Abstract PR08: Tumor masking or tumor aggressiveness? A structural equations modeling approach to estimate the impact of breast density on breast cancer stage, overall and by race

Abstract

Abstract Background: Non-Latina (nL) black women are diagnosed at later stages and with more aggressive forms of breast cancer than their nL white counterparts, including higher-grade tumors and those that lack estrogen and progesterone receptors (ER/PR negative). High breast density (HBD) is a strong risk factor for a late-stage breast cancer diagnosis among women undergoing mammography screening, yet nL black women are slightly less likely to have dense breasts according to the standard clinical classification method. We previously found that after adjusting for differences in obesity using body mass index (BMI), nL black women were substantially more likely to have dense breasts compared to nL white women. Dense breast tissue can mask tumors and cause them to go undetected at screening, only to arise later as a lump, discovered symptomatically when they later stage. It is also conceivable that dense breast tissue could predispose women to develop more aggressive breast cancers that grow more rapidly, are more likely to arise between screens as “true interval” cancers, and therefore might be detected with more frequent screening. We used path analysis (or structural equations modeling, SEM) to estimate the separate contributions of tumor masking and tumor aggressiveness in mediating HBD's associations with later stage, overall and by race. Methods: We used data on 4691 women from the Metropolitan Chicago Breast Cancer Registry, aged 40-79, and diagnosed at stages 1 (early), 2,3 or 4 (later), between 2001 and 2013. Breast density was obtained from the screening mammogram preceding diagnosis. Mode of detection was defined dichotomously as symptomatic versus screen detected. Path analysis (conducted in Mplus) was used to estimate the age, race, and BMI adjusted, direct and indirect effects of HBD on stage at diagnosis. Because estimates can vary depending on model assumptions, we ran simultaneous linear regression models and also ran simultaneous ordered probit regression models. In both sets of models we estimated the relative contributions of tumor masking and tumor aggressiveness in transmitting the association of HBD with stage at diagnosis. Results: A one-unit increase in BD was associated with a 9-percentage-point increased prevalence of later stage (stage 2,3,4, vs. 1) at diagnosis. Roughly one third of the HBD, later-stage association could be explained by tumor masking, and virtually none of the association could be explained by tumor aggressiveness. Similar results were obtained for linear and ordered probit models. Results were broadly similar for nL black and nL white women. Conclusion: Ours is the first analysis we are aware of to formally confirm via mediation analyses that tumor masking is the primary process through which women with dense breasts are disproportionately diagnosed at later stages. To reduce opportunities for tumor masking, women with dense breasts are often referred for supplemental/advanced imaging with breast MRI, ultrasound, or (increasingly) tomosynthesis. We think it is likely that the current criteria for referring women for advanced imaging based on breast density may contribute to racial disparities because (1) they do not account for the greater prevalence of more aggressive breast cancer in nL black patients, and (2) they do not account for the disproportionate prevalence of obesity among nL black women, which may obscure the usefulness of the usual clinical classification method of BMI in making referrals for supplemental imaging. Citation Format: Katherine Y. Tossas-Milligan, Garth H. Rauscher, Richard Campbell, Victoria Seewaldt. Tumor masking or tumor aggressiveness? A structural equations modeling approach to estimate the impact of breast density on breast cancer stage, overall and by race [abstract]. In: Proceedings of the Tenth AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2017 Sep 25-28; Atlanta, GA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2018;27(7 Suppl):Abstract nr PR08.