To study national trends in the mastectomy rate for treatment of early stage breast cancer. We analyzed data from the Surveillance, Epidemiology, and End Results database, including 256,081 women diagnosed with T1–2 N0–3 M0 breast cancer from 2000 to 2008. We evaluated therapeutic mastectomy rates by the year of diagnosis and performed a multivariable logistic regression analyses to determine predictors of mastectomy as the treatment choice. The proportion of women treated with mastectomy decreased from 40.1 to 35.6 % between 2000 and 2005. Subsequently, the mastectomy rate increased to 38.4 % in 2008 (p < 0.0001). Simple logistic regression models demonstrated that mastectomy rates between 2005 and 2008 were moderated by age (p < 0.0001), marital status (p = 0.0230), and geographic location (p < 0.0001). Multivariate logistic regression analysis found that age, race, marital status, geographic location, involvement of multiple regions of the breast, lobular histology, increasing T stage, lymph node positivity, increasing grade, and negative hormone receptor status were independent predictors of mastectomy. Additionally, multivariate analysis confirmed that women diagnosed in 2008 were more likely to undergo mastectomy than women diagnosed in 2005 (odds ratio 1.17, 95 % confidence interval 1.13 to 1.21, p < 0.0001). There is evidence of a reversal in the previously declining national mastectomy rates, with the mastectomy rate reaching a nadir in 2005 and subsequently rising. Further follow-up to confirm this trend and investigation to determine the underlying cause of this trend and its effect on outcomes may be warranted.