Purpose: To determine the accuracy of positron emission tomography with fluorine-18-labeled 2-fluoro-2-deoxy-D-glucose (FDG-PET) in detecting axillary nodal metastases in women with primary breast cancer. Patients and Methods: In this prospective multicenter study, 360 women with newly diagnosed invasive breast cancer underwent FDG-PET. Images were blindly interpreted by three experienced readers for abnormally increased axillary FDG uptake. Imaging results from 308 assessable axillae were compared with axillary node pathology. Results: For detecting axillary nodal metastasis, the mean estimated area under the receiver operator curve for the three readers was 0.74 (range, 0.70 to 0.76). If at least one probably or definitely abnormal axillary focus was considered positive, the mean (and range) sensitivity, specificity, and positive and negative predictive values for PET were 61 % (54% to 67%), 80% (79% to 81%), 62% (60% to 64%), and 79% (76% to 81%), respectively. False-negative axillae on PET had significantly smaller and fewer tumor-positive lymph nodes (2.7) than true-positive axillae (5.1; P < .005). Semiquantitative analysis of axillary FDG uptake showed that a nodal standardized uptake value (lean body mass) more than 1.8 had a positive predictive value of 90%, but a sensitivity of only 32%. Finding two or more intense foci of tracer uptake in the axilla was highly predictive of axillary metastasis (78% to 83% positive predictive value), albeit insensitive (27%). Conclusion: FDG-PET has moderate accuracy for detecting axillary metastasis but often fails to detect axillae with small and few nodal metastases. Although highly predictive for nodal tumor involvement when multiple intense foci of tracer uptake are identified, FDG-PET is not routinely recommended for axillary staging of patients with newly diagnosed breast cancer.