OBJECTIVE: To find out if the sentinel node can be detected in sufficient numbers of women with breast cancer to be useful as a prognostic sign, whether it reflects that state of the entire axilla, and whether it detects micrometastases that would otherwise be missed. DESIGN: Prospective study. SETTING: 3 teaching hospitals, Sweden. SUBJECTS: 75 patients with breast cancer who were listed to have axillary dissection as well as resection of their tumour. INTERVENTIONS: Injection of 99Tc nanocolloid 0.4 ml and patent blue dye 1 ml around the tumour or under the skin above the tumour, followed by preoperative lymphoscintigraphy and then identification of the sentinel node during operation either because it had turned blue or with a gamma probe. Removal of the sentinel node and complete axillary dissection. MAIN OUTCOME MEASURES: Identification of the sentinel node and presence of metastatic nodes in the axilla. RESULTS: The sentinel node was identified in 69/75 (92%). It correctly predicted the state of the axilla in 66/69 (96%), and detected metastases in 24 of the 27 with invaded nodes in the axilla (89%). The false negative rate was 11%. In 14/27 with axillary metastases (52%) the sentinel node was the only involved node. In 3/24, metastases were detected by immunohistochemistry alone. CONCLUSION: Biopsy of the sentinel node predicted the presence or absence of axillary metastases with acceptable accuracy. However, before axillary node dissection is rejected in favour of sentinel node biopsy alone, large multicentre studies are needed to establish the true false negative rate.