Original article Annals of Oncology 14: 450 454, 2003 DOI: 10.1093/annonc/mdh088 Mammographic features and correlation with biopsy findings using 11-gauge stereotactic vacuum-assisted breast biopsy (SVABB) A. Mendez, F. Cabanillas*, M. Echenique, K. Malekshamran, I. Perez & E. Ramos Auxilio Mutuo Cancer, Hospital Auxilio Mutuo, San Juan, Puerto Rico Received 16 June 2003; revised 13 October 2003; accepted 6 November 2003 Background: Mammograms are assigned a BI-RADS (Breast Imaging Reporting and Data System) category, which indicates the level of suspicion for cancer. Objectives: (i) To evaluate the use of BI-RADS categories in a non-academic radiology practice based in a community hospital compared with local radiology private offices; (ii) to determine positive predictive value (PPV), sensitivity and specificity of mammograms; and (iii) to explore the correlation of BI-RAD 3 5 and lesion description with diagnosis of cancer. Patients and methods: We performed 947 SVABBs (stereotactic vacuum-assisted breast biopsies) on 911 patients with BI-RADS 3 5. Lesions were classified as: 1 = microcalcifications; 2 = asymmetric density; 3 = circumscribed mass; and 4 = spiculated mass. Results: BI-RADS category correlated with diagnosis of breast cancer (atypia excluded): category 3 = 4%; category 4 = 15%; and category 5 = 79%. The PPV of BI-RADS 4 and 5 for breast cancer or atypia was 20%, in contrast to 5% for BI-RADS 3. Sensitivity and specificity were 95% and 19%, respectively. For BI-RADS 3 without microcalcifications only 3% were positive, in contrast to 8% for remainder. Conclusions: First, there is a stepwise increase in cancer for each of the BI-RADS categories 3 5. Secondly, in BI-RADS 3 with microcalcifications, a biopsy is indicated according to our findings. Finally, the sensitivity of mammograms is 95% but the specificity is 19%. Key words: biopsy, BI-RADS category, cancer diagnosis, mammography, SVABB Introduction The stereotactic vacuum-assisted breast biopsy (SVABB) technique is now increasingly used because it provides a minimally invasive, faster and less expensive method for sampling nonpalpable abnormalities seen on mammograms but not visualized on sonograms. The Breast Imaging Reporting and Data System (BI-RADS) was proposed by the American College of Radiology [1] in order to provide a common language to unambiguously describe the level of suspicion about a mammographic lesion. Annual follow-up is recommended for lesions classified as BI-RADS category 1 (negative mammogram) or 2 (benign findings). A 6-month follow-up for the breast in question, with 1-, 2- and 3-year follow ups for both breasts, has been recommended for lesions assigned a BI-RADS category 3 (probably benign), while a biopsy is suggested for those classified as category 4 (suspicious) or category 5 (highly suggestive of malignancy). However, some BI-RADS 3 lesions are biopsied if the patient or the referring physician insists [2]. The reasoning for this is that it *Correspondence to: Dr F. Cabanillas, Auxilio Mutuo Cancer, Hospital Auxilio Mutuo, Avenida Ponce de Leon 725, San Juan, PR 00918, Puerto Rico. Tel: +1-787-771-7935; Fax: +1-787-771-7941; E-mail: cabanillas@auxilio.com might at least help to avoid open biopsy, which might be done anyway by the referring physician [2]. If a lesion is visualized by sonography, our approach is to perform a biopsy by sonographic guidance because it is less expensive. If the lesion is not seen using sonomammography then the biopsy is done with either stereotactic guidance or by wire localization. For those cases not visualized on sonogram, we prefer the stereotactic method because it is faster, less expensive, less painful, less invasive and can be easily performed in the outpatient setting. In our center, many of the BI-RADS category 5 cases go directly for a surgical biopsy or sonogram-directed biopsy in view of the high suspicion for cancer, but some have stereotactic biopsies. Only five published series have attempted to correlate the assigned BI-RADS category with the results of stereotactic breast biopsies. In these series, the results of stereotactic breast biopsy correlate well with BI-RADS category, but there is a relatively wide range of yield among the various series. We decided to conduct this study with the following objectives: (i) to evaluate the use of BI-RADS categories in a non-academic radiology practice based in a community hospital as compared with the local radiology private offices; (ii) to determine if the use of BI-RADS categories 3 5 is useful and to explore whether the use of BI-RADS category 3 needs to be better defined; (iii) to 2004 European Society for Medical Oncology
451 determine the positive predictive value (PPV), sensitivity and specificity of mammograms; and (iv) to explore the correlation of BI-RAD 3 5 and description of lesion with the diagnosis of cancer. Materials and methods From August 2000 through December 2002, we performed 947 SVABBs on 911 patients using an upright Siemens unit with a mammotome drive and 11-gauge needle. Specimen radiography was routinely performed on all samples and mammographic lesions were classified into four categories: 1 = microcalcifications; 2 = asymmetric density; 3 = circumscribed mass; 4 = spiculated mass. The major indication for SVABB in our series was a BI-RADS category 4 mammogram (n = 757, 80% of all procedures); however, 156 were done for BI-RADS category 3 and 34 for category 5. Of the 947 biopsies, 730 (77%) were performed for microcalcifications, 97 (10%) for an asymmetric density, 85 (9%) for a circumscribed mass and 35 (4%) for a spiculated mass. Before the SVABB was performed, all patients were evaluated by one of three radiologists assigned to the Breast Imaging Section and all films were reviewed. A BI-RADS category was assigned before the SVABB. Either a surgeon or a radiologist performed the SVABB procedure. During this period of time, four surgeons and three radiologists performed SVABB. In order for a surgeon or a radiologist to perform this procedure he or she had to be credentialed by the Faculty By-Laws Committee. The criteria that were used for credentialization of surgeons are virtually identical to those proposed by the American College of Radiologists [3]. Whenever the surgeon performed the procedure, the radiologist assisted with the targeting. Before submitting the specimen to Pathology, it was first X-rayed in the Department of Radiology. Subsequently, all pathology reports were reviewed by the radiologist and an assessment of concordance or discordance was done on each case. BI-RADS category was assigned using the interpretation provided by the first radiologist who read the films, either the outside radiologist or the institutional radiologist. Of the 947 mammograms done before the SVABB, 313 were performed at our center and 634 were performed outside. We identified 38 cases without microcalcifications in whom there was information on the exact size of the lesion. The median size was 0.9 cm (range 0.2 2.0 cm). In order to calculate the specificity and sensitivity of the mammographic interpretation, we used the following formulae: sensitivity = true-positive mammograms/true-positive mammograms plus false-negative mammograms specificity = true-negative mammograms/false-positive mammograms plus true-negative mammograms True-positive mammograms were defined as those with a BI-RADS >3 and a positive stereotactic biopsy for carcinoma or atypical hyperplasia. Those SVABBs considered to yield inadequate tissue that required a repeat biopsy with a resulting diagnosis of breast cancer or atypical hyperplasia were also considered as true positives, as well as anyone who upon follow-up required a repeat biopsy that was positive after the original SVABB. A true-negative mammogram was defined as anyone with a BI-RADS category 3, with an initially negative stereotactic biopsy. Those with atypical hyperplasia were considered as positive and thus were not included as true negatives. Anyone requiring a repeat biopsy, which yielded a negative result, was also considered as a true negative. The tumor registry was also examined in an attempt to find any cases diagnosed as breast cancer whose initial SVABB was negative. False-positive mammograms were defined as any BI-RADS >3 that yielded a negative result, including atypical hyperplasia, at the time of the first stereotactic biopsy or at rebiopsy for any reason, provided the repeat biopsy was negative. If upon follow-up a repeat biopsy had to be performed and it was positive then the case was considered a true-positive mammogram. Table 1. Correlation of BI-RADS category with a diagnosis of breast cancer or atypical hyperplasia after SVABB BI-RADS n Breast cancer Atypia Total 3 156 7 a (4.5%) 1 (0.6%) 8 (5.1%) 4 757 116 (15.3%) 9 (1.2%) 125 (16.5%) 5 34 27 (79.4%) 0 27 (79.4%) All 947 150 (15.8%) 10 (1.1%) 160(16.9%) a One of the seven was LCIS, which could be an incidental finding. One of the cases was ADH. False-negative mammograms were defined as a BI-RADS 3 with a positive stereotactic biopsy result, including carcinoma or atypical hyperplasia. If a repeat biopsy in a BI-RADS 3 mammogram was performed and it was positive then the mammogram was also considered as a false negative; any initial biopsy in a BI-RADS 3 considered as inadequate tissue and requiring a repeat biopsy that was positive for breast cancer or atypical hyperplasia was also considered a false negative. Also those BI-RADS 3 who upon follow-up required a repeat biopsy that was positive were considered false negatives. Permission was obtained from the local institutional review board for publication of the findings summarized in this study. Results The median age of our population was 57 years (range 25 89). Of 911 patients, 149 (16.4%) were diagnosed with breast cancer; 150 of 947 biopsies (15.8%) were positive for breast cancer. An additional 10 biopsies showed cytological atypia requiring further investigation for a total of 159 (17.5%) positive findings out of 921 patients, or 160 of 947 biopsies (16.9%). The three BI-RADS categories examined showed a stepwise correlation with the diagnosis of breast cancer (Table 1). Of the seven cancers diagnosed in BI-RADS category 3, one was LCIS (lobular carcinoma in situ), most likely an incidental finding, while six were DCIS (ductal carcinoma in situ). When both the initial SVABB result and any subsequent positive biopsy results were considered, including all carcinomas as well as cases with atypia, the PPV of BI-RADS category 4 and 5 mammograms was 161 of 791 (20%) in contrast to eight of 156 (5%) for BI-RADS category 3. In order to compare our results in BI-RADS categories 3 5 with other published results using SVABB, we have summarized the available data in Table 2. There were no statistically significant differences when we compared our diagnostic yield on the first SVABB for the various BI-RADS categories with five published series, except for the fact that our 15% positive rate in category 4 was higher than in Berube s series, which had a rate of 4% (P <0.001), and also for category 5 where our positive rate was 79% compared with 54% (P = 0.005), respectively. Our positive rate was not statistically significantly different from those in the other four series. The specificity and sensitivity (see Materials and methods for formulae and definitions) of the BI-RADS categories were determined by classifying category 3 as a negative mammogram and categories 4 and 5 as positive. The sensitivity of mammography,
452 Table 2. Comparison of our series with others from two academic centers and from three community hospitals BI-RADS category Mendez (our series, community hospital) Liberman (academic center) a [7] Berube (academic center) [8] Margolin (community hospital) b [9] a Includes both percutaneous and sonographic guided biopsies. b For comparison purposes we included only cases who had stereobiopsies, and excluded sonographically directed biopsies from this series. c P <0.001 for our series versus Berube; P >0.10 for our series versus Travade, Margolin, Tate and Liberman (analyzed individually). d P <0.005 for our series versus Berube; P = not significant for comparison with the other series. Tate (community hospital) [10] Travade (community hospital) [11] 3 7/156 (4%) 0/2 (0%) 0/16(0%) 1/16 (6%) 2/87 (2%) 0/12 (0%) 4 116/757 (15%) c 25/141 (18%) 16/397 (4%) 39/299 (13%) 159/961 (17%) 23/160 (14%) 5 27/34 (79%) d 46/50 (92%) 43/82(54%) 5/6 (83%) 36/40 (90%) 30/34 (88%) Table 3. Correlation between type of mammographic abnormality and diagnosis of breast cancer Abnormality Biopsies (n) Breast cancer ADH or ALH Total positive findings Microcalcifications 729 105 (14%) 7 (1%) 113 (15%) Asymmetric density 96 15 (16%) 3 (2%) 18 (19%) Circumscribed mass 85 7 (8%) 0 7 (8%) Spiculated mass 35 23 (66%) 0 32 (66%) Nodule + microcalcification 1 0 0 0 Asymmetric density + microcalcification 1 0 0 0 Total 947 150 (16%) 10 (1%) 168 (17%) including all carcinomas as well as cases of atypia diagnosed initially with SVABB or with further follow-up, was 161 of 170 (95%) and specificity was 147 of 784 (19%). However, when specificity was evaluated exclusively for BI-RADS category 5, it was 95%. The specificity and sensitivity were also calculated classifying BI-RADS category 3 with microcalcifications as a positive mammogram (including also BI-RADS 4 and 5 as positive) and category 3 without microcalcifications as negative. Using this definition the sensitivity increased to 168 of 169 (99%), while the specificity dropped to 65 of 702 (9%). If category 3 mammograms with microcalcifications are considered positive, the positive predictive value of a mammogram would also drop to 159 of 881 (18%) from a level of 20% if only categories 4 and 5 are considered as positive mammograms. With the goal of determining if there was any correlation between the description of the lesion observed in the mammogram and the result of the SVABB, we examined each of the major types of lesions, such as spiculated mass, circumscribed mass, microcalcifications and asymmetric density. The results are summarized in Table 3. The correlation between the mammographic features was not as good as with the BI-RADS category, except for the high frequency of cancer in cases with a spiculated mass. However, when BI-RADS category and type of mammographic lesion were combined, some useful patterns emerged (Table 4). Of the 90 cases of BI-RADS category 3 whose major lesion type was characterized by microcalcifications, six (7%) were diagnosed with cancer and another one with atypia (ADH), giving a total of seven of 90 (8%) positive findings in this subcategory. When BI-RADS category 3 cases without microcalcifications were examined separately, only two of 66 cases (3%) were found to have cancer while none had atypia. The only two cases of breast cancer within the BI-RADS category 3 without microcalcifications were a patient with an asymmetric density who had an invasive ductal carcinoma, and another with a circumscribed mass whose SVABB was initially negative but her follow-up mammogram 9 months later changed to a BI-RADS category 4. A biopsy was therefore performed and an invasive colloid carcinoma with an extensive in situ component was diagnosed. We examined those patients with BI-RADS category 4 and a spiculated mass, and we found that 68% had either cancer or ADH, in contrast to 16% for BI-RADS 4 without a spiculated mass (Table 4). Table 5 shows in detail the types of breast cancers diagnosed in this study. The majority of cases (63%) were pure in situ carcinomas, mostly DCIS, followed by invasive carcinomas in 32%. Another 5% of the cases consisted of mixed invasive and in situ carcinomas. Table 6 shows the correlation between the type of mammographic lesion and the histological types of cancer diagnosed by SVABB. A total of 75% of the malignant lesions characterized by microcalcifications were associated with pure DCIS, while 39 of 45 (87%) of the other malignant lesions, including asymmetric densities, circumscribed masses and spiculated masses, were associated with invasive carcinomas. Of the 82
453 Table 4. Cases diagnosed as cancer or atypia by initial SVABB or by any biopsy later on, classified according to combinations of BI-RADS category and type of lesion BI-RADS Microcalcifications Circumscribed mass Spiculated mass a One of these cases with BI-RADS 3 with microcalcifications consisted of atypia. b Six of these cases of BI-RADS 4 with microcalcifications consisted of atypia. c Three of these cases with BI-RADS 4 and asymmetric density consisted of atypia. Asymmetric density Other Total without microcalcifications 3 7/90 (8%) a 1/37 (2.7%) 0/2 1/25 (4%) 0/2 2/66 (3%) 4 97/621 (16%) b 7/48 (15%) 15/22 (68%) 15/66 (23%) c 36/136 (26%) 5 15/18 (83%) 8/11 (73%) 4/5 (80%) 12/16 (75%) cases of DCIS, 79 (96%) were associated with microcalcifications, while 39 of 56 (70%) of the invasive carcinomas were associated with either a mass or an asymmetric density. Discussion Table 5. Pathologic diagnosis of 150 biopsies diagnosed as breast cancer Diagnosis n % Pure in situ 94 63 DCIS 81 54 LCIS 12 8 Mixed DCIS and LCIS 1 0.7 Pure invasive 48 32 Ductal invasive 27 18 Ductulobular 7 5 Invasive lobular 8 5 Invasive tubular 3 2 Invasive tubulobular 2 1 Invasive colloid 1 0.7 Mixed in situ and invasive 8 5 Mixed DCIS and invasive 4 3 Mixed LCIS and invasive 4 3 Total 150 100 DCIS, ductal carcinoma in situ; LCIS, lobular carcinoma in situ. As expected, there is a discrete stepwise increase in the diagnosis of breast cancer following the first SVABB for each of the BI-RADS categories 3 5 (Table 1). The correlation between BI- RADS category and diagnosis of breast cancer was much better than that observed with description of the lesion per se (Table 3). As expected, most of the SVABBs were done for BI-RADS category 4 patients, the most common indicator for the procedure. Biopsy results for category 3 patients are usually benign, which is why SVABBs are usually performed when either the patient is extremely anxious or has a strong family history, or there is a concern about adequate follow-up. The frequency of cancer in category 3 BI-RADS according to the literature ranges from 0.5% to 2% [4, 5]. However, these data are derived from longitudinal follow-ups using mammography for cases with probably benign mammographic lesions. Very little information on the results of percutaneous biopsies is available for BI-RADS category 3. A total of only 133 cases that were biopsied percutaneously are available from the sum of five different series, in contrast to 156 from our single series (Table 2). Our relatively large number of cases allows us to explore the various subgroups within BI-RADS category 3 in more detail. Our data suggest that BI-RADS category 3 is a heterogeneous group. Those without microcalcifications have a low frequency of cancer, since only two of 66 cases (3%) in this subgroup were positive and none of them had atypia (Table 4). On the other hand, of the 90 cases of BI-RADS category 3 whose mammographic lesion was characterized by microcalcifications, six (7%) were diagnosed with cancer and another one was diagnosed with ADH, making a total of seven of 90 (7.8%) positive findings in this subcategory (Table 4). It might be convenient to subdivide BI-RADS category 3 into 3-a (those without microcalcifications) and 3-b (those with microcalcifications). If biopsies are considered unnecessary in category 3-a, this would translate into 66 less biopsies, or a 7% drop in the number of stereotactic biopsies in our center. Obenauer et al. [6] described an 8.1% (seven of 86) incidence of breast cancer or atypia in BI-RADS category 3 with clustered microcalcifications biopsied by means of SVABB. These results are almost identical to ours. Our findings suggest that BI-RADS category 3 with microcalcifications should be interpreted cautiously, and according to our findings a biopsy is indicated. If such cases are interpreted as a positive mammogram, BI-RADS category 4 with a spiculated mass should probably be considered as a category 5 in view of their 68% incidence of cancer (Table 4). As expected with most screening tests, the sensitivity of mammography is excellent (95%) but the specificity is low (19%). This low specificity, however, is an underestimate because only cases with a BI-RADS category of 3 were biopsied. Those cases that fell into categories 1 and 2 were not biopsied and the likelihood that the vast majority of these were true negatives is higher than for category 3. If they were to be biopsied then the specificity of mammograms would most likely be considerably higher than in our study, which was limited to BI-RADS categories 3 5. Another factor that contributes to the relatively low specificity of
454 Table 6. Types of breast cancer identified according to the type of mammographic lesion Type of lesion n Pure DCIS Pure LCIS Invasive ± in situ component Microcalcifications 105 79 (75%) 9 (9%) 17 (16%) Asymmetric density 15 3 (20%) 12 (80%) Circumscribed mass 7 2 (28%) 5 (71%) Spiculated mass 23 1 (4%) 22 (96%) Total 150 82 12 56 stereotactic biopsy in our series is the small size of the tumors biopsied (median diameter 0.9 cm in cases where the lesion did not consist of microcalcifications). Not surprisingly, 75% of the lesions characterized by microcalcifications were associated with pure DCIS and 87% of the other lesions, including asymmetric densities, circumscribed masses and spiculated masses, were associated with invasive carcinomas. Of the 82 cases of DCIS, 79 (96%) were associated with microcalcifications, while 39 of 56 (70%) of the invasive carcinomas were associated with either a mass or an asymmetric density. In summary, the use of BI-RADS categories 3 5 appears to be effective in predicting the chances of finding either cancer or atypia. However, category 3 needs to be better defined, and subdividing it into categories 3-a and 3-b might be useful in terms of defining which patients in that category should be biopsied. Finally, caution is advised in extrapolating these findings to all cases where mammographic lesions are identified, because our study was limited to cases where the breast lesion was not palpable and not visualized by sonography. The results could be different for cases where the biopsy is obtained by sonography, because the frequency of masses rather than microcalcifications will be higher. References 1. Breast imaging reporting and data system (BI-RADS). American College of Radiology 1998. 2. Rotter K, Haentschel G, Koethe D et al. Evaluation of mammographic and clinical follow-up after 755 stereotactic vacuum-assisted breast biopsies. Am J Surg 2003; 186: 134 142. 3. American College of Surgeons and American College of Radiology. Physician qualifications for stereotactic breast biopsy: a revised statement. Bull Am Coll Surg 1998; 83: 30 33. 4. Sickles E. Periodic mammographic follow up of probably benign lesions: Results of 3184 consecutive cases. Radiology 1991; 79: 463 468. 5. Varas X, Leborgne F, Leborgne J. Nonpalpable, probably benign lesions: Role of follow up mammography. Radiology 1992; 184: 409 414. 6. Obenauer S, Fischer U, Baum F et al. Stereotactic vacuum core biopsy of clustered microcalcifications classified as B1-RADS 3. Rofo Fortschr Geb Rontgenstr Neuen Bildgeb Verfahr 2001; 173: 696 701. 7. Liberman L, Abramson AF, Squires FB et al. The breast imaging reporting and data system: positive predictive value of mammographic features and final assessment categories. Am J Roentgenol 1998; 171: 35 40. 8. Berube M, Curpen B, Ugolini P et al. Level of suspicion of a mammographic lesion: use of features defined by BI-RADS lexicon and correlation with large-core breast biopsy. Can Assoc Radiol J 1998; 49: 223 228. 9. Margolin FR, Leung JW, Jacobs RP, Denny SR. Percutaneous imagingguided core breast biopsy: 5 years experience in a community hospital. Am J Roentgenol 2001; 177: 559 564. 10. Tate PS, Rogers EL, McGee EM et al. Stereotactic breast biopsy: a sixyear surgical experience. J Ky Med Assoc 2001; 99: 98 103. 11. Travade A, Isnard A, Bagard C et al. Stereotactic 11-gauge directional vacuum-assisted breast biopsy: experience with 249 patients. J Radiol 2002; 83: 1063 1071.