Skip to main content
. 2010 Mar 10;91(5):1324–1335. doi: 10.3945/ajcn.2009.28908

TABLE 3.

Predictors of serum 25-hydroxyvitamin D [25(OH)D]: sample from the nested case-control studies of the Women's Health Initiative Calcium plus Vitamin D Clinical Trial (n = 3055)1

Reduced model2
 Final 25(OH)D predictive model3
No. of subjects β-coefficient (SE) P value Model R2 β-coefficient (SE) P value Partial R2 Model R2
Independent variable
 Month of blood draw 0.05 <0.0001 0.03 0.05
  January 260 Reference Reference Reference
  February 226 0.02 (2.04) 0.99 0.26 (1.96)
  March 244 0.99 (2.00) 0.62 1.93 (1.93)
  April 280 3.16 (1.93) 0.10 2.52 (1.89)
  May 230 3.77 (2.03) 0.06 2.32 (2.05)
  June 300 6.86 (1.90) 0.0003 7.60 (1.87)
  July 281 10.05 (1.93) <0.0001 10.37 (1.90)
  August 248 11.35 (1.99) <0.0001 12.15 (2.02)
  September 242 9.95 (2.00) <0.0001 10.27 (2.07)
  October 276 10.20 (1.94) <0.0001 10.23 (1.94)
  November 234 3.97 (2.02) 0.05 4.68 (1.99)
  December 234 4.46 (2.02) 0.03 5.22 (1.96)
 Total vitamin D intake, energy-adjusted (μg/d)4,5 2865 5.28 (0.34) <0.0001 0.12 5.05 (0.34) <0.0001 0.07 0.12
 Waist circumference (cm)4 3054 −0.40 (0.03) <0.0001 0.10 −0.36 (0.03) <0.0001 0.05 0.17
 Recreational physical activity (MET-h/wk)4 2731 0.32 (0.03) <0.0001 0.07 0.21 (0.03) <0.0001 0.01 0.19
 Race-ethnicity 0.07 <0.0001 0.01 0.20
  White 2726 Reference Reference Reference
  African American 164 −15.44 (1.79) <0.0001 −11.37 (1.92)
  Hispanic 76 −10.31 (2.58) <0.0001 −9.22 (2.70)
  Asian/Pacific Islander 51 −1.50 (3.13) 0.63 −8.07 (3.14)
  American Indian/unknown 38 −1.96 (3.62) 0.59 −2.68 (3.56)
 Langleys [(g-cal) per cm2] 3055 0.02 (0.007) 0.001 0.05 0.04 (0.007) <0.0001 0.01 0.21
 Age (y) 3055 2.75 (0.98) 0.005 2.59 (1.00) 0.01
 Age × age (y)6 3055 −0.02 (0.008) 0.004 0.05 −0.02 (0.008) 0.01 0.007 0.21
 Education4 0.05 —7
  ≤High school 741 Reference Reference
  School after high school 1182 0.25 (1.05) 0.81
  College degree or higher 1116 1.54 (1.07) 0.15
 Total time spent walking each week (h)4 2750 0.02 (0.005) <0.0001 0.05 —7
 History of diabetes5 3053 0.05 —7
  No Reference Reference
  Yes −7.73 (1.89) <0.0001
 Use of hormone therapy at year 1, including randomization 0.05 —7
  None 1047 Reference Reference
  Past use 469 2.18 (1.24) 0.08
  Current estrogen alone use 699 2.10 (1.10) 0.06
  Current estrogen plus progesterone use 840 4.49 (1.05) <0.0001
 Osteoporosis-related medication use8 0.04 0.05 —7
  No 2992 Reference
  Yes 63 5.98 (2.86)
 Use of antihypertensive medications 0.003 0.05 —7
  No 2680 Reference
  Yes 375 −3.69 (1.23)
Fully adjusted model R2 0.21
1

The predictive model was built by using a sample size of 3055 participants, but only 2472 were included in the final model because of missing data. MET-h, metabolic equivalent task hours.

2

Stepwise linear regression was implemented to develop the most predictive multivariate model for serum 25(OH)D concentrations (inclusion criteria <0.10 and exclusion criteria >0.05). The reduced model was adjusted for month of blood draw, breast cancer status (yes, no), colorectal cancer status (yes, no), fracture status (yes, no), participation in the hormone therapy trial (yes, no), and randomization to the dietary modification trial [no, yes (on dietary modification), and yes (not on dietary modification)].

3

The final predictive model, developed with stepwise linear regression, for 25(OH)D was adjusted for all covariates in the reduced model plus adjustment for determined predictors of vitamin D status (age, race-ethnicity, Langleys, waist circumference, recreational physical activity, and total vitamin D intake).

4

Total n does not add up to 3055 because of missing data.

5

Total energy-adjusted vitamin D was modeled by using the log base 2 transformed variable.

6

Age is a quadratic term, age × age.

7

Variables determined to not be significant predictors of serum 25(OH)D status.

8

Reported osteoporosis medications were alendronate, salmon calcitonin, raloxifene, and etidronate.