CMStatistics 2020: Start Registration
View Submission - CMStatistics
B1054
Title: Accounting for preinvasive conditions in analysis of invasive cancer risk: Application to breast cancer Authors:  Jung In Kim - The Pennsylvania State University (United States) [presenting]
Jason Fine - University of North Carolina at Chapel Hill (United States)
Shanshan Zhao - NIEHS (United States)
Abstract: Ductal carcinoma in situ (DCIS), non-invasive cancer where abnormal cells have been found in the lining of the breast milk duct, is considered as the earliest stage of breast cancer. In epidemiology studies, there are several ways to deal with DCIS. DCIS cases are usually considered as censored cases by restricting the outcome only to invasive breast cancer. Alternatively, the first of either DCIS or invasive breast cancer is regarded as being the outcome. The former makes the strong assumption that DCIS and breast cancer are independent, while the latter fails to distinguish the risk of DCIS from that of breast cancer. In the Sister Study data, almost all women who had been diagnosed with DCIS were treated with lumpectomy or mastectomy, and they will not have invasive breast cancer after DCIS, clearly violating the independent censoring assumption. We propose a competing risks framework for analyzing breast cancer risks in the presence of DCIS by addressing the limitations of the conventional approaches. We demonstrate our approach via comprehensive simulation studies.