In the last decade, new knowledge has emerged concerning the efficacy of treatment for breast cancer. For that reason, the National Institutes of Health devoted a consensus conference to this topic. To determine whether the consensus conference had influenced practice patterns, and to evaluate the level of quality of care given to women with breast cancer, the authors abstracted and analyzed the medical records of 573 patients treated in ten hospitals throughout Washington state. Results showed no changes with respect to the consensus conference's recommendations for use of a total mastectomy with axillary dissection or the use of a two-step procedure in which the biopsy is performed first and therapeutic options are discussed before a definitive surgery is undertaken. Analyses of quality of care issues not addressed by the consensus conference revealed that 4 percent of the sample were explicitly staged preoperatively and 29 percent postoperatively and that little changed over time in the use of sentinel laboratory tests. These results also show that consensus recommendations will not necessarily change physicians' behavior even where change is possible, and that quality of care in diagnosis and treatment of breast cancer still needs to be addressed.