Abstract P4-03-07: The contribution of mammographic appearance, basal-like phenotype, and disease extent to prediction of breast cancer death

Abstract

Abstract Background: Advances in molecular medicine and subtyping breast cancer, in addition to conventional treatment indicators, are playing an increasing role in predicting prognosis and tailoring breast cancer treatment. To date, the emphasis principally has been to ensure that treatment is sufficiently aggressive, but in an era when women are increasingly being diagnosed with localized, favorable prognosis cancers, equal emphasis is needed to avoid overtreatment. Objectives: We differentiated the risk of breast cancer death in a longitudinal cohort of women diagnosed with breast cancer, using three parameters of breast tumor extent and biology, specifically basal-like phenotype, histologic tumor distribution (unifocal, multifocal, diffuse), and mammographic appearance of the tumor, allowing for conventional tumor attributes and treatment modalities. Methods: Using a prospective cohort study design, 498 consecutive patients who had been diagnosed with invasive breast cancer at Falun Central Hospital between 1996 and 1998 were enrolled as the test cohort. We assessed the independent effects of the three parameters on the 14-year risk of breast cancer death by estimating the relative contribution of each predictor in proportional hazards regression. Based on these results, we classified prognosis into four categories. External validation was then performed using independent data from a validation cohort of 848 consecutive patients diagnosed with invasive breast cancer in the same institution between 2006 and 2008. Results: Mammographic appearance of the tumor was an independent predictor of risk of breast cancer death (P = 0.0003); the presence of casting type calcifications and architectural distortion on the mammogram were associated with 3.13-fold (95% CI: 1.46-6.70) and 3.19-fold (95% CI: 1.55-6.56) risks of breast cancer death, respectively. The basal-like phenotype independently conferred a 2.68-fold (95% CI: 1.33-5.39) risk of breast cancer death compared with the non-basal type. The association between histologic tumor distribution and risk of breast cancer death was marginally statistically significant (P = 0.0741). The observed deaths did not differ significantly from expected in the validation cohort, indicating adequacy of model prediction. Conclusions: Combining the three parameters, i.e., basal-like phenotype, histologic tumor distribution (unifocal, multifocal, diffuse), and mammographic tumor features accurately predicted long-term risk of breast cancer death. Compared with current treatment planning protocols, risk classification based on basal-like phenotype, histologic tumor distribution (unifocal, multifocal, diffuse), and mammographic appearance, allowing for conventional tumor attributes and treatment modalities, has the potential to reduce overtreatment and is worthy of further validation. Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P4-03-07.