OBJECTIVE To evaluate the accuracy of percutaneous, image-guided core-needle breast biopsy (CNBx) and to compare the surgical management of patients with breast cancer diagnosed by CNBx with patients diagnosed by surgical needle-localization biopsy (SNLBx). BACKGROUND Percutaneous, image-guided CNBx is a less invasive alternative to SNLBx for the diagnosis of nonpalpable mammographic abnormalities. CNBx potentially spares patients with benign lesions from unnecessary surgery, although false-negative results can occur. For patients with malignant lesions, preoperative diagnosis by CNBx allows definitive treatment decisions to be made before surgery and may affect surgical outcomes. METHODS Between 1992 and 1999, 939 patients with 1,042 mammographically detected lesions underwent biopsy by stereotactic CNBx or ultrasound-guided CNBx. Results were categorized pathologically as benign or malignant and, further, as invasive or noninvasive malignancies. Only biopsy results confirmed by excision or 1-year-minimum mammographic follow-up were included in the analysis. Patients with breast cancer diagnosed by CNBx were compared with a matched control group of patients with breast cancer diagnosed by SNLBx. RESULTS Benign results were obtained in 802 lesions (77%), 520 of which were in patients with adequate follow-up. Ninety-five of the 520 evaluable lesions (18%) were subsequently excised because of atypical hyperplasia, mammographic-histologic discordance, or other clinical indications. There were 17 false-negative CNBx results in this group; 15 of these lesions were correctly diagnosed by excisional biopsy within 4 months of CNBx. In two patients (0.9%), delayed diagnoses of ductal carcinoma in situ were made at 15 and 19 months after CNBx. Malignant results were obtained in 240 lesions (23%), 220 of which were surgically excised from 202 patients at our institution. Two lesions diagnosed as ductal carcinoma in situ were reclassified as atypical ductal hyperplasia and considered false-positive results (0.4%). For malignant lesions, the sensitivity and specificity of CNBx for the detection of invasion were 89% and 96%, respectively. During the first surgical procedure, 115 of 199 patients (58%) diagnosed by CNBx underwent local excision; 194 of 199 patients (97%) evaluated by SNLBx underwent local excision. For patients whose initial surgery was local excision, those diagnosed before surgery by CNBx had larger excision specimens and were more likely to have negative surgical margins than were patients initially evaluated by SNLBx. Overall, patients diagnosed by CNBx required fewer surgical procedures for definitive treatment than did patients diagnosed by SNLBx. CONCLUSIONS Diagnosis by CNBx spares most patients with benign mammographic abnormalities from unnecessary surgery. With the selective use of SNLBx to confirm discordant results, missed diagnoses are rare. When compared with SNLBx, preoperative diagnosis of breast cancer by CNBx facilitates wider initial margins of excision, fewer positive margins, and fewer surgical procedures to accomplish definitive treatment than diagnosis by SNLBx.