BACKGROUND Percutaneous core needle biopsy (CNB) is considered to be the standard technique for histological diagnosis of breast lesions. But, it is less reliable for diagnosing atypical ductal hyperplasia (ADH). The purpose of the present study was to predict, based on clinical and radiological findings, which cases of ADH diagnosed by CNB would be more likely to be associated with a more advanced lesion on subsequent surgical excision. METHODS Between February 2002 and December 2007, consecutive ultrasound-guided CNBs were performed on suspicious breast lesions at Seoul St. Mary's Hospital. A total of 69 CNBs led to a diagnosis of ADH, and 45 patients underwent follow-up surgical excision. We reviewed the medical records and analyses retrospectively. RESULTS Sixty-nine patients were diagnosed with ADH at CNB. Of these patients, 45 underwent surgical excision and 10 (22.2%) were subsequently diagnosed with a malignancy (ductal carcinoma in situ, n = 8; invasive cancer, n = 2). Univariate analysis revealed age (>or= 50-years) at the time of core needle biopsy (p = 0.006), size (> 10 mm) on imaging (p = 0.033), and combined mass with microcalcification on sonography (p = 0.029) to be associated with underestimation. When those three factors were included in multivariate analysis, only age (p = 0.035, HR 6.201, 95% CI 1.135-33.891) was an independent predictor of malignancy. CONCLUSIONS Age (>or= 50) at the time of biopsy is an independent predictive factor for breast cancer at surgical excision in patients with diagnosed ADH at CNB. For patients diagnosed with ADH at CNB, only complete surgical excision is the suitable treatment option, because we could not find any combination of factors that can safely predict the absence of DCIS or invasive cancer in a case of ADH.