Table 2.
Adjusted hazard ratios for invasive breast cancer by BI-RADS density or longitudinal breast density (continuous/categorical)
| Model | Density measure | Adjusted hazard ratio (95% CI) (time-varying breast density covariate) |
|
|---|---|---|---|
| 1 | BI-RADS density | ||
| Fatty | 0.49 (0.40–0.60) | ||
| Scattered | 1 [Reference] | ||
| Heterogeneous | 1.71 (1.56–1.86) | ||
| Extremely dense | 2.11 (1.84–2.42) | ||
| ΔLR-χ2 (df = 3) = 307.7 | |||
| 2 | Continuous longitudinal density | ||
| Linear (per unit*) | 5.53 (3.34–9.14) | ||
| Quadratic (per unit2*) | 0.83 (0.76–0.92) | ||
| ΔLR-χ2 (df = 2) = 379.6 | |||
| 3 | Categorical longitudinal density | ||
| Category 1: | ‘Fatty’ I (< 1.5) | 0.48 (0.36–0.63) | |
| Category 2: | ‘Fatty’ II (1.5 to < 2.0) | 0.83 (0.69–1.00) | |
| Category 3: | ‘Scattered’ I (2.0 to < 2.2) | 1 [Reference] | |
| Category 4: | ‘Scattered’ II (2.2 to < 2.6) | 1.43 (1.24–1.64) | |
| Category 5: | ‘Heterogeneous’ I (2.6 to < 2.9) | 1.82 (1.59–2.07) | |
| Category 6: | ‘Heterogeneous’ II (2.9 to < 3.2) | 2.44 (2.11–2.81) | |
| Category 7: | ‘Extremely dense’ I (3.2 to < 3.6) | 2.66 (2.23–3.17) | |
| Category 8: | ‘Extremely dense’ II (≥ 3.6) | 3.03 (2.48–3.70) | |
| ΔLR-χ2 (df = 7) = 375.1 | |||
n = 132,439
Adjusted hazard ratios from proportional-hazards Cox models for BI-RADS density (time-varying) or the new longitudinal breast density measure (time-varying) as a continuous or eight-category value
Hazard ratios adjusted for age at baseline (continuous) and time-varying BMI (continuous)
‘Time-varying’ means that a woman’s values for breast density and BMI are updated through time, i.e. at each screening examination
Cut points for the eight-category new breast density measure were chosen so that, at baseline mammogram, the proportion of women in the Fatty, Scattered, Heterogeneous and Extremely dense categories are the same as for BI-RADS density, and then split equally into the I and II subgroups within each category
95% CIs from Wald tests
ΔLR-χ2 represents the additional information in the likelihood-ratio statistic when including breast density
BI-RADS Breast Imaging Reporting and Data System, BMI body mass index, df degrees of freedom, 95% CI 95% confidence interval
*Units of longitudinal density are non-standard—interpretation may be facilitated by Fig. 1