Role of Breast MRI in the Preoperative Evaluation of Patients with Newly Diagnosed Breast Cancer

Transcription

1 Women s Imaging Original Research Schell et al. Preoperative Breast MRI Women s Imaging Original Research WOMEN S IMAGING Amy M. Schell 1 Kari Rosenkranz 2 Petra J. Lewis 3 Schell AM, Rosenkranz K, Lewis PJ Keywords: breast, breast cancer, contralateral breast, ipsilateral breast, MRI, staging DOI: /AJR Received July 16, 2008; accepted after revision November 4, The data collection and analysis for this study were funded in part by a grant from the Prouty Pilot Projects (Norris Cotton Cancer Center at Dartmouth Hitchcock Medical Center). 1 Department of Medicine, Dartmouth Hitchcock Medical Center, Lebanon, NH. 2 Department of Surgery, Dartmouth Hitchcock Medical Center, Lebanon, NH. 3 Department of Radiology, Dartmouth Hitchcock Medical Center, 1 Medical Center Dr., Lebanon, NH Address correspondence to P. J. Lewis (petra. lewis@hitchcock.org). AJR 2009; 192: X/09/ American Roentgen Ray Society Role of Breast MRI in the Preoperative Evaluation of with Newly Diagnosed Breast Cancer OBJECTIVE. The purpose of our study was to evaluate in a noninvestigational setting the effect of preoperative breast MRI on the diagnosis of unsuspected additional malignant tumors and on surgical management choices. SUBjectS AND METHODS. One hundred ninety-nine patients with a new diagnosis of breast cancer who underwent preoperative bilateral breast MRI were enrolled. Data included additional imaging and biopsies performed, final surgical management, and pathologic correlation. To reduce bias due to patient preference and variable surgical approaches, ipsilateral malignant lesions found on breast MR images were retrospectively reviewed by a surgeon blinded to final surgical treatment. RESULTS. MRI detected additional suspicious previously unsuspected lesions in 74 patients (37%). Fifty-four of these lesions, in 38 patients (19%), were identified as malignant, of which 41 (76%) were invasive. Retrospective review of the MRI-identified ipsilateral malignant lesions resulted in hypothetical recommendations that would have altered the surgical treatment of 26 of the patients (13%), principally as mastectomy or wider excision. CONCLUSION. For patients with a new diagnosis of breast cancer, breast MRI in a mid-sized regional hospital depicts unsuspected malignant lesions in both the ipsilateral and contralateral breasts in proportions consistent with the results of earlier studies at larger institutions. Whether clinical outcome is improved by changes in surgical management consequent to MRI detection of unsuspected malignant lesions remains unproven. B reast MRI is increasingly being used for the preoperative evaluation of patients with newly diagnosed breast cancer. It has been found in several studies to depict unsuspected ipsilateral additional foci of malignancy in 6 27% of patients [1 5] and contralateral unsuspected malignant lesions in 3 9% [1, 2, 5 9]. In studies in which the comparison has been made, breast MRI appears to depict more malignant lesions than either mammography or ultrasound. The routine use of preoperative breast MRI has not been without controversy, however, and the effect on long-term survival has not been assessed to our knowledge. The false-positive rate of MRI is high and varies markedly among studies (9 78%) [10 12]. Additional imaging, such as mammography and ultrasound, is frequently needed, often resulting in additional biopsies. The effect on surgical management is not always evaluated for example, small additional lesions close to the primary tumor often are excised anyway during routine partial mastectomy. Interpreta- tion of breast MR images is difficult and has a marked learning curve. There also is the question of whether the results obtained at centers with extensive experience in breast MRI can be extended to centers with less experience. Our institution is a rural academic medical center with a well-established multidisciplinary breast program but relatively short experience with breast MRI. Initial imaging of ipsilateral malignant lesions was begun in early When we decided to institute routine bilateral breast MRI of all patients with newly diagnosed breast cancer, a locally funded grant was obtained to allow tracking, an audit of our findings, and postoperative follow-up. The aim was to evaluate the utility of breast MRI at our institution compared with the results in the literature, including the influence on surgical treatment of these patients. Subjects and Methods Patient Population Between September 2005 and December 2006, all patients with newly diagnosed breast cancer 1438 AJR:192, May 2009

2 Preoperative Breast MRI who underwent staging breast MRI were entered prospectively into this study. They included patients with disease diagnosed with core needle biopsy or with excisional biopsy. underwent bilateral breast MRI (unilateral if they had previously undergone mastectomy) according to the protocol in Breast MRI Technique. This study was approved by our hospital institutional review board, who deemed that no written or oral patient consent was required. Breast MRI Technique For the first 6 months of the study, owing to software limitations, breasts were imaged individually within 1 week of each other. For the rest of the study, bilateral acquisition was performed in one session. All imaging was performed with a 1.5-T MRI system (Signa, GE Healthcare). Unilateral acquisition was performed with a four-channel dedicated breast coil and bilateral acquisition with a seven-channel dedicated breast coil. Vibrant software (GE Healthcare) was used in all cases. The unilateral acquisitions were primarily in the sagittal plane, and the bilateral acquisitions were primarily in the axial plane. Sequences for both bilateral and unilateral imaging were T2-weighted spoiled gradient-recalled echo with and without fat saturation before gadolinium administration and dynamic spoiled gradient-recalled echo images after gadolinium administration. Gadolinium was administered with a mechanical injector at a dose of 0.2 ml/ kg. Slice thickness was 2 3 mm depending on the protocol and sequence. Dynamic acquisitions lasted less than 90 seconds each with three or four contrast-enhanced acquisitions, depending on the protocol. A final sequence was performed in the plane perpendicular to the dynamic series. Data were exported to a workstation for kinetic analysis with CADstream software (Confirma). MRI Interpretation One of four radiologists with experience in breast imaging on a PACS workstation used the American College of Radiology BI-RADS criteria and terminology to interpret the images. On the basis of both the morphologic features of the lesion and kinetic enhancement information, each lesion was assigned to a BI-RADS category 1 6. Interpreters were not blinded to clinical data or to previous imaging findings. detected were assigned a sequential identifying number per breast and located by both clock position and distance from the nipple. To reduce confusion, the lesion number persisted through all further imaging, surgical management, and pathologic examinations. These data were prospectively entered into a database constructed with Filemaker Pro software (Filemaker). These MRI findings, in conjunction with earlier clinical data, were the basis for subsequent treatment, which included 6-month follow-up MRI, recall for a second-look ultrasound examination or mammography with or without proceeding to core needle biopsy, MRIguided biopsy, surgical excision of the indicator lesion, surgical excision of additional lesions after needle localization, mastectomy, no further surgery, or a combination of these options. Data Recording and Analysis For the purposes of this study, the electronic clinical information system at our institution was reviewed retrospectively by two of the authors for the 199 patients in the study group. For five patients treated by surgeons outside our institution this information was obtained from the outside surgeon. Data recorded included lesion descriptions and BI-RADS category determined at preoperative breast MRI interpretation; lesion-by-lesion correlation, when available, of MRI findings with the histologic results from the electronic pathologic record; callbacks for additional imaging and any additional biopsies performed because of the MRI results; and final surgical procedure. The BI-RADS cate gory after additional imaging was used for data analysis (i.e., patients assigned BI-RADS category 0 before additional imaging would have had their category changed to BI-RADS 2 5 after the additional imaging). When necessary in complex cases, review of the MR images was repeated by one investigator. For patients with benign results of ultrasound or stereotactic biopsy of additional lesions when no obvious concordant mass is found and for all patients with benign results of MRI-guided biopsy, our routine is to perform postbiopsy short-interval MRI, usually in 6 8 weeks. Clips are always left in place. For some patients the short follow-up interval was 6 months. The size of MRI-detected lesions, including the primary tumor, was compared with the size of lesions recorded on the pathology report. Only lesions examined by pathologists at our institution were included in this study because accurate measurements often were not included in outside pathology reports. with postoperative seroma were excluded from the study, as were pa tients who underwent neo adjuvant chemo therapy after MRI but before final surgery. Also excluded were patients with extensive ductal carcinoma in situ (DCIS) in addition to invasive carcinoma, but in whom the measurement of the DCIS component was not included in the pathology report. The difference between the pathologically determined maxi mal size and the maximal diameter recorded on the initial MRI report was calculated in millimeters. The Filemaker Pro and Excel (Microsoft) programs were used as the primary means of database analysis to summate numbers of benign, malignant, and unknown lesions identified at MRI. Maximum lesion size at pathologic examination and MRI were correlated when possible; however, pathologic measurements of mixed invasive and DCIS tumors included only the invasive component, limiting effective comparison. Standardized Review by Blinded Observer Since early 2005, when bilateral breast MRI of all patients with newly diagnosed breast cancer became routine at our institution, patients do not consult with surgeons about treatment options until after breast MRI results become available. Because of anxiety about a breast cancer diagnosis and detection of additional lesions, surgical treatment is often influenced by factors beyond those dictated by the clinical evidence itself. Some patients choose mastectomy for personal reasons despite the feasibility of less aggressive but equally effective alternatives. In an effort to minimize these subjective elements in evaluating the effect of MRI on surgical management, one breast surgeon retrospectively reviewed the images of all patients in whom additional malignant lesions had been detected with MRI. Reviews and surgical management recommendations were made without knowledge of the actual surgical treatment of these patients. This blinded review was also intended to minimize the potential influence of management variations among the breast surgeons treating these patients. For example, one surgeon performs double lumpectomy on some patients with two small malignant lesions in different quadrants, whereas others do not. The distance between the edge of the pri mary lesion and the farthest edge of any new additional malignant lesion was measured with electronic calipers. Because we strive for 1-cm surgical margins at our institution, the reviewer conservatively assumed that any additional lesions less than 1.5 cm from the index lesion would have been excised during primary lump ectomy and would not need additional wire localization. Additional lesions within this 1.5-cm margin therefore were not considered to result in management changes based on the MRI findings. If the distance between the index and additonal malignant lesions was 1.5 cm or more, the reviewing surgeon determined whether patients were candidates for conservative breast surgery before MRI and whether they remained so after detection of the additional malignant lesions. In these cases, the reviewing surgeon assessed the need for more extensive lumpectomy or an ad ditional needle localization procedure. for whom the surgical procedure AJR:192, May

3 Schell et al. TABLE 1: Histologic Results on Index Index Lesion Type DCIS 10 DCIS with microinvasion 4 IDC 91 IDC and DCIS 39 IDC and invasive lobular carcinoma 2 IDC with lobular features 9 IDC with medullary features 3 Invasive lobular carcinoma 15 Invasive lobular carcinoma and DCIS 1 Inflammatory 5 Adenoid cystic carcinoma 1 Mucinous 4 Paget disease of nipple 1 Intraductal papillary carcinoma 1 Lymphoma 1 Poorly differentiated carcinoma 1 Unknown a 13 Total 202 b Note DCIS = ductal carcinoma in situ, IDC = infiltrating ductal carcinoma. a Pathologic findings on 13 lesions were unknown because the examination was performed outside our hospital, and the primary pathology report could not be obtained. b One patient had bilateral cancer, and two patients had two index lesions in the same breast. was converted to mastectomy or who underwent neoadjuvant chemo therapy on the basis of the MRI findings also were identified. It was assumed that detection of any con tralateral malignant disease affected management. Results Study Population Between September 2005 and December 2006, 199 patients underwent breast MRI after a new diagnosis of cancer. The mean patient age was 57.4 years (range, years; median, 57 years). Index tumor types for these patients are shown on Table 1. In 97% (194/199) of the cases, the index pathology report was read or reinterpreted by pathologists at our institution. Two hundred ipsilateral MRI examinations and 185 contralateral MRI examinations were performed. One patient underwent two ipsilateral MRI examinations because of the presence of known bilateral primary malignant tumors of the breast. TABLE 2: Types of Contralateral and Ipsilateral Biopsies Performed Because of MRI Finding of Suspicious Lesion (BI-RADS 3, 4, or 5) Sample Source Ipsilateral Contralateral Combined Lesion excised at surgery with index lesion Mastectomy a Partial mastectomy Biopsy separate from excision of index lesion MRI guided Separate needle localization excision Ultrasound guided Stereotactic Totals Excised at surgery Separate biopsies All biopsies a These two patients underwent prophylactic contralateral mastectomy that included a lesion found suspicious at MRI. Ipsilateral Breast Additional suspicious lesions (BI-RADS 3, 4, or 5) were detected in 29% of patients (75 lesions in 58 patients). Five lesions in five patients regressed with neoadjuvant chemotherapy. These lesions were excluded from further analysis because it was believed that they could be either benign or malignant, so 70 additional ipsilateral suspicious lesions were used for the denominator. Sixteen patients (8%) needed additional imaging, either ultrasound or mammography, after MRI. Ten patients underwent ultrasound examination only, one patient underwent mammography only, and five patients needed both additional studies. Pathologic results on 69 of the suspicious lesions were available for analysis. One additional lesion grew during chemotherapy, as did the index lesion, and was assumed malignant. Thirty-three of these lesions were evaluated at mastectomy, and another 12 lesions were excised with the index tumor in a partial mastectomy. Twenty-four lesions were biopsied separately from the surgical procedure for the index tumor. The types of biopsies are summarized in Table 2. Additional malignant lesions visualized at MRI were found in the ipsilateral breast in 16% of patients (44 lesions in 32 patients), as shown in Table 3. An example is shown in Figure 1. These lesions had previously been undetected at either mammography or clinical examination. Most (80%, 35 of 44) of the malignant lesions detected were invasive cancer (Table 4). Excluding the lesions that regressed during chemotherapy, the truepositive rate for MRI detection of ipsilateral malignant tumors was 63% (44/70). Twentysix of the 70 lesions were benign, giving a false-positive rate of 37%. Thirteen patients (6.5% of all of the patients) had undergone unnecessary ipsilateral biopsies, defined as a separate procedure to sample a lesion that turned out to be benign. To determine the potential effect of our findings on the surgical treatment of the patients, the additional ipsilateral malignant lesions were reviewed by a breast surgeon at our institution. The results are shown in Table 5. We decided that lesions more than 1.5 cm from the index lesion were unlikely to be routinely included in a partial mastectomy if not separately needle localized. According to this criterion, 38 of the 44 additional malignant tumors (86%) would not have been included in this surgical margin around the index tumor. Nine of the 32 patients with additional ipsilateral malignant tumors (5% of all of the patients) had MRI findings that would have necessitated conversion from ipsilateral breast conservation therapy to mastectomy to include all lesions. Thirteen patients (6.5% of all of the patients) would have needed wider excision of the primary lesions 1440 AJR:192, May 2009

4 Preoperative Breast MRI TABLE 3: Additional Ipsilateral and Contralateral Identified on MR Images Location to include the additional malignant lesions. In another three patients (1.5% of all of the patients) who were still considered candidates for lumpectomy, a separate needle localization excision would have been necessary to include the additional malignant tumors. At least one patient underwent neoadjuvant chemotherapy on the basis of the MRI results. Contralateral Breast One hundred eighty-five patients underwent MRI of the contralateral breast. Additional suspicious lesions were detected in 19% of the patients (41 lesions in 36 patients). Twenty-five patients (14%) needed additional imaging after MRI: 10 patients underwent ultrasound examination only, three underwent mammography only, and 12 underwent both mammographic and ultrasound examinations. Of the additional 41 lesions detected, pathologic results were available for 34. The types of biopsy are shown in Table 2. Ten of 34 sampled lesions were malignant (Fig. 2), for a true-positive rate of 29%. Thus synchronous contralateral cancer not previously detected with conventional imaging was found Total Additional Additional Malignant Additional High-Risk a Additional Benign b in 4% (8/185) of the patients (Table 3). The pathologic results on these malignant lesions are shown in Table 4. Sixty percent of the additional contralateral malignant tumors were invasive, and 40% were DCIS. High-risk lesions (i.e., isolated atypical ductal hyperplasia and lobular carcinoma in situ) were found in 9% (3/34) of sampled contralateral breasts. Contralateral breast (185 patients) (19) 10 8 (4) (12) 3 3 (1.6) Ipsilateral breast (199 patients) (29) (16) (13) 0 0 Combined (37) (19) (23) 3 3 (1.5) Note Values in parentheses are percentages. a Malignant tumors diagnosed with separate biopsy or with excision with primary lesion at mastectomy. Excludes lesions that regressed during chemotherapy because those lesions were deemed indeterminate. b Includes lesions that had regressed at 6-month follow-up MRI and were assumed benign. A B Fig year-old woman with MRI-detected ipsilateral malignant tumor. A, Axial gadolinium-enhanced spoiled gradientrecalled echo MR image of left breast shows index tumor is 2.5-cm infiltrating ductal carcinoma (arrow) in 12-o clock position. B, Axial gadolinium-enhanced spoiled gradientrecalled echo MR image of left breast shows enhancing 1-cm mass (arrow) in left lower inner quadrant. Biopsy result after second-look ultrasound examination was invasive ductal carcinoma. Mastectomy was performed. Seven contralateral lesions were evaluated with subsequent MRI or mammography and were assumed to be benign if the lesion had regressed or disappeared. Unnecessary biopsy had been performed on 10% (18/185) of patients who underwent MRI of the contralateral breast. The false-positive rate for contralateral lesions was 68%. Eight patients underwent contralateral mastectomy. Only one of these patients had been found to have a contralateral malignant tumor. Another patient was found to have highrisk histologic findings (atypical ductal hyperplasia). Six patients chose mastectomy even though they had normal MRI findings, either TABLE 4: Pathologic Results on Additional Ipsilateral and Contralateral Malignant Tumors (n = 199 ) Pathologic Finding Ipsilateral Contralateral Combined Ductal carcinoma in situ Invasive ductal carcinoma Invasive ductal carcinoma with extensive ductal carcinoma in situ Invasive lobular carcinoma Other Progressed during chemotherapy Total a b a Three patients had lesions with more than one type of malignancy. b Two patients had different pathologic findings in the ipsilateral and contralateral breast. AJR:192, May

5 Schell et al. TABLE 5: Potential Effect of Detection of Additional Ipsilateral Malignant on Treatment of with Biopsy-Proven Additional Ipsilateral Malignant Tumors Who Underwent Surgery (n = 31) Management Effect because of a family history of breast cancer (two patients) or young age (four patients). Combined Ipsilateral and Contralateral Data Thirty-four of the 199 patients (17%) needed additional imaging with either ultrasound or mammography. At MRI, 74 patients (37%) were found to have a total of 115 suspicious lesions. Forty-seven patients (24%) underwent a separate procedure for acquisition of tissue for diagnosis. Fifty-four additional unsuspected malignant tumors were found in either breast in 38 patients (19%), including two patients found to have bilateral additional malignant lesions. Seventy-six percent of MRI-detected malignant tumors were invasive, and 22% were DCIS. One tumor grew on chemotherapy but was not biopsied. The median size of these malignant tumors was 10 mm (mean, 16 ± 14.7 [SD] mm; range, 3 65 mm). The rate of positive biopsy findings was 35% for the 56 lesions biopsied separately None 5 (16) Change to mastectomy 9 (28) Wider excision needed 13 (42) Additional needle localization needed 3 (9) Other management change 1 (3) Note Values in parentheses are percentages. A Fig year-old woman with MRI-detected contralateral malignant tumor. A, Axial gadolinium-enhanced spoiled gradient-recalled echo MR image of left breast shows index tumor is 2-cm invasive ductal carcinoma (arrow). B, Axial gadolinium-enhanced spoiled gradient-recalled echo MR image of right breast shows enhancing 1-cm mass (arrow). Biopsy result after second-look ultrasound examination was invasive lobular carcinoma. from surgery. Thirty-five percent (8/23) also was the rate of positive findings of malignancy at MRI-guided biopsy. B Lesion Size: MRI Versus Pathologic Finding One hundred fifty lesions, including both index lesions and additionally identified malignant tumors, were of sufficient pathologic size for meaningful comparison of sizes at MRI and pathologic examinations. The results are shown in Table 6. The mean absolute difference between the size on MR images and the size at pathologic examination was 6.8 ± 9.0 mm (median, 3 mm). A difference greater than 15 mm was found in 5.3% (8/150) of lesions and was considered potentially clinically significant. At MRI, true size was underestimated in three of the eight patients, and in five it was overestimated. These differences were barely statistically significant (p = 0.03; 95% CI, to 2.2); however, the upper range of the 95% CI was 2.2 mm, which is extremely unlikely to be clinically significant. The pathologic MRI correlation was inadequate in seven patients because, first, in four cases the pathology report included the measurement of only the invasive component, not the associated extensive DCIS (hence the MRI-overestimated lesion size). The second cause of inadequate correlation was that in three cases the pathology report combined adjacent multiple lesions to give one pathologic measurement, whereas at MRI each lesion was measured separately (hence the MRI-underestimated tumor size). Including these seven patients in the analysis, however, only minimally changes the level of significance (p = 0.03; 95% CI, 0.1 to 2.6). Overall, among the 150 lesions with adequate correlation, in 38% of cases the pathologic size was underestimated by 1 mm or more at MRI, and in 47% of cases the pathologic size was overestimated. Discussion In this study, we found that 19% of patients with recently diagnosed breast cancer had a contralateral or ipsilateral malignant tumor that was detected only with MRI. Sixteen percent of patients had another ipsilateral malignant tumor, and 4% had a synchronous contralateral malignant tumor. These data are similar to those reported in other studies and provide support for the use of preoperative screening breast MRI outside the investigational setting. Rates of detection of additional ipsilateral malignant tumors in other reports have ranged between 6% and 27% [1 5]. The reported incidence of contralateral malignant tumors is also similar to our finding, ranging between 3% and 9% [1, 2, 5 9]. Although our numbers are significant, the overall influence of the use of MRI is difficult to assess and comes with substantial cost, both financial and emotional, to the patient. In our total study population of 199 patients, we found additional suspicious lesions in 37% of patients (n = 74); 17% of patients underwent additional imaging studies, and 24% of patients underwent separate biopsy procedures not planned before MRI. These procedures can delay surgery days or weeks. Although it is unlikely to markedly affect the course of the disease, the delay can certainly increase patient anxiety. We were unable to evaluate delays in treatment owing to the large number of referrals from outside our institution (an inevitable cause of delay) and our lack of comparative data before beginning routine preoperative screening. The anxiety associated with finding an additional 1442 AJR:192, May 2009

6 Preoperative Breast MRI TABLE 6: Size Comparison of MRI and Pathologic Findings on Malignant (n = 150) Value Size of Lesion on MR Image (mm) Size of Lesion at Pathologic Examination (mm) Difference Between MRI and Pathologic Finding (mm) Mean a Median SD Range Note At MRI, the size of 38% of the lesions was underestimated, and the size of 47% of the lesions was overestimated. a Significant at p = 0.03, 95% CI, to 2.2. Fig year-old woman with false-positive MRI finding. Sagittal gadolinium-enhanced spoiled gradient-recalled echo MR image of right (contralateral) breast shows 8-mm enhancing mass (arrow). Lesion was biopsied with ultrasound and found to be a papilloma. abnormality on a breast MR image, even if it turns out to be benign, such as the lesion in Figure 3, can influence patients to select mastectomy over conservative breast therapy. We are studying this question. To accurately assess the potential effect of breast MRI on patient care, one surgeon blinded to the final surgical management separately evaluated the ipsilateral breast MRI data. The rationale behind this approach was multifold: patients do not usually see a surgeon at our institution until after the breast MRI report is available, hence any initial surgical plan is influenced by this report; surgeons differ in their approaches to managing larger tumors and multifocal cases of breast cancer; and a patient s choice of mastectomy is often based on factors other than medical necessity. Our findings with this approach show that nine (5%) of all of the patients in the study would have had conservative breast surgery converted to mastectomy and that 17 (9%) of the patients would have undergone more extensive local surgery (Fig. 4) or preoperative neoadjuvant chemotherapy. We assumed that detection of any contralateral malignant tumor (4% of all patients) substantially affected these patients treatment. This detection rate, although small, is extremely important for patient care. The results of our study confirm the utility of breast MRI in the detection of mammographically occult malignant tumors in both the ipsilateral and contralateral breasts. Effective local therapy for breast carcinoma is known to be essential. Punglia et al. [13] discussed the importance of local control of breast cancer, noting that current systemic therapies for breast cancer are effective mainly against microscopic disease and that improved local control is associated with considerable improvement in the 15- year survival rate. This survival advantage is obtained with either more extensive surgery or whole-breast irradiation. Fischer et al. [14] addressed the question whether breast MRI affects clinical outcome. The findings among women with breast cancer who underwent breast MRI as part of the preoperative evaluation were retrospectively compared with the results for women who did not undergo breast MRI. The investigators found that the rate of local tumor recurrence an average of 40 months after treatment was significantly reduced among the patients who had undergone preoperative MRI (1.2% vs 6.8%, p < 0.001). In addition, the incidence of a contralateral cancer diagnosis 40 months after treatment decreased (1.7% vs 4%, p < 0.001) in women who underwent MRI. In a later study [15], however, other investigators found that only 1.7% of 476 patients had local recurrence during a median 5-year follow-up period after partial mastectomy and radiation therapy and that 2.7% had a contralateral malignant tumor. This finding suggests that routine preoperative breast MRI can be an extremely expensive way of making a very small impact on recurrence rates. An even more complex question is whether detection of additional malignant tumors affects survival. Use of mammography has been criticized for leading to overdiagnosis of DCIS. However, 76% of the additional malignant tumors in our study were invasive and had a median size of 10 mm. of this size may not be successfully controlled with whole-breast radiation therapy to the ipsilateral breast. To our knowledge, studies of the survival benefit of preoperative MRI have not been conducted and because of the extremely large numbers of patients needed may never be performed. Morrow [16] questioned the widespread use of breast MRI. She pointed out that there is little published information that supports breast MRI as a way to improve clinical outcome and that clinical outcome among patients undergoing breast conservation therapy based on mammographic and clinical selection is already excellent. Among patients Fig year-old woman with MRI finding of changing extent of disease. Sagittal gadoliniumenhanced spoiled gradient-recalled echo MR image of right (ipsilateral) breast shows 8-mm enhancing mass (long arrow), which was the known lesion, and 3 cm of nonmasslike linear enhancement (short arrow). Biopsy result was ductal carcinoma in situ. Larger lumpectomy was performed. AJR:192, May

7 Schell et al. who receive brachytherapy with an implantable device such as the MammoSite system (Cytyc) and therefore would not be receiving whole-breast radiation, the effect is likely to be greater. A strong argument can be made for performance of breast MRI on all patients before brachytherapy. Improved knowledge of tumor extent before surgery in theory should lead to decreased need for reexcision because of inadequate surgical margins. Findings presented by Comstock et al. [17] suggest this theory is correct. Use of breast MRI, however, has been found to lead to both overestimation and underestimation of the extent of disease [18 20]. Although our data show a statistically significant difference between lesion size at pathologic examination and size measured at breast MRI, the 95% CI was only 2.2 mm. In 48% of these cases, the breast MRI measurement was greater than that seen at pathologic examination, and in 40% it was smaller. This difference is highly unlikely to be surgically significant. However, in 5% of patients, there was a 15-mm or greater difference between the breast MRI measurement and the pathologic measurement. In those patients, the discrepancy can lead to larger excision than necessary, even mastectomy, or to positive margins. The limitations of this study included the assumption that MRI-detected lesions that were not sampled were either benign on the basis of follow-up imaging findings (16 lesions) or malignant on the basis of lack of response to chemotherapy (one lesion). Another limitation was that pathologic size measurements in patients with both invasive and in situ disease at a site included only the invasive component, whereas the breast MRI measurement likely included both the invasive and in situ components owing the difficulties of differentiating the two. In several patients, the pathologic measurements combined multiple small nodules into one measurement. The latter two limitations resulted in exclusion of seven patients from the size correlations. The 1.5-cm cutoff for deciding whether an additional lesion was or was not likely to be included in a routine excision was subjective and based on local practice but was considered to be a conservative figure to avoid overestimation of the influence of breast MRI. We assessed surgical importance with the aid of only one surgeon, thus the result was an ideal rather than real effect. A further limitation was that technical improvements during the course of the study led to a change at our institution from unilateral sagittal acquisition to bilateral axial acquisition. We also increased the number of clinical interpreters during this time, each of whom had a marked learning curve in developing skills in breast MRI interpretation. We conclude, as others have, that breast MRI depicts a considerable number of both ipsilateral and contralateral malignant tumors not detected with mammography or clinical breast examination. The results in the literature are reproducible outside of the investigational setting, but the high rate of false-positives findings and need for additional biopsies and careful correlative imaging may limit the widespread use of breast MRI to centers with extensive expertise in breast imaging. Although there appears to be a substantial effect on patient care, the cost-effectiveness of breast MRI and its overall effect on patient survival warrant further study. Acknowledgment We thank Dan Deneen for editing the manuscript. References 1. Berg WA, Gutierrez L, NessAiver MS, et al. Diagnostic accuracy of mammography, clinical examination, US, and MR imaging in preoperative assessment of breast cancer. Radiology 2004; 233: Dao TN, Lamont JP, Knox SM. Clinical utility of breast magnetic resonance imaging in patients presenting with primary breast cancer. Proc (Bayl Univ Med Cent) 2007; 20: Deurloo EE, Peterse JL, Rutgers EJ, Besnard AP, Muller SH, Gilhuijs KG. Additional breast lesions in patients eligible for breast-conserving therapy by MRI: impact on preoperative management and potential benefit of computerised analysis. Eur J Cancer 2005; 41: Liberman L, Morris EA, Dershaw DD, Abramson AF, Tan LK. MR imaging of the ipsilateral breast in women with percutaneously proven breast cancer. AJR 2003; 180: Wiener JI, Schilling KJ, Adami C, Obuchowski NA. Assessment of suspected breast cancer by MRI: a prospective clinical trial using a combined kinetic and morphologic analysis. AJR 2005; 184: Lee SG, Orel SG, Woo IJ, et al. MR imaging screening of the contralateral breast in patients with newly diagnosed breast cancer: preliminary results. Radiology 2003; 226: Lehman CD, Blume JD, Thickman D, et al. Added cancer yield of MRI in screening the contralateral breast of women recently diagnosed with breast cancer: results from the International Breast Magnetic Resonance Consortium (IBMC) trial. J Surg Oncol 2005; 92: Lehman CD, Gatsonis C, Kuhl CK, et al. MRI evaluation of the contralateral breast in women with recently diagnosed breast cancer. N Engl J Med 2007; 356: Liberman L, Morris EA, Kim CM, et al. MR imaging findings in the contralateral breast of women with recently diagnosed breast cancer. AJR 2003; 180: Del Frate C, Borghese L, Cedolini C, et al. Role of pre-surgical breast MRI in the management of invasive breast carcinoma. Breast 2007; 16: Zhang Y, Fukatsu H, Naganawa S, et al. The role of contrast-enhanced MR mammography for determining candidates for breast conservation surgery. Breast Cancer 2002; 9: Bilimoria KY, Cambic A, Hansen NM, Bethke KP. Evaluating the impact of preoperative breast magnetic resonance imaging on the surgical management of newly diagnosed breast cancers. Arch Surg 2007; 142: Punglia RS, Morrow M, Winer EP, Harris JR. Local therapy and survival in breast cancer. N Engl J Med 2007; 356: Fischer U, Zachariae O, Baum F, von Heyden D, Funke M, Liersch T. The influence of preoperative MRI of the breasts on recurrence rate in patients with breast cancer. Eur Radiol 2004; 14: Gorechlad JW, McCabe EB, Higgins JH, et al. Screening for recurrences in patients treated with breast-conserving surgery: is there a role for MRI? Ann Surg Oncol 2008; 15: Morrow M. Magnetic resonance imaging in the preoperative evaluation of breast cancer: primum non nocere. J Am Coll Surg 2004; 198: Comstock C, Hunt P, Middleton M. Effect of preoperative MRI on mastectomy rates, lumpectomy negative margin rates and time to surgery in patients with known breast cancer. (abstr). RSNA Oak Brook, IL: Radiological Society of North America. V2007/conference/event_display.cfm?em_ id= Accessed February 2, Blair S, McElroy M, Middleton MS, et al. The efficacy of breast MRI in predicting breast conservation therapy. J Surg Oncol 2006; 94: Mann RM, Veltman J, Barentsz JO, Wobbes T, Blickman JG, Boetes C. The value of MRI compared to mammography in the assessment of tumour extent in invasive lobular carcinoma of the breast. Eur J Surg Oncol 2008; 34: Van Goethem M, Schelfout K, Dijckmans L, et al. MR mammography in the pre-operative staging of breast cancer in patients with dense breast tissue: comparison with mammography and ultrasound. Eur Radiol 2004; 14: AJR:192, May 2009