Learning Curve for Stereotactic Breast Biopsy: How Many Cases Are Enough?

Laura Liberman 1 Catherine L. Benton D. David Dershaw Andrea F. Abramson Linda R. LaTrenta Elizabeth A. Morris Received July 17, 2000; accepted after revision September 7, 2000. Supported by grant C015709 from the New York State Department of Health. 1 All authors: Breast Imaging Section, Department of Radiology, Memorial Sloan-Kettering Cancer Center, 1275 York Ave., New York, NY 10021. Address correspondence to L. Liberman. AJR 2001;176:721 727 0361 803X/01/1763 721 American Roentgen Ray Society Learning Curve for Stereotactic Breast Biopsy: How Many Cases Are Enough? OBJECTIVE. The objective of this study was to evaluate the learning curve for stereotactic breast biopsy. MATERIALS AND METHODS. Retrospective review was performed of 923 consecutive lesions that underwent stereotactic breast biopsy performed by one of six radiologists. Four hundred fourteen lesions had 14-gauge automated core biopsy, and 509 subsequent lesions had vacuum-assisted biopsy (14-gauge in 163 and 11-gauge in 346). Medical records were reviewed to determine the technical success rate and false-negative rate as a function of operator experience. RESULTS. For 14-gauge automated core biopsy, a significantly lower technical success rate was seen for the first five cases of each radiologist than for subsequent cases (25/30 = 83.3% versus 366/384 = 95.3%, p < 0.02) and for the first 20 cases than for subsequent cases (90/100 = 90% versus 284/296 = 95.9%, p < 0.05). For 11-gauge vacuum-assisted biopsy, a significantly lower technical success rate was seen for the first five cases than for subsequent cases (17/20 = 85.0% versus 310/322 = 96.3%, p < 0.05) and for the first 15 cases than for subsequent cases (54/60 = 90.0% versus 273/283 = 96.5%, p = 0.03). The false-negative rate was higher for the first 15 cases compared with subsequent cases both for stereotactic 14- gauge automated core biopsy (4/31 = 12.9% versus 3/115 = 2.6%, p < 0.04) and for stereotactic 11-gauge vacuum-assisted biopsy (2/27 = 7.4% versus 0/85 = 0%, p < 0.06). CONCLUSION. A learning curve exists for stereotactic breast biopsy. Significantly higher technical success rates and lower false-negative rates were observed after the first five to 20 cases for 14-gauge automated core biopsy and after the first five to15 cases for 11-gauge vacuum-assisted biopsy. Even after a radiologist has experience with stereotactic biopsy, changes in equipment may result in a new learning curve. S tereotactic breast biopsy is being increasingly used as an alternative to surgical biopsy for the evaluation of breast lesions [1 5]. Several aspects of stereotactic breast biopsy require learning, even for individuals with experience in breast imaging; these aspects include lesion targeting using stereotactic images and the mechanical use of different tissue acquisition devices. Few data address the learning curve for stereotactic breast biopsy [6 8]. This study was undertaken to determine the technical success rate and false-negative rate of stereotactic breast biopsy as a function of operator experience. Materials and Methods Biopsy Technique Retrospective review was performed of 923 consecutive lesions that underwent stereotactic biopsy performed by one of six radiologists at our institution from August 1992 through June 1998. Biopsies were performed with patients prone on a dedicated table using digital imaging (LoRad with Digital Spot Mammography; LoRad, Danbury, CT) with a 14- gauge automated needle (Automatic Cutting Needle, Manan Medical Products, Northbrook, IL; Bard Biopty-Cut, Bard Urological, Covington, GA; or Ultra-Core Biopsy Needle, Medical Device Technologies, Gainesville, FL) in 414 lesions, a 14-gauge directional vacuum-assisted biopsy device (Mammotome, Biopsys/Ethicon Endo-Surgery, Cincinnati, OH) in 163 lesions, and an 11-gauge directional vacuum-assisted biopsy device (Biopsys/Ethicon Endo-Surgery) in 346 lesions. From August 1992 to December 1995, all 414 stereotactic biopsies were performed using a 14-gauge automated needle with a long excursion gun (Manan Pro-Mag 2.2, Manan; or Biopty, Bard Urological) in 404 (97.6%) of 414 lesions and a short excursion gun (Manan Pro-Mag 1.2) in 10 (2.4%) of 414 lesions. From December 1995 to October 1996, all 139 stereo- AJR:176, March 2001 721

Liberman et al. tactic biopsies were performed with a 14-gauge directional vacuum-assisted biopsy probe (Mammotome, Biopsys/Ethicon Endo-Surgery). From October 1996 to June 1998, stereotactic biopsies were performed with the 11-gauge directional vacuum-assisted biopsy probe in 93.5% (346/370) of lesions and with the 14- gauge directional vacuum-assisted biopsy probe in 6.5% (24/370) of lesions, at the discretion of the radiologist performing the biopsy. For stereotactic 14-gauge automated core biopsy, the median number of specimens obtained per lesion was five (range, 1 22 specimens); five or more specimens were obtained in 404 (97.6%) of 414 lesions [9]. For stereotactic 14-gauge directional vacuum-assisted biopsy, the median number of specimens obtained per lesion was 14 (range, 1 50 specimens); eight or more specimens were obtained in 149 (91.4%) of 163 lesions [10]. For stereotactic 11-gauge directional vacuum-assisted biopsy, the median number of specimens obtained per lesion was 15 (range, 4 43 specimens); eight or more specimens were obtained in 355 (97.5%) of 364 lesions [11]. For lesions evident as calcifications, specimen radiography was performed to confirm calcification retrieval [12, 13]. Radiologist Experience Each biopsy was performed by one of six attending radiologists at an academic institution where more than 30,000 mammograms were performed annually during the study period. Five of the six radiologists specialized in breast imaging and one specialized in sonography. At the outset of their participation in the study, three of the radiologists had a median of 7 years of breast imaging experience (individual levels of experience were 2, 7, and 11 years), one had 4 years of experience as an interventional radiologist, one had completed a fellowship in breast and body imaging, and one had completed residency training. Five of the radiologists had no prior clinical experience with stereotactic breast biopsy before performing biopsies as attending radiologists at our hospital, and one had participated in approximately 30 stereotactic 14-gauge automated core breast biopsies during fellowship training under the direct supervision of an attending radiologist. All radiologists participated in hands-on training sessions using phantoms before performing clinical stereotactic biopsies with each new tissue acquisition device. These training sessions included at least one 0.5-hr group lesson with an applications specialist and additional individual sessions at the discretion of the radiologist. The median number of stereotactic biopsy procedures performed per radiologist during the study period of approximately 6 years was 180 (range, 30 295). Management After Biopsy Management after stereotactic biopsy was guided by the following protocol. If stereotactic biopsy yielded benign findings concordant with the imaging characteristics, the patient was referred for follow-up mammography, usually at 1 year. If stereotactic biopsy yielded carcinoma, the patient was referred for definitive treatment. Surgical excision was recommended if it was suggested by the pathologist; if a specific disorder was encountered such as atypical ductal hyperplasia, fibroepithelial lesions thought to represent phyllodes tumors, or high-risk lesions such as radial scars; or if there was discordance between histologic and imaging findings [14 18]. Technical Success Technical success was defined as retrieving calcifications documented by specimen radiography (for calcifications) or obtaining a histologic diagnosis concordant with imaging features (for masses). The technical success rate as a function of operator experience was calculated: cumulatively for the group as a whole, by summing the first n cases of each radiologist, with n ranging from one to 30, for all radiologists who had performed at least n cases; cumulatively for each individual radiologist; and sequentially for each individual radiologist, by determining technical success rates in sequential groups of five cases. False-Negative Rate The false-negative rate was defined as the proportion of pathologically proven cancerous lesions that yielded benign stereotactic biopsy findings without atypia or high-risk lesions [7, 17]. Follow-up data were obtained to calculate the false-negative rate. For 14- gauge automated core biopsy, surgical (n = 188) or imaging (n = 171) follow-up data were available in 359 (86.7%) of 414 lesions; the median duration of imaging follow-up was 42 months (range, 6 87 months). For 14-gauge directional vacuum-assisted biopsy, surgical (n = 67) or imaging (n = 64) follow-up data were available in 131 (80.4%) of 163 lesions; the median duration of imaging follow-up was 36 months (range, 6 50 months). For 11-gauge directional vacuum-assisted biopsy, surgical (n = 150) or imaging (n = 147) follow-up data were available in 297 (85.8%) of 346 lesions; the median duration of imaging follow-up was 24 months (range, 6 42 months). Data Analysis Data were prospectively entered into a database using spreadsheet software (Excel; Microsoft, Redmond, WA). Medical records were reviewed to determine the technical success rate and false-negative rate as a function of operator experience. Tests for statistical significance were performed with the chi-square and Fisher s exact tests using statistics software (Epi- Info; Centers for Disease Control, Atlanta, GA). TABLE 1 Outcome Note. Numbers in parentheses are percentages. Results Overall Outcome Outcome as a Function of Biopsy Method 14-Gauge Automated Overall outcome as a function of biopsy method is shown in Table 1, and the technical success rate for each biopsy method for each radiologist is shown in Table 2. A trend was seen toward a higher calcification retrieval rate with 11-gauge directional vacuum-assisted biopsy than with 14-gauge automated core biopsy (246/256 = 96.1% versus 131/146 = 89.7%; p < 0.06; odds ratio = 2.44 [95% confidence interval, 0.96 6.39]). Another trend was seen toward a lower false-negative rate with 11- or 14-gauge directional vacuum-assisted biopsy than with 14-gauge automated core biopsy (2/171 = 1.1% versus 7/146 = 4.8%; p < 0.09; odds ratio = 0.23 [95% confidence interval, 0.02 1.27]). No significant differences were observed in the overall technical success rate or mass concordance rate as a function of biopsy method. 14-Gauge Automated Core Breast Biopsy The learning curves for stereotactic 14-gauge automated core breast biopsy are shown in Figure 1. Comparison of technical success in early versus late experience for the group as a whole is shown in Table 3, and the sequential experience of each individual radiologist performing stereotactic 14-gauge automated core biopsy is shown in Table 4. For stereotactic 14-gauge automated core biopsy, a significantly lower technical success rate was observed for the first 20 cases of each radiologist as compared with subsequent cases (Table 3). A trend toward a higher technical success rate was noted in radiologists who performed an average of 12 or more cases per year than among radiologists who performed fewer than 12 cases (328/344 = 95.3% versus 63/70 = 90%; p < 0.09; odds ratio = 2.28 [95% confidence interval, 0.76 6.13]). A steeper learning curve was observed for calcifications than for masses. The technical success rate for calcifications was 50.0% for the first calcification case of each radiologist, 14-Gauge Vacuum- Assisted 11-Gauge Vacuum- Assisted Technical success rate Calcifications 131/146 (89.7) 80/86 (93.0) 246/256 (96.1) Masses 260/268 (97.0) 76/77 (98.7) 85/90 (94.4) All lesions 391/414 (94.4) 156/163 (95.7) 331/346 (95.7) False-negative rate 7/146 (4.8) 1/58 (1.7) 2/113 (1.8) 722 AJR:176, March 2001

Learning Curve for Breast Biopsy 72.2% for the first three cases, 80.0% for the first five cases, 93.3% for the first 10 cases, and 95.6% for the first 15 cases (Fig. 1B). The technical success rate for masses was 83.3% for the first mass case of each radiologist, 88.9% for the first three cases, 90.0% for the first five cases, 91.7% for the first 10 cases, and 94.7% for the first 15 cases (Fig. 1C). Technical failures occurred in 23 (5.6%) of 414 lesions that had stereotactic 14-gauge automated core biopsy, including 15 (10.3%) of 146 calcific lesions and eight (3.0%) of 268 masses. Subsequent surgical excision, performed in 15 of these technical failures, revealed carcinoma in 11 lesions (73.3%), including eight (88.9%) of nine calcific lesions and three (50%) of six masses. Stereotactic biopsy histologic findings in these 11 lesions were atypical ductal hyperplasia in four and benign lesion in seven; surgical histologic findings were ductal carcinoma in situ in six and infiltrating carcinoma in five. The remaining eight lesions in which technical failure occurred remained stable at mammographic follow-up (range, 12 69 months; median, 26 months). Among 146 pathologically proven cancerous lesions that underwent stereotactic 14-gauge automated core biopsy, false-negative results were Fig. 1. Bar graphs show learning curves for lesions that underwent stereotactic 14-gauge automated core biopsy. Data reflect cumulative technical success rate for first n cases of each radiologist, with all radiologists considered as a group. A, Learning curve for all lesions. Note lower initial technical success rate and plateau reached at 15 20 cases. B, Learning curve for calcifications. Note low initial technical success rate and steep slope of curve. C, Learning curve for masses. Initial technical success was higher for masses than for calcifications, and slope is less steep. encountered in seven (4.8%). All seven falsenegative cases were identified promptly because of failure to sample calcifications (n = 4) or discordant results in a mass lesion (n = 3). In four of these false-negative cases, surgery revealed infiltrating carcinoma (ductal in two and lobular in two) with a median size of 0.8 cm (range, 0.6 1.0 cm); axillary lymph node dissection, performed in three of these four infiltrating carcinomas, revealed axillary metastases in one. In the remaining three false-negative cases, surgery revealed ductal carcinoma in situ. The TABLE 2 Radiologist A false-negative rate was significantly higher for the first 10 cases of each radiologist than for subsequent cases (4/20 = 20.0% versus 3/126 = 2.4%; p < 0.01; odds ratio = 10.25 [95% confidence interval, 1.54 74.30]) and for the first 15 cases than for subsequent cases (4/31 = 12.9% versus 3/115 = 2.6%; p < 0.04; odds ratio = 5.53 [95% confidence interval, 0.87 39.37]). 14-Gauge Directional Vacuum-Assisted Breast Biopsy For stereotactic 14-gauge vacuum-assisted biopsy, a trend was noted toward a lower Overall Technical Success Rates for Individual Radiologists 14-Gauge Automated Technical Success Rate 14-Gauge Vacuum-Assisted 11-Gauge Vacuum-Assisted Note. Numbers in parentheses are percentages. NA = not applicable. a This radiologist had previously participated in approximately 30 stereotactic biopsies under supervision during fellowship training. Total A 144/151 (95.4) 33/34 (97.1) 106/110 (96.4) 283/295 (95.9) B 102/107 (95.3) 43/45 (95.6) 48/51 (94.1) 193/203 (95.1) C a 17/18 (94.4) 41/43 (95.3) 120/123 (97.6) 178/184 (96.7) D 82/86 (95.3) 30/32 (93.8) 53/58 (91.4) 165/176 (93.8) E 21/22 (95.5) 9/9 (100) 4/4 (100) 34/35 (97.1) F 25/30 (83.3) NA NA 25/30 (83.3) Total 391/414 (94.4) 156/163 (95.7) 331/346 (95.7) 878/923 (95.1) B C AJR:176, March 2001 723

Liberman et al. technical success rate for the first 10 cases of each radiologist than for subsequent cases (36/40 = 90% versus 111/114 = 97.4%, p < 0.08). A lower (but not significantly lower) technical success rate was observed for the first 15 cases of each radiologist than for TABLE 3 subsequent cases (55/60 = 91.7% versus 92/ 94 = 97.9%, p = 0.11). A learning curve was observed for calcifications: the technical success rate was 80.0% for the first five calcification cases of each radiologist, 87.5% for the first 10 cases, and 91.7% Technical Success Rates in Early Versus Late Experience No. of Cases First n Cases After n Cases p Odds Ratio a [95% CI] 14-Gauge Automated Core Biopsy 3 14/18 (77.8) 377/396 (95.2) <0.02 5.67 [1.23 20.35] 5 25/30 (83.3) 366/384 (95.3) <0.02 4.07 [1.08 12.60] 10 51/60 (85) 340/354 (96.0) <0.003 4.29 [1.54 11.22] 12 63/72 (87.5) 328/342 (95.9) <0.01 3.35 [1.22 8.69] 15 80/90 (88.9) 311/324 (96.0) <0.02 2.99 [1.13 7.67] 20 90/100 (90.0) 284/296 (95.9) <0.05 2.63 [0.98 6.87] 25 91/100 (91.0) 262/274 (95.6) 0.14 2.16 [0.77 5.78] 30 111/120 (92.5) 242/254 (95.3) 0.39 1.64 [0.59 4.36] 11-Gauge Vacuum-Assisted Biopsy 3 13/15 (86.7) 318/331 (96.1) 0.13 3.76 [0.37 19.45] 5 17/20 (85.0) 310/322 (96.3) <0.05 4.56 [0.75 19.12] 10 36/40 (90.0) 291/302 (96.4) <0.09 2.94 [0.65 10.56] 12 44/48 (91.7) 283/294 (96.3) 0.24 2.34 [0.52 8.32] 15 54/60 (90.0) 273/282 (96.8) <0.04 3.37 [0.94 11.06] 20 74/80 (92.5) 253/262 (96.6) 0.12 2.28 [0.64 7.42] 25 94/100 (94.0) 233/242 (96.3) 0.38 1.65 [0.47 5.36] 30 114/120 (95.0) 213/222 (95.9) 0.35 1.87 [0.49 7.15] Note. Data reflect cumulative technical success rates for the first n cases as compared with subsequent cases for the group as a whole. Numbers in parentheses are percentages. CI = confidence interval. a For comparison of late versus early experience. TABLE 4 Radiologist Sequential Technical Success Rates for Individual Radiologists No. of Successful Biopsies/No. of Biopsies Performed for the first 15 cases. No appreciable learning curve was observed for mass lesions, in which the technical success rate was 100% for the first five, 10, and 15 mass cases, respectively, of each radiologist. Technical failures occurred in seven (4.3%) of 163 lesions that underwent stereotactic 14- gauge vacuum-assisted biopsy, including six (7.0%) of 86 calcific lesions and one (1.3%) of 77 masses. Surgical excision, performed in five technical failures (four calcific lesions and one mass), revealed carcinoma in one lesion (20%). Of the remaining two technical failures, followup is pending in one; the other showed stable mammographic findings at 50 months. Among 58 pathologically proven cancerous lesions that underwent stereotactic 14-gauge directional vacuum-assisted biopsy, false-negative results were encountered in one (1.7%). This was a calcific lesion yielding benign findings with no calcifications at stereotactic biopsy; surgery revealed infiltrating ductal carcinoma measuring 2 mm and ductal carcinoma in situ with negative findings in nodes. The false-negative rate was higher among the first 15 cases of each radiologist than among subsequent cases (1/21 = 4.8% versus 0/31 = 0%, p = 0.4), but this difference did not achieve statistical significance. 11-Gauge Directional Vacuum-Assisted Breast Biopsy The learning curves for stereotactic 11- gauge directional vacuum-assisted biopsy are shown in Figure 2. Comparison of technical 1st Five 2nd Five 3rd Five 4th Five 5th Five 6th Five 14-Gauge Automated Core Biopsy A 5/5 (100) 3/5 (60) 5/5 (100) 5/5 (100) 5/5 (100) 5/5 (100) B 5/5 (100) 5/5 (100) 5/5 (100) 5/5 (100) 5/5 (100) 5/5 (100) C 4/5 (80) 5/5 (100) 5/5 (100) NA NA NA D 4/5 (80) 4/5 (80) 5/5 (100) 5/5 (100) 5/5 (100) 5/5 (100) E 4/5 (80) 5/5 (100) 5/5 (100) 5/5 (100) NA NA F 3/5 (60) 4/5 (80) 4/5 (80) 4/5 (80) 5/5 (100) 5/5 (100) Total 25/30 (83.3) 26/30 (86.7) 29/30 (96.7) 24/25 (96.0) 20/20 (100) 20/20 (100) 11-Gauge Vacuum-Assisted Biopsy a A 5/5 (100) 5/5 (100) 5/5 (100) 5/5 (100) 5/5 (100) 5/5 (100) B 5/5 (100) 4/5 (80) 5/5 (100) 5/5 (100) 5/5 (100) 5/5 (100) C 4/5 (80) 5/5 (100) 5/5 (100) 5/5 (100) 5/5 (100) 5/5 (100) D 3/5 (60) 5/5 (100) 3/5 (60) 5/5 (100) 5/5 (100) 5/5 (100) Total 17/20 (85) 19/20 (95) 18/20 (90) 20/20 (100) 20/20 (100) 20/20 (100) Note. Data reflect the technical success rate of individual radiologists in sequential groups of five cases. Numbers in parentheses are percentages. NA = not applicable. a Data from radiologist E, who had technical success in 4/4 (100%) 11-gauge vacuum-assisted biopsies, are excluded from this table. 724 AJR:176, March 2001

Learning Curve for Breast Biopsy Fig. 2. Bar graphs show learning curves for lesions that had stereotactic, 11- gauge vacuum-assisted biopsy. Data reflect cumulative technical success rate for first n cases of each radiologist, with all radiologists considered as a group. A, Learning curve for all lesions. Note initial technical success rate is higher for 11- gauge vacuum-assisted biopsy than for 14-gauge automated core biopsy (Fig. 1A), but technical success still improves with experience. B, Learning curve for calcifications. Note low initial technical success rate and steep slope of curve. C, Learning curve for masses. Note high initial technical success with no appreciable curve. success in early versus late experience for the group as a whole is shown in Table 3, and the sequential experience of each individual radiologist performing 11-gauge vacuum-assisted biopsy is shown in Table 4. For 11-gauge vacuum-assisted biopsy, a significantly lower technical success rate was observed for the first five to 15 cases of each radiologist than for subsequent cases (Table 3). A learning curve was more evident for calcifications than for mass lesions. Among calcification lesions, the technical success rate was 60.0% for the first calcification case of each radiologist, 75.0% for the first three cases, 85.0% for the first five cases, 90.0% for the first 10 cases, and 93.3% for the first 15 cases (Fig. 2B). Among mass lesions, the technical success rate was 100% for the first mass case, 95.0% for both the first five and the first 10 cases, and 97.8% for the first 15 cases (Fig. 2C). Technical failures occurred in 15 (4.3%) of 346 lesions that had stereotactic 11-gauge directional vacuum-assisted biopsy, including 10 (3.9%) of 256 calcific lesions and five (5.6%) of 90 masses. Surgical excision, performed in 10 technical failures, revealed carcinoma in three (30%), including two (50%) of four masses and one (16.7%) of six calcific lesions. In the calcific lesion, which yielded atypical ductal hyperplasia at stereotactic biopsy, surgery revealed infiltrating ductal carcinoma and ductal carcinoma in situ; in the two masses, which yielded benign findings at stereotactic biopsy, ductal carcinoma in situ was found at surgery. Of the remaining five lesions in which technical failure occurred, follow-up data are available in four: one had a successful second stereotactic biopsy with benign results, and three have remained stable at mammographic follow-up (range, 14 27 months; median, 24 months). Among 113 pathologically proven cancerous lesions that underwent stereotactic 11-gauge directional vacuum-assisted breast biopsy, falsenegative results were encountered in two (1.8%). Both lesions were evident as areas of architectural distortion on mammography and yielded benign stereotactic biopsy results that were considered discordant with the imaging features. Prompt surgical excision was suggested in both cases; both lesions yielded a radial scar and ductal carcinoma in situ. The first 15 cases of each radiologist as compared with subsequent cases had a trend toward a higher false-negative rate (2/28 = 7.1% versus 0/85 = 0%, p < 0.06). Discussion A learning curve exists for most endeavors in life. In the field of medicine, learning curves A have been shown for a wide variety of procedures ranging from surgical knot-tying to sentinel lymphadenectomy [19 23]. Many studies analyzing learning curves for different procedures have shown improved outcome with experience. Understanding the learning curve for a particular procedure may help us determine how we can ascend the learning curve while providing high-quality care [22, 23]. Previous investigators have suggested that a learning curve exists for stereotactic breast biopsy. In a multiinstitutional study of stereotactic 14-gauge automated core biopsy, Brenner et al. [6] found a higher frequency of failing to diagnose the lesion during early experience (defined as the first 20 cases per institution) than during later experience (3/37 = 8% versus 4/142 = 3%, p = 0.16). In a study of lesions not diagnosed at stereotactic 14-gauge automated core biopsy, Liberman et al. [7] found that four (80%) of five cases in which inaccurate needle placement led to failure to diagnose cancer occurred during the first 9 months of experience. In a study of stereotactic 14-gauge directional vacuum-assisted biopsy, Liberman et al. [8] found a higher frequency of failure to retrieve calcifications during the first 4 months of experience than during the second 4 months (5/24 = 21% versus 0/31 = 0%, p = 0.01). B C AJR:176, March 2001 725

Liberman et al. Our data support the existence of a learning curve for stereotactic breast biopsy with the automated needle and with the directional vacuum-assisted biopsy probes. We observed a significantly lower technical success rate and a higher false-negative rate among the first five to 20 cases than with subsequent cases for stereotactic 14-gauge automated core biopsy, and among the first five to 15 cases as compared with subsequent cases for stereotactic 11- gauge directional vacuum-assisted biopsy. For both methods, a technical success rate of greater than 95% was achieved after the first five cases for the group of radiologists as a whole; however, examination of the data of the individual radiologists indicates that the performance of the individual radiologists showed further improvement after the first five to 20 cases for 14-gauge automated core biopsy and after the first five to 15 cases for 11-gauge vacuum-assisted biopsy. Our findings have several implications. First, training programs should endeavor to allow trainees to perform an adequate number of biopsies under supervision to achieve sufficiently high technical success and sufficiently low false-negative findings. Our data suggest that this number may be between five and 20 cases, but further work is needed to determine if these results can be generalized. Second, a better outcome may be achieved if stereotactic biopsy is performed by a small number of individuals with more experience rather than by a large number of individuals with less experience. Third, when embarking on the performance of stereotactic biopsy, close supervision by an experienced individual during the early cases may be desirable. Finally, the learning curve, although present, is relatively short. That stereotactic biopsy can be learned relatively rapidly contributes to its potential usefulness in general clinical practice. We encountered a learning curve for stereotactic 11-gauge vacuum-assisted biopsy among our radiologists who had a collective experience of more than 500 prior stereotactic 14-gauge automated core or 14-gauge vacuum-assisted biopsy procedures. This finding engenders respect for changes in technology. Even for radiologists experienced at performing stereotactic biopsy, new equipment poses new challenges. With the introduction of new devices, hands-on training sessions with phantoms, as pioneered in the work of Georgian- Smith et al. [24], may be helpful. For stereotactic biopsy, lesion targeting can be practiced using commercially available phantoms (e.g., Stereotactic Breast Biopsy Phantom Model 164; Gammex/RMI, Middleton, WI), and the process of tissue acquisition can be practiced using a home-built phantom, such as a pepper stuffed with pizza dough. We found a steeper learning curve for stereotactic biopsy of calcifications than for masses, with a lower technical success rate in the early calcification cases than in early mass cases. The technical difficulties encountered during stereotactic biopsy of calcifications have been extensively discussed in the literature and relate to the target being tiny and discontinuous [25 27]. Our observed differences in learning curves for calcifications and masses may also in part reflect our definition of technical success: calcification retrieval (for calcific lesions) is a more objective and stringent criterion for success than imaging and histologic concordance (for masses). We found, as have others [26, 27], that the use of stereotactic 11-gauge directional vacuum-assisted biopsy rather than 14- gauge automated core biopsy resulted in a higher technical success rate for the biopsy of calcifications, but did not eliminate the learning curve. A higher false-negative rate was observed during early than during late experience. Our false-negative findings were identified promptly on review of stereotactic biopsy findings because of the failure to retrieve calcifications or imaging histologic discordance. In previous studies, imaging histologic discordance has been reported in 0.9 6.2% of lesions undergoing stereotactic breast biopsy; of lesions yielding discordant results, subsequent excision revealed carcinoma in 0 63.6% [14 18]. Correlation of imaging and histologic findings is important to prevent a deleterious delay in the diagnosis of breast cancer and should be emphasized during training. The false-negative rate may be minimized by close coordination with more experienced operators while ascending the learning curve. To initially qualify to perform stereotactic breast biopsy under the accreditation program of the American College of Radiology, the physician is required to have performed at least 12 stereotactic breast biopsies, or at least three hands-on stereotactic breast biopsy procedures under the supervision of a physician who is qualified to interpret mammography under the Mammography Quality Standards Act and who has performed at least 24 stereotactic biopsies [28]. To maintain accreditation, a physician is required to perform at least 12 stereotactic breast biopsies per year or to refulfill the initial requirements. Our findings support both the usefulness of establishing guidelines for the number of cases required for initial accreditation and the specific requirements of the American College of Radiology. The guideline of 12 cases for initial qualification is within the five to 20 range suggested by our data, and the guideline of 24 cases for those providing supervision corresponds to the plateau of the learning curve (Figs. 1 and 2). In summary, a learning curve exists for stereotactic breast biopsy, but it is relatively short. In our experience, significantly higher technical success rates and lower false-negative rates were achieved after the first five to 20 cases for stereotactic 14-gauge automated core biopsy and after the first five to 15 cases for stereotactic 11-gauge vacuum-assisted biopsy. Even after experience with stereotactic breast biopsy, changes in equipment can result in a new learning curve. The learning curves reported here are those of academic radiologists specializing in breast imaging. Further work is necessary to determine the learning curves for stereotactic breast biopsy in other settings, to assess the number of cases required to maintain proficiency, to determine the impact of training and supervision on the learning curve, and to evaluate the learning curves for new breast biopsy technology as it becomes available. Acknowledgments We thank Hiram S. Cody III for critical review of the manuscript and David C. Perlman for invaluable assistance. References 1. Parker SH, Lovin JD, Jobe WE, et al. Stereotactic breast biopsy with a biopsy gun. Radiology 1990;76:741 747 2. Parker SH, Lovin JD, Jobe WE, Burke BJ, Hopper KD, Yakes WF. Nonpalpable breast lesions: stereotactic automated large-core biopsies. Radiology 1991;180:403 407 3. Elvecrog EL, Lechner MC, Nelson MT. Nonpalpable breast lesions: correlation of stereotaxic large-core needle biopsy and surgical biopsy results. Radiology 1993;188:453 455 4. Gisvold JJ, Goellner JR, Grant CS, et al. Breast biopsy: a comparative study of stereotaxically guided core and excisional techniques. AJR 1994;162:815 820 5. Liberman L. Percutaneous imaging-guided core breast biopsy: state of the art at the millennium. AJR 2000;174:1191 1199 6. Brenner RJ, Fajardo L, Fisher PR, et al. Percutaneous core biopsy of the breast: effect of operator experience and number of samples on diagnostic accuracy. AJR 1996;166:341 346 7. Liberman L, Dershaw DD, Glassman JR, et al. Analysis of cancers not diagnosed at stereotactic core breast biopsy. Radiology 1997;203:151 157 8. Liberman L, Hann LE, Dershaw DD, Morris EA, Abramson AF, Rosen PP. Mammographic find- 726 AJR:176, March 2001

Learning Curve for Breast Biopsy ings after stereotactic 14-gauge vacuum biopsy. Radiology 1997;203:343 347 9. Liberman L, Dershaw DD, Rosen PP, Abramson AF, Deutch BM, Hann LE. Stereotaxic 14-gauge breast biopsy: how many core biopsy specimens are needed? Radiology 1994;192:793 795 10. Burbank F, Parker SH, Fogarty TJ. Stereotactic breast biopsy: improved tissue harvesting with the Mammotome. Am Surg 1996;62:738 744 11. Parker SH, Klaus AJ. Performing a breast biopsy with a directional, vacuum-assisted biopsy instrument. RadioGraphics 1997;17:1233 1252 12. Meyer JE, Lester SC, Frenna TH, White FV. Occult breast calcifications sampled with large-core biopsy: confirmation with radiography of the specimen. Radiology 1993;188:581 582 13. Liberman L, Evans WP III, Dershaw DD, et al. Radiography of microcalcifications in stereotaxic mammary core biopsy specimens. Radiology 1994;190:223 225 14. Dershaw DD, Morris EA, Liberman L, Abramson AF. Nondiagnostic stereotaxic core breast biopsy: results of rebiopsy. Radiology 1996;198:323 325 15. Meyer JE, Smith DN, Lester SC, et al. Large-needle core biopsy: nonmalignant breast abnormalities evaluated with surgical excision or repeat core biopsy. Radiology 1998;206:717 720 16. Philpotts LE, Shaheen NA, Carter D, Lange RC, Lee CH. Comparison of rebiopsy rates after stereotactic core needle biopsy of the breast with 11- gauge vacuum suction probe versus 14-gauge needle and automatic gun. AJR 1999;172:683 687 17. Jackman RJ, Nowels KW, Rodriguez-Soto J, Marzoni FA, Finkelstein SI, Shepard MJ. Stereotactic, automated, large-core needle biopsy of nonpalpable breast lesions: false-negative and histologic underestimation rates after long-term follow-up. Radiology 1999;210:799 805 18. Liberman L, Drotman M, Morris EA, et al. Imaging-histologic discordance at percutaneous breast biopsy: an indicator of missed cancer. Cancer 2000;89:2538 2546 19. Flood AB, Scott WR, Ewy W. Does practice make perfect? I. The relation between hospital volumes and outcome for selected diagnostic categories. Med Care 1984;22:98 114 20. Vossen C, Van Ballaer P, Shaw RW, Koninckx PR. Effect of training on endoscopic intracorporeal knot tying. Hum Reprod 1997;12:2658 2663 21. Cody HS III, Hill ADK, Tran KN, Brennan MF, Borgen PI. Credentialing for breast lymphatic mapping: how many cases are enough? Ann Surg 1999;229:723 728 22. Hatlie MJ. Climbing the learning curve : new technologies, emerging obligations. (editorial) JAMA 1993;270:1364 1365 23. Gates EA. New surgical procedures: can our patients benefit while we learn? Am J Obstet Gynecol 1997;176:1293 1299 24. Georgian-Smith D, Lyon RD, Schwartzberg BS, Shiels WE II. A smorgasbord of interventional phantoms. In: Mueller PRR, ed.; Nemceck AA, guest ed. Seminars in Interventional Radiology, vol. 14. The use of ultrasonography in interventional radiology. New York: Thieme Medical, 1997:377 387 25. Liberman L, Smolkin JH, Dershaw DD, Morris EA, Abramson AF, Rosen PP. Calcification retrieval at stereotactic, 11-gauge, directional vacuum-assisted breast biopsy. Radiology 1998;208: 251 260 26. Jackman RJ, Burbank FH, Parker SH, et al. Accuracy of sampling microcalcifications by three stereotactic breast biopsy methods. (abstr) Radiology 1997;205(P):325 27. Reynolds HE, Poon CM, Goulet RJ, Lazaridis CL. Biopsy of breast microcalcifications using an 11-gauge directional vacuum-assisted device. AJR 1998;171:611 613 28. American College of Radiology Commission on Standards and Accreditation. Stereotactic breast biopsy accreditation program introductory memorandum. Reston, VA: American College of Radiology, 1998 AJR:176, March 2001 727