Core needle biopsy (CNB) is used to sample both mammographically and ultrasound detected breast lesions. A diagnosis of ductal carcinoma in situ (DCIS) by CNB does not ensure the absence of invasive cancer upon surgical excision and as a result an upstaged patient may need to undergo additional surgery for axillary nodal evaluation. This study evaluates the accuracy of CNB in excluding invasive disease and the preoperative features that predict upstaging of DCIS to invasive breast cancer. Two hundred fifty-four patients over an 8-year period from 1994 to 2002 with a diagnosis of DCIS alone by CNB were retrospectively reviewed. Underestimation of invasive cancer by CNB was determined. Radiographic, pathologic, and surgical features of the cohort were compared using univariate and multivariate analysis. The mean age was 55 years (range 27-84) and mean follow-up was 25 months with one patient unavailable for follow-up. There were a total of six patient deaths, all of which were not disease-specific. A total of 21 out of 254 patients (8%) with DCIS by CNB were upstaged to invasive cancer following surgical excision. There was a significant inverse relationship between the number of core biopsies and the incidence of upstaging (p < 0.006) in that patients with fewer core samples were more likely to be upstaged at surgical pathology. No relationship was noted between the size of the core samples and the likelihood of upstaging (p > 0.4). Of 21 patients with invasion, all but two had comedonecrosis by CNB. Comedonecrosis by CNB significantly increased the likelihood of upstaging (p < 0.001). Of the 21 patients who were upstaged, 12 required subsequent surgery for nodal evaluation while nine had sentinel node biopsy at initial operation. Finally, upstaged patients were significantly more likely to have a positive margin (p < 0.008). Ductal carcinoma in situ with comedonecrosis on CNB can help to predict the possibility of invasion. Increasing the number of core biopsies reduced the likelihood of sampling error.