Abstract Introduction: The clinical classification of breast cancer into at least four intrinsic subtypes has advanced targeted therapy and improved prognosis. However, for etiological classification, it has been proposed that breast cancer is comprised of just two main subtypes: basal-like and non-basal-like. Evidence for these two etiologic subtypes emerges strongly from bimodal age frequency distributions at diagnosis. In the absence of RNA-based intrinsic subtyping, estrogen receptor (ER) expression is a useful surrogate for these two distinct etiologic classes. Using data from the population-based Carolina Breast Cancer Study (CBCS), we examined evidence for a two-component (ER-positive vs. ER-negative) mixture model for breast cancer biologic/etiologic heterogeneity. Methods: Automated digital scoring of ER expression was performed on immunohistochemistry-stained tissue microarrays comprising 1,920 invasive breast cancer cases from CBCS. Clinical classification of ER status has changed over time as new data have emerged regarding optimal treatment-relevant thresholds, but optimal etiologic thresholds have not been established. Therefore, we considered ER status as a quantitative, categorical variable with cut points of <1% (ER-negative) vs. ≥1% (ER-positive), with ER-positive cases further categorized as highly positive (≥80-100%), intermediate (≥40-<80%), low (≥10-<40%) or borderline (≥1-<10%). Smoothed age frequency distributions at diagnosis (i.e., density plots) were constructed and logistic regression adjusted for age and race was conducted to assess associations between patient and tumor characteristics and level of ER positivity. Results: As expected for etiologically-distinct entities, ER-negative and highly ER-positive tumors showed predominantly unimodal early-onset and late-onset age distributions at diagnosis with peak frequencies near ages 50 and 70 years, respectively. However, tumors with low and intermediate positivity showed bimodal patterns, consistent with a mixture of two main subtypes. Consistent with these age distribution patterns, young age (<40 years) at diagnosis was associated with an elevated odds ratio (OR) for low positive (OR 1.8; 95% CI 1.1-2.9), borderline (OR 1.9; 95% CI 1.1-3.3) and ER-negative disease (OR 2.2; 95% CI 1.5-3.2). Relative to highly ER-positive tumors, low ER-positive tumors were more likely to be node-positive (OR 1.4; 95% CI 1.0-1.9), higher grade (combined grade III; OR 1.9; 95% CI 1.3-2.9), and were more likely to harbor a p53 mutation (OR 2.0; 95% CI 1.2-3.5). Conclusions: While etiologic differences between dichotomized ER-negative and ER-positive breast cancer categories have been well described in many epidemiologic studies, differences in breast cancer etiology across quantitative levels of ER expression have not been so well-characterized. In this study, we report that ER-positive tumors with low positivity share etiologic features of ER-negative tumors, including young age at diagnosis and aggressive tumor characteristics. These data provide additional support for a two-component breast cancer mixture model, with quantitative level of ER positivity reflecting the relative distributions of ER-positive and ER-negative tumor populations. Citation Format: Allott EH, Tse C-K, Carey L, Anderson WF, Olshan AF, Troester MA. Etiologic heterogeneity in breast cancer across quantitative levels of estrogen receptor expression. [abstract]. In: Proceedings of the Thirty-Eighth Annual CTRC-AACR San Antonio Breast Cancer Symposium: 2015 Dec 8-12; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2016;76(4 Suppl):Abstract nr P1-07-04.