Table 3.

Predictors of mammographic density in a multilevel nonlinear growth model

VariableInitial status (age, 32 y)
Rate of change (per decade)
Estimate*SEPEstimateSEP
Intercept55.006.85<0.0001
Age (linear)−5.632.500.02
Age2 (nonlinear)1.640.44<0.0001
Case status
    Control (reference)0.000.00
    Case10.172.68<0.0001−1.420.880.11
Ethnicity
    Caucasian (reference)0.000.00
    Hawaiian1.734.560.700.181.560.91
    Japanese−5.624.540.223.791.530.01
    Other6.676.340.29−0.042.260.99
BMI (kg/m2)
    <22.52.682.650.311.340.900.14
    22.5-24 (reference)0.000.00
    25-29−11.762.69<0.00011.860.920.04
    30+−21.553.87<0.00013.581.370.01
Age at menarche (y)
    <13−2.984.790.53−0.041.560.98
    13-144.284.910.38−1.331.600.41
    >14 (reference)0.000.00
Age at first live birth (y)
    None (reference)0.000.00
    <21−12.176.180.050.782.130.71
    21-30−10.885.370.041.581.840.39
    >30−16.706.040.014.232.050.04
Parity
    0-1 (reference)0.000.00
    2-37.124.370.10−2.571.460.08
    ≥43.425.210.51−2.671.700.12
Menopausal status−1.183.660.75−1.251.650.45
HRT
    None (reference)0.00
    Estrogen only1.570.720.03
    Estrogen and progestin3.280.76<0.0001
Soy intake§
    Childhood−7.554.200.071.591.420.26
    Adulthood8.944.250.04−2.081.400.14
  • NOTE: Age, age2 (the square of age), menopausal status, BMI, and HRT are time-varying variables (level 1). All other variables were measured only once at entry into the cohort (level 2).

  • * Estimate is the regression coefficient.

  • Menopausal status was coded 0 for premenopausal, 0.5 for perimenopausal, and 1 for postmenopausal.

  • Hormone replacement use cannot be estimated before menopause.

  • § There is no reference group for soy intake because it is not dummy coded (“as child” and “as adult” can both equal 1).