Wolfe's parenchymal pattern and percentage of the breast with mammographic densities: redundant or complementary classifications?

Mammographic breast densities are one of the strongest breast cancer risk factors. The two most frequently used classifications of breast densities are Wolfe's parenchymal pattern and the percentage of the breast with densities. In this analysis, associations of these two classifications with breast cancer risk were compared, and the dose response curve of risk with densities was examined. Three case-control studies were combined totaling 1060 cases with newly diagnosed breast cancer and 2352 controls. A single observer had assessed parenchymal pattern and percent density without any information on subjects. Relative risks (RRs) were estimated with logistic regression and spline functions adjusting for age and body weight. The two classifications were strongly correlated (r = 0.81, P = 0.0001). Breast cancer risk increased progressively with percent density reaching a 5-6-fold increase for women with 85% or more of the breast with densities compared with women with no density. In contrast, women with P2 or DY patterns had only a 2-3-fold increase in risk compared with women with N1 pattern. More importantly, among women with P2 or DY, RR varied substantially with percent density, whereas, among women with a given percent density, RR varied little with parenchymal pattern. Comparisons of multivariate models reveal that in the presence of parenchymal pattern, inclusion of percent density in the model improved the prediction of breast cancer risk (chi(2) = 35.5, P = 0.0082) but not the opposite (chi(2) = 1.1, P = 0.7662). These findings show that the percentage of the breast with densities provide more information on breast cancer risk than Wolfe's parenchymal patterns and that, once percent breast density is taken into account, no more information on breast cancer risk is given by assessing parenchymal pattern.