Sentinel lymph node biopsy for ductal carcinoma in situ (DCIS) of the breast is not standard of care. However, nodal involvement for DCIS patients is reported. Aim of our study was to identify preoperative features predictive of nodal involvement in DCIS patients. We have retrospectively reviewed 175 patients with a preoperative diagnosis of DCIS following a vacuum-assisted breast biopsy, and undergoing surgery with sentinel node biopsy. Variables distribution was compared between patients upstaged to invasive cancer at final pathology and patients with a confirmed DCIS, and between positive vs negative sentinel node patients. Univariate and multivariate analyses were performed for risk of a positive node. Lymph node biopsy was positive in 13 (7.4%) patients, with 8 (61.5%) macrometastases and 5 (38.5%) micrometastases. In these patients, Breast Imaging Reporting and Data System (BI-RADS) index >4 (OR 4.69, 95% CI 1.282-17.224, P = .02), lesion extension ≥20 mm (OR 4.25, 95% CI 1.255-14.447, P = .02), multifocal disease (OR 4.12, 95% CI 0.987-17.174, P = .05), comedo type (OR 3.54, 95% CI 1.044-11.969, P = .04), and upstaging (OR 4.56, 95% CI 1.080-19.249, P = .04) were all predictive of nodal involvement, although upstaging could not be predicted preoperatively. By multivariate analysis, the only independent factor predictive for positive sentinel node was multifocal disease (OR 5.14, 95% CI 1.015-26.066, P < .05). A preoperative diagnosis of DCIS, also including advanced biopsy systems such as vacuum-assisted breast biopsy, may be not always sufficient to exclude patients from sentinel node biopsy. DCIS patients with associated BI-RADS >4, lesion extension ≥20 mm, comedo type, and above all multifocal disease should be considered for axillary evaluation.