There is sparse evidence on community practice patterns in treating women with breast cancer. This study compared care of women with breast cancer with evidence from meta-analyses and US National Comprehensive Cancer Network (NCCN) clinical guidelines. Records of 4395 women with breast cancer were abstracted from practices of 19 surgeon oncologists in six specialty practices in the Philadelphia region during 1995-1999. Patients were followed through December 2001. Low-frequency data were obtained on all patients. All other data were from a random sample of 464 women, minimum of 50 patients per practice. Actual care provided was compared to NCCN guidelines and results of meta-analyses. Fewer than half the women received treatments reflecting meta-analysis results or NCCN guidelines, by disease stage/TNM status. Adherence to either standard varied from 0% for LCIS to 87% for stages IIA or IIB node positive. There are multiple interactive reasons for low adherence to guidelines or meta-analyses results, including insufficient health system supports to clinicians, inadequate organisation and delivery systems and ineffective continuing medical education. The paucity of written information from patient records on physician/patient interactions limits the understanding of treatment decisions.