A clinicopathologic correlation of mammographic parenchymal patterns and associated risk factors for human mammary carcinoma.

Abstract

The five-year screening experience for 10,131 asymptomatic women evaluated at the Louisville Breast Cancer Detection Demonstration Project (LBCDDP) disclosed 144 breast carcinomas in 1,209 patients (12%) aged 35 to 74 years in whom 904 biopsies and 305 aspirations were performed. This study included 44,711 high-quality xeromammograms (XM) prospectively classified by the modified Wolfe mammographic parenchymal patterns into low-risk (N(1), P(1)) versus high-risk (P(2), DY) groups, with expansion of the P(2) cohort into three additional categories. Using BMDP computer-program analysis, each XM pattern was collated with 21 nonneoplastic and 18 malignant pathologic variables and commonly associated risk factors. A separate analysis of epithelial proliferative and nonproliferative fibrocystic disease of the breast (FCDB) was performed. The histopathology for each biopsy, with distinction of FCDB and neoplasms, was analyzed with regard to the statistical probability of influencing the XM pattern. An average of 1.05 biopsies per patient were performed in women with findings suggestive of carcinoma at clinical and/or XM examinations. An equal distribution of the N(1), P(1), and P(2) DYXM patterns was observed in the 10,131 screenees. Of 8.5% of the screened population having biopsies, 623 were observed to have nonproliferative FCDB and 137, proliferative FCDB. For women 50 years of age or younger, these pathologic variables were seen more frequently in the P(2) DY patterns (p < 0.001), whereas no difference in XM pattern distribution was observed for the screenee 50 years of age or older for proliferative FCDB (p = 0.65). Sixteen percent of the biopsied/aspirated lesions were carcinomas, yielding a biopsy/cancer ratio of 6.25:1. These in situ and invasive neoplasms were more commonly (p < 0.04) observed in 55% of the P(2) (P(2f), P(2n), P(2c)) categories, while 64% of all cancers appeared more frequently in the P(2) DY subgroup (p <0.001), compared with this pattern in the screened population. An equal distribution frequency of the XM classification existed for screenees 50 years of age or older with cancer (p = 0.50), while screenees 35-49 years of age were more often observed to have the high-risk P(2), DY patterns (p <0.001). Analysis of 1,759 histologic characteristics in biopsies of 863 patients with FCDB revealed ductal and lobular hyperplastic lesions, sclerosing adenosis, or epithelial cyst(s) to be the major constituents of 64-69% of the high-risk P(2) (P(2f), P(2n), P(2c)) image (p < 0.001). These data suggest that XM parenchymal patterns observed in asymptomatic screenees incompletely correlate with known pathologic variables and risk factors. Additionally, benefit for recognition of these preinvasive proliferative pathologic factors and carcinomas appears restricted to the younger screenee. The clinical integration of these risk factors with XM patterns may allow preselection of patients deserving of frequent follow-up for breast cancer; however, these data do not support the contention that Wolfe XM patterns are predictors for screening strategies or that they decisively enhance patient management.