PURPOSE: Ultrasound-guided core needle biopsy (US CNB) is increasingly used in the histologic evaluation of non-palpable solid breast lesions. We retrospectively investigated the diagnostic accuracy of this technique, using an 18-gauge needle in 422 nonpalpable breast lesions. MATERIALS AND METHODS: 583 female patients with an average age of 56 (range, 22-90) years underwent 590 US CNBs. Between January 1994 and December 1999, using 18-gauge needles, an average of four cores per lesion was obtained. Three hundred and eighty-five lesions were subsequently surgically excised; for 14 of these, the pathologic diagnosis was breast carcinoma metastasis, while 23 with benign diagnoses were clinically followed up for > or = 2.5 years and were considered for analysis. RESULTS: Of the 422 lesions, 340 (80.6%) were malignant [308 invasive, 24 ductal carcinoma in situ (DCIS), 7 DCIS with undetermined invasion and 1 DCIS vs. lobular carcinoma in situ], 67 (15.9%) were benign [30 fibroadenoma (FA) and 37 other diagnoses], and five (1.2%) were fibroepithelial lesions. The remaining ten samples (2.4%) included six cases of atypical ductal hyperplasia (ADH), two of atypical hyperplasia (AH), and two of lobular neoplasia. The sensitivity, specificity, positive predictive value, and negative predictive value of CNBs were 99%, 100%, 100%, and 96%, respectively. Two cases of invasive carcinoma were missed at CNB; there was no false-positive diagnosis. Five of six ADHs and one of two AHs were found to be carcinomas (3 DCIS and 3 infiltrating duct carcinomas). Sixteen of 24 (66.7%) cases of DCIS were found at excision to be invasive carcinomas. Of 31 FAs, two (6.5%) were found to be low-grade phyllodes tumor (PT). The five fibroepithelial lesions were shown at excision to be either PT (n=4) or FA (n=1). CONCLUSION: US CNB using an 18-gauge needle is a safe and reliable means of diagnosing breast carcinoma. Because of the high prevalence of ductal carcinoma in these lesions; findings of ADH/AH at US CNB indicate that surgical excision is needed. In order to rule out low-grade PT, a diagnosis of FA at CNB requires close follow-up.