Patients would have been restricted from undergoing biopsy if they weighed more than 300 pounds (136 kg), had bleeding diathesis, or were undergoing anticoagulant therapy that could not be temporarily stopped, but no patients had those conditions during the study period. No restrictions were based on lesion size, position, or conspicuity; breast size; or presumed inability of the patient to cooperate with the procedure. In all cases in which a lesion necessitated image-guided biopsy and the patient was referred from within our multispecialty clinic, stereotactic biopsy of 98% of the lesions was technically feasible and was performed on 97% of the lesions [17]. The other 3% of lesions were evaluated with surgical biopsy.

We performed stereotactic biopsy with the patient prone on a biopsy table (Mammotest or Mammotest Plus/S, Fischer Imaging) with two successive vacuum-assisted techniques. From mid March 1995 to mid June 1996, biopsies were performed with 14-gauge probes, and from mid May 1996 through December 2001, they were performed with 11-gauge probes. During the 1-month overlap between the two periods, availability of the 11-gauge probe determined which method was used. No lesion or pa tient variables were used to determine the method. Four radiologists, including one of the authors, performed the biopsies. They had 3.75, 3.75, 2.75, and 0.25 years of stereotactic biopsy experience performing 14-gauge automated large-core needle biopsy before performing vacuum-assisted biopsy but no experience performing vacuum-assisted biopsy before biopsying the study lesions.

The original histologic diagnoses were accepted for this study. One of the authors classified the histologic diagnoses from the histology reports as malignant, high-risk, or benign lesion. Malignant lesions included invasive carcinoma, ductal carcinoma in situ, and metastasis to the breast. High-risk lesions were those known by sampling error to be frequently associated with underestimation of the presence of carcinoma at the biopsy site. We considered atypical ductal hyperplasia (ADH) a high-risk lesion throughout the study [18]. Non-ADH lesions added to the high-risk category as new information became available from the literature and in our experience were lobular neoplasia (i.e., atypical lobular hyperplasia or lobular carcinoma in situ) [19, 20], radial scar [15, 21], phyllodes tumor [22, 23], papilloma [24, 25], mucocele-like lesions [26, 27], and any nonmalignant lesions with atypia [28]. Those lesions were categorized as high-risk in our results. Lesions not categorized as histologically malignant or high-risk were classified as benign.

Patient variables were age at needle biopsy, lesion palpability (categorized before biopsy by referring clinicians as discretely palpable, vaguely palpable, or nonpalpable), and occurrence of death during the follow-up period (found by review of medical records by one of two authors).

Mammographic variables recorded at biopsy by the radiologist performing the biopsy were mammographic lesion type (microcalcifications alone or masses, including asymmetries and areas of architectural distortion, with or without associated calcifications), maximum mammographic lesion diameter, mammographic tissue density (extremely dense, heterogeneously dense, scattered fibroglandular density, or almost entirely fat), BI-RADS category (2–5) [29], number of lesions biopsied per patient, and involvement of the right or the left breast.

Biopsy variables recorded at biopsy by the radiologist performing the biopsy were gauge of the vacuum probe (14- or 11-gauge), number of tissue specimens obtained from each lesion, presence or absence of calcifications on specimen radiographs, and which radiologist performed the biopsy. A decision about the number of specimens to obtain was left to the discretion of the radiologist performing the biopsy. In general, we tried to obtain at least 12 specimens per lesion, with a tendency to obtain fewer samples of lesions smaller than 6 mm in diameter and more specimens when initial specimen radiographs did not show calcifications or when a large lesion necessitated more than one skin entry site for adequate sampling. One of the authors reviewed the biopsy database to determine the number of vacuum-assisted biopsies the involved radiologist had performed before making each false-negative biopsy finding.

Specimen radiographs were routinely obtained for microcalcification lesions, irregularly obtained (at the discretion of the radiologist performing the biopsy) for calcified masses, and not obtained for uncalcified masses. Visualization of one or more calcifications was considered a positive finding.

The radiologist performing the biopsy, often in consultation with the pathologist evaluating the histologic slides, decided whether the imaging and histologic findings were concordant or discordant. Biopsy results were considered discordant if the histologic findings did not seem to explain the mammographic findings [30]. Until approximately mid 1999, we considered biopsy of a microcalcification lesion technically adequate if calcifications were identified histologically and/or on a specimen radiograph [30]. After that we considered the biopsy finding regarding a histologically benign lesion discordant if calcifications were found at histologic examination but not on the specimen radiograph [6, 12].

Surgical excision was recommended to patients with malignant lesions and lesions categorized as high-risk at biopsy. Because ADH was the only lesion considered high-risk throughout the study, patients with non-ADH high-risk lesions generally underwent mammographic follow-up in the early years of the study and surgical excision in the later years.

Immediate repetition of the biopsy by surgical excision or needle technique was recommended to patients with discordant benign lesions. Patients with concordant benign lesions were advised to undergo 6-month and 12-month postbiopsy mammograms and then annual mammographic follow-up for at least 3 years after biopsy. Delayed repetition of the biopsy was recommended if lesion progression was found at postbiopsy mammographic follow-up [31–34]. Medical records were reviewed by one of two authors for results of follow-up imaging and repeated biopsy.

One of the authors analyzed the data using statistical software (Stata version 9.2, Stata). A p value of < 0.05 on Fisher's exact test was considered significant.

Of the 1,280 lesions, the histologic diagnoses were malignant in 38% (n = 489), high-risk in 9% (n = 117), and benign in 53% (n = 674) of the cases. Malignant lesions were ductal carcinoma in situ in 46% (n = 224), invasive carcinoma in 24% (n = 115), and combined ductal carcinoma in situ and invasive carcinoma in 31% (n = 150) of the cases.

High-risk lesions were ADH in 53% (n = 62), ADH plus another high-risk lesion in 12% (n = 14), lobular neoplasia in 16% (n = 19), radial scar in 5% (n = 6), lobular neoplasia plus radial scar in 1% (n = 1), papilloma in 3% (n = 4), mucocele-like lesion in 2% (n = 2), possible phyllodes tumor in 1% (n = 1), and miscellaneous nonhyperplastic atypical lesion in 7% (n = 8) of the cases. These atypical lesions were angioma (n = 2), sclerosing adenosis (n = 2), spindle cell tumor (n = 2), apocrine metaplasia (n = 1), and tubular adenoma (n = 1).

Benign lesions were fibrocystic change in 48% (n = 325), fibroadenoma (including fibroadenomatous change and fibro adenoma toid change) in 22% (n = 149), fibrosis in 14% (n = 92), lymph node in 1% (n = 5), sclerosing adenosis in 1% (n = 4), and miscellaneous benign in 15% (n = 99) of the cases.

Patient, mammographic, biopsy, and follow-up variables for the five false-negative lesions are shown in Table 1. Details of interest beyond those in Table 1 and the figures are as follows. Lesion 2 was difficult to classify. The finding at 14-gauge vacuum biopsy was concordant, but at 11-gauge vacuum biopsy performed when lesion progression was found at imaging 90 months after the initial biopsy, the finding was ductal carcinoma in situ (Fig. 1A, 1B, 1C, 1D). Because the progressive calcifications were at the exact site of the original calcifications, it is most accurate to call the lesion false-negative despite the long follow-up period.

Lesion 3 was a false-negative discordant calcification lesion (Fig. 2A, 2B, 2C, 2D) with absence of calcifications on the specimen radiograph related to inaccurate needle position. Lesion 4, a false-negative discordant mass lesion (Fig. 3A, 3B), was a faint 4-mm indistinct mass that was difficult to see during stereotactic biopsy. Lesion 5, a false-negative mass lesion, was a 14-mm indistinct isodense mass. Because benign ducts were concentrated in a portion of the biopsy samples, we considered this a concordant lesion consistent with a focal area of fibroglandular tissue. In retrospect, we should have considered this biopsy finding discordant, necessitating immediate repetition of the biopsy.

For lesions with a final diagnosis of malignancy, variables of lesions that were false-negative at initial biopsy were compared with those of lesions not false-negative at initial biopsy. Gauge of biopsy probe and findings on specimen radiographs were the only significant variables (Table 2). Biopsy of the few palpable lesions and the two BI-RADS category 3 lesions was performed at the request of referring surgeons.

Specimen radiography was performed on 765 of 766 calcification lesions, 102 of 112 calcified masses, and none of 402 uncalcified masses. By lesion type, radiographs showed no calcifications in the cases of 1.7% (13 of 765) of calcification lesions and 2% (two of 102) of calcified masses (p = 0.693). By biopsy device, radiographs showed no calcifications in the cases of 5% (five of 101) of lesions evaluated with 14-gauge biopsy and 1.3% (10 of 766) of lesions evaluated with 11-gauge biopsy (p = 0.022).

Biopsy was repeated on nine discordant benign lesions without imaging follow-up. The repeated biopsy was performed by surgical excision (n = 6) or vacuum needle technique (n = 3). The discordant lesions manifested themselves as calcifications (n = 4) or masses (n = 5). Discordant lesions were false-negative for one calcification lesion and one mass lesion. No false-negative findings were made in the cases of seven lesions with absence of calcifications on specimen radiographs thought to be concordant in the early years of the study because of histologic calcifications. Six of the seven patients had stable mammographic follow-up findings (mean follow-up period, 90 months; range, 42–129 months), and one was lost to follow-up.

Malignant lesions can be completely removed at percutaneous biopsy [35, 36], so all 489 malignant lesions were considered malignant in our analysis whether or not the malignancy was found at surgical excision.

Adequate follow-up of high-risk lesions requires either repeated biopsy or imaging follow-up for 24 months or longer [15, 21, 25, 27]. The 117 high-risk lesions were followed by repeated biopsy (n = 98) or mammography for 24 months or longer (n = 13). Follow-up of six lesions was inadequate. Histologic diagnoses of the six lesions with inadequate follow-up were ADH (n = 3), ADH plus another high-risk lesion (n = 1), papilloma (n = 1), and miscellaneous atypical lesions (n = 1). The patients with these six lesions underwent mammographic follow-up for less than 24 months (n = 5) or were lost to follow-up (n = 1).

Malignancy was found in 13% (14 of 111) of high-risk lesions with adequate follow-up. These underestimated malignant lesions were originally diagnosed as ADH (nine of 59 lesions), ADH plus another high-risk lesion (two of 13 lesions), lobular neoplasia (one of 19 lesions), miscellaneous atypical lesion (two of seven lesions), or other (zero of 13 lesions).

Adequate follow-up of benign lesions requires either repetition of biopsy or imaging follow-up for 24 months or longer [31–34, 37]. The 674 benign lesions in this study were subjected to repeated biopsy (n = 46) or mammographic follow-up for 24 months or longer (n = 506; median, 68 months; range, 24–122 months). Follow-up of 122 lesions (18% of all benign lesions) was inadequate. The patients with these 122 lesions underwent less than 24 months of mammographic follow-up (n = 57; median, 12 months; range 3–21 months) or were lost to follow-up (n = 65, 10% of all benign lesions).

A significant difference between benign lesions with and those without adequate follow-up occurred in one variable. The patients with 14% (17 of 122) of lesions with inadequate follow-up died during the follow-up period, as did the patients with 1.4% (eight of 552) of the lesions with adequate follow-up (p = 0.0001). No deaths were attributed to breast cancer. Malignancy was found in five of the 46 lesions on which biopsy was repeated. This number represents 0.91% (five of 552) of benign lesions with adequate follow-up and 0.74% (five of 674) of all benign lesions.

Targeted repeated biopsy of 33 lesions was performed by surgical excision (n = 13) or vacuum-assisted needle biopsy (n = 20). Biopsy of the 33 lesions was repeated because of discordance (n = 9), because excision was suggested by the pathologist (n = 3) (single lesions diagnosed as a spindle cell tumor, unusual stromal proliferation, and unusual adenoma), or because follow-up imaging showed the lesion had grown (n = 21). Incidental surgical excision was performed on 13 benign lesions because of carcinoma elsewhere in the breast (n = 12) or reduction mammoplasty (n = 1). The five false-negative lesions were found in three of 21 lesions on which biopsy was repeated because follow-up imaging showed the lesion had grown, two of nine discordant lesions, none of three lesions with histologic concerns on the part of the pathologist, and none of 13 lesions incidentally excised.

Of the 508 lesions with a final diagnosis of malignancy, the initial needle biopsy diagnosis was malignant (n = 489), high-risk (n = 14), or benign (n = 5). The false-negative rate thus was 1% (five of 508 malignant lesions).

In evaluating 1,280 consecutively detected lesions by percutaneous stereotactic biopsy performed using a vacuum-assisted probe with the patient prone on the table, we had a false-negative rate of 1% (five of 508 of malignant lesions). False-negative lesions were compared on the basis of patient, mammographic, and biopsy variables (Table 2) and were significantly related to two of the four primary variables studied.

False-negative biopsy findings were related to the gauge of the biopsy probe and to specimen radiographic findings in respect to calcifications. False-negative diagnoses were made in 4.4% (three of 68) of cases in which 14-gauge vacuum biopsy was performed and in 0.45% (two of 440) of cases in which 11-gauge vacuum biopsy was performed (p = 0.019). False-negative diagnoses were made in 25% (one of four) of cases in which specimen radiographs showed no calcifications and 0.67% (two of 300) of cases in which they did show calcifications (p = 0.0390) (Table 2). False-negative biopsy findings, despite the presence of calcifications on specimen radiographs, were made in two cases in which 14-gauge vacuum biopsy was performed and have been reported with 11-gauge vacuum biopsy [3, 5, 6], the cause being uncertain. Intralesional histologic heterogeneity, which can cause sampling problems and lead to false-negative results of needle biopsy, is known to occur with both microcalcification and mass lesions [38, 39]. In addition, malignant tumors are known to occur incidentally adjacent to calcifications in benign tissue, and sampling of the calcifications at needle biopsy can miss the malignant growth [40]. Perhaps one of those factors is responsible for a false-negative biopsy result despite the presence of calcifications on a specimen radiograph.

The rate of false-negative biopsy findings was not related to the biopsy experience of the radiologist. False-negative findings were made in the cases of 3.6% (one of 28) of malignant lesions in the first 15 biopsies by each radiologist and 0.83% (four of 480) of malignant lesions after the first 15 biopsies (p = 0.248). The rate of false-negative findings also was not related to the number of specimens obtained per lesion (p = 1.000) (Table 2). No secondary variables were significant. Most important, false-negative biopsy findings were not related to lesion type, being made in the cases of 1.2% (three of 248) of calcification lesions and 0.8% (two of 260) of mass lesions (p = 0.679).

We are aware of two other articles [1, 2] describing the evaluation of false-negative rates of stereotactic 14-gauge vacuum-assisted biopsy. Those articles give conflicting results in comparisons of 14- and 11-gauge vacuum-assisted false-negative rates [1, 2]. Liberman et al. [1] reported similar false-negative rates for the two probe sizes. False-negative findings were made in the cases of 1.7% (one of 58) of lesions subjected to 14-gauge vacuum biopsy and 1.8% (two of 113) of lesions subjected to 11-gauge biopsy. Shah et al. [2] had substantially different results: false-negative findings in the cases of 22.2% (four of 18) of lesions subjected to 14-gauge biopsy and 3.3% (two of 61) of lesions subjected to 11-gauge biopsy. The relatively small number of lesions in those studies may account for the disparity in results.

Like Shah et al. [2], we had significantly more false-negative biopsy findings with 14-gauge devices, but we biopsied substantially more lesions and had substantially lower false-negative rates with both devices. Our false-negative rates were 4.4% (three of 68 lesions) for 14-gauge biopsy and 0.45% (two of 440 lesions) for 11-gauge biopsy (p = 0.19).

More robust data are available for 11-gauge vacuum-assisted stereotactic biopsy [3–8], and our results with 11-gauge vacuum probes are compared with those data in Table 3. The false-negative rates range from 0 to 3.3%, the rate for the combined six studies being 1.1% (20 of 1,755 malignant lesions). Liberman et al. included 11-gauge-biopsy lesions from an earlier study [1] in a later study [5] reported in Table 3. Shah et al. [2] did not provide enough data for inclusion of lesions subjected to 11-gauge biopsy in Table 3.

Follow-up of benign lesions is variable (Table 3); 100% follow-up was achieved in the two studies with the smallest number of lesions [4, 6]. Three other studies [3, 5, 7, 8] lacked follow-up of 30–37% of benign lesions, compared with our lack of follow-up of 10% of benign lesions. Ours was the only study in which the majority of benign lesions were followed up for more than 24 months. Three numbers are substantially higher in the study by Pfarl et al. [6] than in the other studies of 11-gauge biopsy (Table 3). Pfarl et al. found cancer at initial biopsy in 63% of lesions, suggesting different selection criteria for lesions being biopsied; surgically excised all 318 study lesions; and had a false-negative rate of 3.3%.

Most articles, including those reporting false-negative rates at stereotactic 14-gauge and 11-gauge vacuum biopsy [1–8], do not address the issue of selection bias. The articles that do address it provide variable information [12]. With 97% of lesions in our study being biopsied with the stereotactic technique and with no selection bias for the vacuum biopsy probe, we believe we had less selection bias than in other reports. Selection bias for biopsy method may influence false-negative rates and other measures of success.

Limitations of our study include its retrospective nature, reliance on the histologic presence of calcifications despite absence of calcifications on specimen radiographs in the early years of the study, lack of repetition of biopsy of all benign lesions with absence of calcifications on specimen radiographs, and lack of adequate follow-up of all benign lesions. A necessary limitation in our long-term study was the addition of different histologic lesions to the group considered high-risk lesions as the literature and our experience evolved. In our study, the number of lesions subjected to 14-gauge vacuum biopsy was larger than previously reported in the literature (a strength) but was substantially smaller than the number of lesions subjected to 11-gauge vacuum biopsy in our study (a weakness).

Strengths of our study include the large number of lesions on which 11-gauge biopsy was performed, minimal bias in selection of lesions necessitating image-guided biopsy, lack of selection bias regarding the two successive stereotactic vacuum biopsy methods, performance of similar biopsies of calcifications and masses, and thoroughness of mammographic follow-up of benign lesions on which biopsy was not repeated.

False-negative findings at stereotactic vacuum-assisted needle biopsy were less common with an 11-gauge probe (0.45%, two of 440 lesions) than with a 14-gauge probe (4.4%, three of 68 lesions) (p = 0.19). The 1.1% rate of false-negative findings at stereotactic 11-gauge vacuum-assisted needle biopsy discerned in the literature review (Table 3) compares favorably with the 2.0% false-negative rate for needle-localized surgical breast biopsy [41] and the 4.0% rate with stereotactic 14-gauge automated large-core needle biopsy [31] found in the literature. False-negative biopsy findings were significantly related to specimen radiographic findings with respect to calcifications but were not significantly related to lesion type. False-negative findings were made in the cases of 0.67% (two of 300) of lesions in which calcifications were present on specimen radiographs and 25% (one of four) of lesions with absence of calcifications on specimen radiographs (p = 0.039). False-negative findings also were made in the cases of 1.2% (three of 248) of calcification lesions and 0.8% (two of 260) of mass lesions (p = 0.6790). Results of a comparison of the accuracy of biopsy of mass lesions using sonographic guidance and various needles versus biopsy using stereotactic guidance and vacuum-assisted needles would be of interest.