I read with interest the article by Kettritz et al. 1 regarding the accuracy of stereotactic 11-gauge vacuum-assisted breast biopsy (VAB) for diagnosing breast cancer in nonpalpable breast lesions. Of the 2893 VAB procedures performed, 19 (0.7%) yielded findings that were not representative, whereas 2874 were performed successfully. Atypical ductal hyperplasia (ADH), which was diagnosed in 141 lesions, was upgraded to cancer in 32 of 135 lesions that underwent surgical excision. No followup information was available for lobular carcinoma in situ (LCIS), which was diagnosed in 18 cases. Benign histology was diagnosed in 2090 lesions, with cancer going undetected in 1 of 24 lesions that underwent surgical excision. The false-negative rate was defined as the percentage of lesions diagnosed as being benign on VAB that later proved to be malignant. This rate, which was calculated excluding nonrepresentative lesions and including benign lesions with (n Ο 1438) or without (n Ο 628) follow-up imaging data, was 0.05% (1 of 2090).
Although the reported results are impressive, I believe that the analysis may overstate the accuracy of this technique. I prefer an alternative false-negative rate, defined as the percentage of pathologically confirmed malignant lesions that yielded benign stereotactic biopsy findings without findings of atypia or high-risk lesions. 2 This approach requires delineation of atypical and high-risk lesions (including ADH and, variably, LCIS, atypical lobular hyperplasia, radial scar, papilloma, and/or benign phyllodes tumor), knowledge of the total number of cancers found in the study group at any time, and knowledge of how many of those cancers yielded benign (including discordant or nonrepresentative) findings on original biopsy. Falsenegative rates calculated in this manner have been reported for stereotactic 11-gauge VAB 2,3 and automated 14-gauge large-core needle biopsy. 2,4,5 To calculate this alternative false-negative rate for the study conducted by Kettritz et al., 1 more data are needed.
For the 19 nonrepresentative lesions and the 18 lesions diagnosed as LCIS, we must know how many of each were rebiopsied and how many cancers subsequently were identified. The total number of cancers among the 2893 lesions evaluated with VAB equals the number of cancers detected on rebiopsy of nonrepresentative lesions and rebiopsy of lesions diagnosed as LCIS plus the number of cancers detected on VAB (n Ο 625), rebiopsy of lesions diagnosed as ADH (n Ο 32), and rebiopsy of lesions found to be benign (n Ο 1). The total number of false-negative lesions equals the number of cancers found on rebiopsy of nonrepresentative lesions plus the one cancer detected on rebiopsy of a lesion that initially was found to be benign.