Background: In a previous study of the relationship between breast density and primary tumor features, we observed a higher mastectomy rate in patients with extremely dense breasts. Here we examine possible reasons for this finding. Methods: Data were obtained from a prospectively maintained database of 1,056 invasive breast cancer patients from January 2005 to June 2007. Mammographic density was assigned by Breast Imaging-Reporting and Data System (BI-RADS) classification. Initial and final surgical procedures, and patient and tumor variables were recorded. Results: Breast-conserving surgery (BCS) was attempted in 758 patients (72 %), 385 (51 %) of whom had preoperative magnetic resonance imaging (MRI). Initial BCS was less common among patients with the highest (BI-RADS 4) breast density compared to patients with less-dense breasts (52 vs. 74 %; p < 0.0001), but MRI use was more common (65 vs. 33 %; p < 0.0001). Adjusting for clinical and pathologic variables, patients with the highest breast density had 1.94-times (95 % confidence interval 1.44-2.62; p < 0.0001) the odds of initial mastectomy compared to patients with less-dense breasts. After initial BCS, 387 patients (51 %) had positive shaved margins, 96 (25 %) of whom converted to mastectomy. MRI did not correlate with the rate of positive margins overall or among those with dense breasts. Adjusting for clinical and pathologic variables, density did not predict margin status or conversion to mastectomy. In a multivariate model, age, histologic grade, extensive intraductal component, and multicentricity/multifocality were independently associated with conversion to mastectomy. Conclusions: Density alone seems to influence the decision to proceed with initial mastectomy. When BCS was attempted, breast density was not associated with positive margins or conversion to mastectomy. A benefit of MRI in decreasing positive margins was not observed. These data do not support the use of breast density as a selection criterion for BCS.