Breast Cancer Relative Hazard Estimates From Case–Control and Cohort Designs With Missing Data on Mammographic Density