shaded region) levels as a smoothed function of 25OHD3. Lipid and 25OHD3 measurements at years 1 and 3 were included in GAMMs to incorporate both cross-sectional and longitudinal information. GAMMs included an indicator variable for pre/post-CaD randomization visit and were adjusted for age, race/ethnicity, body mass index, smoking status, history of high blood cholesterol, diabetes mellitus, prior HT use, physical activity, total calcium intake, HT randomization arm, DM randomization arm, CaD randomization arm, and CaD randomizationYvisit interaction. A random intercept was also included to account for within-participant correlation. The smoothness of each spline fit was chosen objectively by generalized cross-validation. CaD, calcium/vitamin D; HT, hormone therapy; LDL-C, low-density lipoprotein cholesterol; 25OHD3, 25-hydroxyvitamin D3; HDL-C, high-density lipoprotein cholesterol.
Results
Baseline Characteristics
Of the 600 women in the CaD trial subsample, 24 were excluded from the analyses: 1 participant did not have valid 25OHD3 data and the remaining 23 had their visit 1 blood drawn after CaD randomization (Figure 1). Among the 576 women with valid data, 291 (50.5%) were receiving CaD and 285 (49.5%) were receiving placebo. The mean (SD) age at baseline was 61.8 (6.7) years. The women, on average, were 14.6 (9.7) years from menopause. The mean (SD) BMI and waist circumference were 30.5 (6.6) kg/m and 90.7 (13.9) cm, respectively. The mean (SD) systolic blood pressure was 130.1 (16.5) mm Hg, and the mean (SD) LDL-C, HDL-C, and TG levels were 124.5 (33.1) mg/dL, 58.6 (15.0) mg/dL, and 149.2 (71.2) mg/dL, respectively. Demographics and comparisons between treatment and placebo groups are presented in Table 1 . No significant differences in baseline characteristics were seen between those on CaD and those on placebo.
Because 25OHD3 is the exposure variable of interest, Table 2 presents the variables associated with 25OHD3. Analyzing baseline characteristics by LDL-C concentrations before randomization (broken into quartiles), we identified several findings across the range of 25OHD3 concentrations. For example, 25OHD3 concentration was inversely associated with being a current smoker (P < 0.001). African-American race was a risk factor for decreased 25OHD3, white race was a risk factor for increased 25OHD3, and Hispanic ethnicity had no influence. Concentrations of 25OHD3 were negatively associated with BMI (P < 0.001) but positively associated with expenditure of energy from recreational physical activity (P < 0.001). Table 3 presents the baseline characteristics of women in the CaD trial subsample by quartiles of LDL-C (mg/dL) at year 1.
Effects of CaD on 25OHD3 Concentrations
Women on CaD had significantly increased mean (95% CI) postintervention 25OHD3 concentrations compared with placebo (24.3 [22.9-25.7] vs 18.2 [17.1-19.3] ng/mL, respectively; P < 0.001; Table 4 ). The mean (95% CI) postintervention 25OHD3 concentration in women on CaD was 1.38 (1.29-1.47) times higher than that in women on placebo after covariate adjustment. Women taking CaD supplementation were more than twice as likely (risk ratio, 2.35; 95% CI, 1.71-3.22; P < 0.001) to have 25OHD3 concentrations of 30 ng/mL or higher: 35.4% in the intervention arm versus 15.1% in the control arm. Similarly, women taking CaD supplementation were 1.58 times more likely (95% CI, 1.38-1.82; P < 0.001) to have 25OHD3 concentrations of 20 ng/mL or higher: 75.6% in the intervention arm versus 47.7% in the control arm. The effect of CaD on 25OHD3 concentration was not modified by covariates such as age, race/ethnicity, BMI, smoking, alcohol consumption, physical activity, HT use, or seasonal temporality ( Table 4 ).
Effects of CaD on Lipids
Women randomized to CaD had a 4.46-mg/dL decrease in LDL-C (95% CI, 0.41-8.51) compared with placebo (P=0.03; Figure 3A, path c; Table 5 ). There was a nonsignificant increase in HDL-C (P = 0.82) and a nonsignificant decrease in TG (P = 0.21). When 25OHD3 concentration was included in the model, the effect of CaD on LDL-C was attenuated to a 3.24-mg/dL decrease and was no longer significant (P=0.13). Instead, 25OHD3 concentration was a significant predictor of LDL-C (P = 0.04), where a 38% increase in 25OHD3 was associated with a 1.28 -mg/dL decrease in LDL-C (P=0.04). The multiple imputation analysis yielded similar results where the effect of CaD was attenuated and no longer significant (P = 0.17), whereas 25OHD3 was significantly associated with LDL-C (P=0.01).
Association Between 25OHD3 and Lipids
Modeling associations at both prerandomization and postrandomization visits and serum 25OHD3 concentrations were significantly associated with all three cholesterol parameters (Figure 4). More specifically, higher 25OHD3 concentrations were associated with higher HDL-C levels (P = 0.003). Lower TG was also associated with higher 25OHD3 (P < 0.001), but it seemed that a certain threshold value of 25OHD3 was needed (approximately 15 ng/mL) before this association became evident. Similarly, higher 25OHD3 was associated with lower LDL-C levels (P = 0.02). The associations between 25OHD3 and lipids were not modified by visit year (P interaction > 0.10) or treatment assignment (P interaction > 0.10).
(Enlarge Image)
Figure 4.
Multivariable-adjusted generalized additive mixed model (GAMM) estimates of the mean LDL-C, HDL-C, and triglyceride (95% CI; shaded region) levels as a smoothed function of 25OHD3. Lipid and 25OHD3 measurements at years 1 and 3 were included in GAMMs to incorporate both cross-sectional and longitudinal information. GAMMs included an indicator variable for pre/post-CaD randomization visit and were adjusted for age, race/ethnicity, body mass index, smoking status, history of high blood cholesterol, diabetes mellitus, prior HT use, physical activity, total calcium intake, HT randomization arm, DM randomization arm, CaD randomization arm, and CaD randomizationYvisit interaction. A random intercept was also included to account for within-participant correlation. The smoothness of each spline fit was chosen objectively by generalized cross-validation. CaD, calcium/vitamin D; HT, hormone therapy; LDL-C, low-density lipoprotein cholesterol; 25OHD3, 25-hydroxyvitamin D3; HDL-C, high-density lipoprotein cholesterol.