Background. Pilot studies indicate that probe-guided resection of radioactive sentinel nodes (the first nodes that receive drainage from tumors) can identify regional metastases in patients with breast cancer. To confirm this finding, we conducted a multicenter study of the method as used by 11 surgeons in a variety of practice settings. Methods. We enrolled 443 patients with breast cancer. The technique involved the injection of 4 ml of technetium-99m sulfur colloid (1 mCi [37 MBq]) into the breast around the tumor or biopsy cavity. 'Hot spots' representing underlying sentinel nodes were identified with a gamma probe. Sentinel nodes subjacent to hot spots were removed. All patients underwent a complete axillary lymphadenectomy. Results. The overall rate of identification of hot spots was 93 percent (in 413 of 443 patients). The pathological status of the sentinel nodes was compared with that of the remaining axillary nodes. The accuracy of the sentinel nodes with respect to the positive or negative status of the axillary nodes was 97 percent (392 of 405); the specificity of the method was 100 percent, the positive predictive value was 100 percent, the negative predictive value was 96 percent (291 of 304), and the sensitivity was 89 percent (101 of 114). The sentinel nodes were outside the axilla in 8 percent of cases and outside of level 1 nodes in 11 percent of cases. Three percent of positive sentinel nodes were in nonaxillary locations. Conclusions. Biopsy of sentinel nodes can predict the presence or absence of axillary-node metastases in patients with breast cancer. However, the procedure can be technically challenging, and the success rate varies according to the surgeon and the characteristics of the patient.