IDENTIFYING SIGNIFICANT MAMMOGRAPHY FINDINGS AND RISK FACTORS TO REDUCE OVERDIAGNOSIS IN AGING WOMEN

Purpose: While breast cancer mortality is higher in women over age 65, these women are also more likely to experience “overdiagnosis”, which occurs when screening identifies cancer such as ductal carcinoma in situ (DCIS), which may not lead to breast cancer mortality. Therefore, it is important to improve the early diagnosis of invasive breast cancer while minimizing unnecessary invasive procedures that will reduce overdiagnosis. Our objective is to determine the demographic risk factors and mammography predictive of overdiagnosis in aging women.

Method: We constructed logistic regression (LR) models to determine the most predictive mammography findings for estimating breast cancer subtypes in three different age groups:  older age group (age>65), middle age group (age is between 50 and 64), and younger age group (age<50). The LR models use 5607 consecutive mammography records, which contained demographic risk factors and mammography findings. We studied the discrimination problem for two different binary classifications: 1) Benign vs. Malignant (B~M) and 2) Benign/DCIS-Grade1 vs. DCIS-Grade2/DCIS-Grade3/Invasive (B1~M1), i.e. B1~M1 discrimination model classified DCIS-Grade 1 as a benign outcome. We obtained significance levels to identify which mammography findings are the most predictive for each of these two study groups in three different age groups.

Result: For the older age group, personal history of breast cancer, mass margins, mass shape, mass density, calcification distribution, architectural distortion, presence of palpable lump and mass size were statistically significant predictors for B1~M1. In the younger age group, all of these factors were significant for B1~M1 except mass density. When DCIS Grade1 cases are classified as a malignant outcome in the older age group (B~M), architectural distortion (p=0.042) and mass size (0.016) were no longer significant with a significance level of 0.01, while they had p values at 0.0083 and 0.0037 before (B1~M1). Considering that architectural distortion and mass size are distinctive for more invasive characteristics, this observation provides an important policy for distinction of DCIS Grade1 cases in the older age group by classifying them as benign, and helping to reduce the false positive rate.

Conclusion: For women over age 65, the mammographic findings of architectural distortion and mass size can be readily used as distinctive features to differentiate DCIS-Grade1 as a benign outcome, which in turn, may lead to a reduction in overdiagnosis in this age group.