Targeted Ultrasound of the Breast in Women With Abnormal MRI Findings for Whom Biopsy Has Been Recommended

Transcription

1 Women s Imaging Original Research Meissnitzer et al. Ultrasound of the Breast After Abnormal MRI Women s Imaging Original Research WOMEN S IMAGING Matthias Meissnitzer 1 D. David Dershaw Carol H. Lee Elizabeth A. Morris Meissnitzer M, Dershaw DD, Lee CH, Morris EA Keywords: breast biosy, breast cancer, breast imaging, breast MRI, breast ultrasound, directed ultrasound DOI: /AJR Received January 29, 2009; acceted after revision Aril 14, All authors: Deartment of Radiology, Breast Imaging Section, Memorial Sloan-Kettering Cancer Center, 300 East 66th Street, #723, New York, NY Address corresondence to D. D. Dershaw. AJR 2009; 193: X/09/ American Roentgen Ray Society Targeted Ultrasound of the Breast in Women With Abnormal MRI Findings for Whom Biosy Has Been Recommended OBJECTIVE. This study was erformed to identify characteristics of susicious lesions seen on breast MRI that are most likely to have an ultrasound correlate and to determine how often the resumed ultrasound correlate actually corresonds to the MRI finding. MATERIALS AND METHODS. From Setember 2005 through December 2007, targeted ultrasound was erformed for 519 susicious MRI-detected lesions in 361 women. Retrosective review was erformed to determine lesion tye (mass vs nonmass), lesion descritors, lesion size, BI-RADS category, indication for MR examination, atient age, and biosy outcome. The results of 80 follow-u MRI examinations among 154 cases with concordant benign results on ultrasound-guided biosy were noted. RESULTS. A resumed ultrasound correlate was found in 290 (56%) of the 519 lesions with masses more likely than nonmass lesions to be seen with ultrasound (62% of masses and 31% of nonmass lesions). Increasing lesion size, assessment of BI-RADS category 5 versus BI-RADS category 4, rim enhancement in masses, and clumed enhancement in nonmass lesions were also significantly more likely to have an ultrasound correlate. On follow-u imaging in 80 benign, concordant ultrasound-guided biosies, the sonograhic lesion did not corresond to the MRI finding in 10. Nine of these 10 lesions underwent subsequent MRI-guided biosy and five cancers were diagnosed. CONCLUSION. The MR characteristics of lesions most likely to be seen with an ultrasound correlate were mass versus nonmass, increasing size, and increased level of susicion of the lesion. Cli lacement and follow-u imaging after ultrasound-guided biosy that yields benign concordant results should be erformed to detect cases in which the resumed ultrasound correlate is inaccurate to detect unsusected false-negative biosies. S ome susicious lesions on MRI can be biosied more easily under ultrasound guidance than MRI guidance. Advantages of ultrasound guidance include less time needed for the biosy, lower cost for the biosy, greater access to ultrasound than to MRI, greater accessibility of some sites in the breast and axilla, and increased atient comfort [1 6]. However, exerience with targeted ultrasound after MRI has shown that a correlate is not found in u to 77% of cases, wasting time, effort, and exense in the diagnostic worku [6, 7]. Furthermore, there are currently no guidelines as to which MRI lesions should undergo targeted ultrasound for the uroses of biosy. Thus, individual radiologists have varying criteria for this recommendation. In addition, when an ultrasound sonograhic correlate to a susicious MRI finding is found and biosy is erformed using ultrasound guidance, it is unknown how often the resumed correlate actually does or does not reresent the MRI lesion. This study was undertaken to review our exerience with the recommendation for targeted ultrasound of susicious MRI lesions to identify MR characteristics of lesions for which targeted ultrasound should be recommended. In addition, this study evaluated the outcome of biosy and follow-u of cases with a resumed ultrasound correlate to determine whether the correlation is ever inaccurate. Materials and Methods This retrosective study was undertaken with aroval of the institutional review board. During a 28-month eriod (Setember 1, 2005 December 31, 2007), 361 women with 519 MRI-discovered BI-RADS category 4 or 5 lesions underwent AJR:193, October

2 Meissnitzer et al. directed breast sonograhy within 3 weeks to assess the ossibility of ultrasound-guided biosy for these lesions. These 519 lesions constitute the cases in our study. All MRI examinations were interreted by one of eight breast imagers with between 1 and 10 years of exerience interreting breast MRI examinations and other breast imaging studies and erforming breast biosy rocedures using MRI or sonograhic guidance. Correlation of MRI findings with rior mammograhic or sonograhic findings was erformed at the time of MRI interretation if these other breast imaging examinations were available. There was no standard rotocol for the recommendation of targeted ultrasound when a susicious finding that was believed to require biosy was encountered on breast MRI, and the decision as to whether to erform ultrasound was made at the discretion of the interreting radiologist. For data analysis, all cases in which targeted ultrasound was reorted by the interreting radiologist to show a lesion that was thought to reresent the MRI finding were included in the grou that had an ultrasound correlate even if follow-u imaging showed that assessment to be inaccurate. MRI was erformed on 1.5-T magnets with dedicated breast coils using standard technique with IV contrast material and sagittal imaging with 3-mm-thick contiguous slices. Sonograhy (Acuson Sequoia 512, Siemens Healthcare) was erformed using 8-15 MHz small-arts transducers. The sonograhy examination was initially erformed by a trained and certified breast sonograher, and additional scanning was erformed by the interreting breast imaging radiology technologist at the radiologist s discretion. MRI studies were available at the time sonograhy was erformed, and the two studies were correlated in the reort of the sonograhy examination. MRI-guided biosies were erformed with 9-gauge vacuum-suction MR-comatible biosy systems. Sonograhically guided biosies were er formed with a variety of 14-gauge gun-needle breast biosy devices obtaining a minimum of three cores. All biosies were erformed by breast imaging radiologists. For each lesion, reorts were reviewed to determine lesion tye, descritors, and size; BI-RADS category; the indication for MR examination; and atient age. Lesion characteristics were recorded on the basis of the BI-RADS lexicon for breast MRI. were divided by tye into masses and nonmasses: There were 424 (82%) masses and 95 (18%) nonmass lesions. For masses, lesion characteristics recorded included rim enhancement and internal enhancement attern. For nonmass lesions, clumed or nonclumed attern was recorded. Among the 519 lesions, biosy was erformed in 422, including 253 with an ultrasound correlate TABLE 1: Characteristics of Patients Parameter erformed with ultrasound guidance and 169 without an ultrasound correlate erformed with MRI guidance. In the remaining 97, the biosy was not erformed because lesions were lost to follow-u (n = 27), mastectomy was erformed eliminating need of biosy for the secific MR lesion (n = 21), or lesions were reassessed as benign or robably benign based on sonograhic imaging findings (n = 35). Biosy was not erformed because the lesion could not be found at the time of attemted MR biosy in nine lesions, stage IV disease obviated biosy in three lesions, and a similar adjacent lesion underwent biosy with benign results and this histology was also thought to be concordant for the nonbiosied lesion in two lesions. All histoathologic results were correlated to MR images of the biosied lesion by the breast imaging radiologist who erformed the rocedure to determine concordance. Among the 154 benign histologically concordant lesions with an ultrasound correlate, 80 had MR follow-u and the results of the follow-u examinations were noted. Selection of these cases for follow-u was at the discretion of the radiologist erforming the biosy and does not follow any established rotocol. The remaining 74 had either benign concordant results, such as large fibroadenomas for which follow-u imaging was deemed not to be necessary, or were lost to follow-u. MRI characteristics of No. (%) of Patient age (y) < (55) 34 (45) (51) 99 (49) (61) 59 (39) (61) 24 (39) (58) 13 (42) Indication for MRI Screening high-risk atient (52) 91 (48) Screening contralateral breast (53) 85 (47) Staging (63) 25 (37) Problem solving (68) 17 (32) Identify unknown rimary cancer (56) 8 (44) Other 9 6 (67) 3 (33) Breast density 494 a b Fatty 4 1 (25) 3 (75) Scattered fibroglandular (57) 109 (43) Heterogeneously dense (52) 98 (48) Extremely dense (53) 15 (47) a Twenty-five lesions had no recorded breast density. b Fisher s exact text. lesions that underwent ultrasound-guided biosy in which the resumed ultrasound correlate were found to be inaccurate and those in which the results of the ultrasound-guided biosy was discordant with MRI imaging findings were noted. Statistical analysis was erformed alying the chi-square test and Fisher s exact test using SPSS statistical analysis software, and values of less than 0.05 were considered significant. Results Among the 519 lesions, a resumed ultrasound correlate was found for 290 (56%) and no correlate was found for 229 (44%). A significant difference ( 0.001) was found for finding an ultrasound correlate for masses versus nonmasses, with 62% of all MR masses seen with ultrasound versus 31% of nonmasses. A breakdown of atient age, the indication for MR study, and density of breast in the study oulation is shown in Table 1. No significant difference was found between those for whom ultrasound found a correlate and those for whom it did not based on atient age, indication for MRI, or breast density. Lesion tye, lesion size, and BI-RADS category for the 519 lesions are shown in Table 2. The lesions ranged in size as measured on 1026 AJR:193, October 2009

3 Ultrasound of the Breast After Abnormal MRI TABLE 2: Characteristics of Lesion tye Parameter MRI from 2 to 70 mm with a mean of 12 mm. Masses had a mean size of 10 mm (range, 2 70 mm), which was significantly smaller than nonmass lesions (mean, 22 mm; range, 6 70 mm). For both masses and nonmass lesions, increasing size resulted in statistically significant increasing ultrasound consicuity, with an ultrasound correlate found in 50% of masses that were 5 mm or less, 56% of those 6 10 mm, 72.5% of masses mm, and 86% of masses larger than 15 mm. For nonmass lesions, an ultrasound correlate was found for 13% of those measuring 6 10 mm, 25% of those mm, and 42% of those larger than 15 mm. A more susicious aearance (i.e., BI-RADS category 5 vs 4) was also a statistically significant redictor of ability to find a lesion with ultrasound, and this finding was true for both masses and nonmasses. Malignant lesions were more likely to be found by ultrasound than benign lesions, although the absence of an ultrasound correlate did not exclude malignancy (Table 3). Among the 253 lesions with an ultrasound correlate that underwent biosy, 87 (34%) were malignant. Among the 169 lesions without an ultrasound correlate that underwent No. (%) of All (56) 229 (44) < Mass (62) 163 (38) Nonmass (31) 66 (69) Size of masses (mm) a < (50) 20 (50) (56) 115 (44) (72.5) 22 (27.5) > (86) 6 (14) Size of nonmasses (mm) a (13) 20 (87) (25) 18 (75) > (42) 28 (58) BI-RADS of masses Category (59) 155 (41) Category (81) 8 (19) BI-RADS of nonmasses Category (26) 64 (74) Category (75) 2 (25) a Based on MRI measurements. b Fisher s exact text b biosy, 34 (19%) were malignant. Among malignancies, invasive carcinoma was significantly more likely to be evident on ultrasound than ductal carcinoma in situ, and invasive ductal carcinoma was more likely to be seen on ultrasound than invasive lobular or mixed tye, although the difference was not statistically significant ( = 0.219). Among the 253 lesions with an ultrasound correlate that underwent ultrasound-guided biosy, results were benign and concordant in 154. However, on follow-u MRI in 80 of these cases, 10 of the sonograhic lesions that were biosied did not corresond to the MRI lesion. Subsequent MRI-guided biosy was erformed in nine of these cases and malignancy was diagnosed in five. In the remaining case, the MRI lesion was reassessed as robably benign and follow-u MRI, rather than biosy, was recommended. In 12 (5%) of 253 cases with an ultrasound correlate that underwent ultrasound-guided biosy, results were thought to be discordant with the MRI findings. Subsequent MRI-guided biosy was erformed in eight and malignancy was diagnosed in four. Among the remaining four cases, mastectomy was erformed in two, the lesion had disaeared at the time of attemted MRI-guided biosy in one, and the lesion was reassessed as robably benign based on its MR aearance on follow-u in the final case. In all, nine malignancies were missed by ultrasound-guided biosy and were discovered either by follow-u MRI that showed what was resumed to be an ultrasound correlate did not actually corresond to the MR lesion or because of erceived discordance between the MR aearance and ultrasound biosy results. Table 4 comares the characteristics of lesions found on ultrasound that were true correlates of the MR lesion versus those that were incorrectly identified as the ultrasound correlate of the MR lesion. Overall, 231 (91%) of 253 MRI lesions seen with ultrasound were accurately correlated. Twenty-two lesions including 12 with benign biosy results that were deemed discordant by the radiologist who erformed the rocedure and 10 benign lesions that were thought to be concordant but that were shown by follow-u MRI that the ultrasound lesion did not corresond to the MR lesion. Nine missed cancers included four found among the discordant grou and five among those with an inaccurate ultrasound correlate. Additionally, the nine cancers falsely identified as a benign lesion on ultrasound were comared. There was no identifiable characteristic that could searate these grous. We did not identify any characteristic that would increase the ossibility of incorrectly matching an MR lesion with an ultrasound correlate. Discussion Breast MRI is an increasingly imortant tool in breast imaging based on a sensitivity for the detection of breast cancer that aroaches 100% [8, 9]. When a susicious finding is found on MRI that is occult to hysical examination, MRI-guided biosy is necessary to establish a diagnosis. Because MRI-guided biosy is exensive, time consuming, and not always available, biosy guidance using ultrasound is referable if the target lesion can be seen with ultrasound. It is well known that breast lesions that form masses are more easily seen with ultrasound than those that do not, such as intraductal disease, and that as lesions increase in size, lesion consicuity on ultrasound increases [4, 5, 10 15]. Therefore, it is not surrising that masses and invasive cancers were tyes of lesions more readily found in our atients. Similar results have been reorted by Wiratkaun and AJR:193, October

4 Meissnitzer et al. TABLE 3: Versus Lesion Pathology Pathology of Biosied coworkers [5] who identified more ultrasound correlates in masslike lesions (55%, 23/42) than nonmass lesions (42%, 23/55) and also correlated increasing lesion size with increased ultrasound detection rate. Our results differ from theirs in that the difference between detection of lesions with masslike and nonmass enhancement was more ronounced (62% vs 31%, resectively) and that size was imortant in both mass and nonmass lesions to give rise to an ultrasound correlate. Our data suort the findings of other investigators that MRI-discovered lesions with an ultrasound correlate are more likely to be malignant than those without. LaTrenta and coworkers [6] found that 43% of lesions with a ultrasound correlate were malignant versus 14% of those without. Sim and colleagues [7] reorted a higher incidence of malignancy in both grous, with 92% of cancers having an ultrasound correlate versus 58% of benign lesions. As with other investigators [5 7], we also found that the absence of ultrasound correlate does not obviate biosy. The characteristics of cancers on MRI are siculated margin, rim enhancement, and irregular shae for masses and segmental or clumed ductal enhancement for nonmasses [16, 17]. Additionally, washout describing kinetic curve assessment is a good indeendent redictor for malignancy [18]. We hyothesized in our study that lesion characteristics No. (%) of All biosied lesions (60) 169 (40) Benign (55) 135 (45) Malignant (72) 34 (28) Biosied masses (65) 124 (35) Benign (61) 96 (39) Malignant (73) 28 (27) Biosied nonmasses (37.5) 45 (62.5) Benign (28) 39 (72) Malignant (67) 6 (33) Histology DCIS (37) 17 (63) < Invasive cancer (80) 17 (20) Invasive ductal cancer (85) 10 (15) a Invasive lobular cancer (68) 6 (32) Invasive mixed cancer 3 2 (67) 1 (33) Note DCIS = ductal carcinoma in situ. a Fisher s exact text. on MRI that are good redictors for malignancy are also good redictors for sonograhic correlation. Rim-enhancing masses were successfully correlated in 83% and BI-RADS category 5 lesions had at least a 75% detection rate on ultrasound. However, washout did not influence the robability for correlation in masses, which was already 62%. Not surrisingly, MR lesion characteristics with a low likelihood for malignancy also had a low likelihood of sonograhic correlation. Nonmass lesions smaller than 1 cm were also least likely to have an ultrasound correlate, with only 13% seen by ultrasound. Diagnostic efforts for correlation were most often noncontributory to diagnostic worku in these subgrous. The limitations of our study include the fact that it is retrosective and there was no defined rotocol for which cases underwent targeted ultrasound. These factors may bias results in favor of finding an ultrasound correlate because those cases deemed by the interreting radiologist to be unlikely to have a correlate were not evaluated with ultrasound. Therefore, the cases included in this series are heavily weighted toward masses versus nonmasslike enhancement. In addition, it is ossible that ultrasound was recommended for very susicious lesions desite their size, as shown by the fact that 40 masses measuring 5 mm or less were included. This factor would tend to bias the results in favor of finding an ultrasound correlate and, indeed, a correlate was found in 50% of these very small masses. Another limitation is that follow-u MRI was available for only 80 of 154 cases biosied with sonograhic guidance with benign results deemed to be concordant. No aarent characteristic searate those lesions with follow-u MRI from those without. However, it is imossible to exclude undetected bias searating these two grous, and this undetected bias may have imacted the data reorted herein. Nonetheless, even among the limited number of cases with MR follow-u, our results emhasize the imortance of correlating the location, size, and morhology of the lesion on MRI with the sonograhic characteristics [13, 14, 19, 20] and of erforming follow-u MRI for these cases [21]. Among our oulation, 22 inaccurately correlated lesions were identified based on either an MRI follow-u examination or imaging-discordant histology. Among these lesions, nine carcinomas were detected by subsequent MRI-guided biosy. Our analysis of the 22 inaccurately correlated lesions and the nine missed cancers could not identify a secific subgrou for which ultrasound correlation was more likely to be inaccurate. As with all imaging-guided biosies, discordant results should romt the recommendation for reeat biosy. In addition, a cli should be laced when ultrasound-guided biosy for an MRI correlate is erformed and follow-u MRI should be erformed for all benign concordant lesions. In addition to the data reorted in this study, another consideration in atient selection is cost. It is beyond the scoe of this article to resent a cost analysis of various management otions. Such a study would also be imacted by the variety of reimbursements for services throughout the United States and elsewhere. However, analysis of cost and inclusion of cost in considering management otions would be aroriate. In conclusion, ultrasound consicuity of MR-susicious lesions is greater in mass versus nonmass lesions and increases with increasing lesion size and increasing likelihood of malignancy. Even among masses measuring 5 mm or less for which targeted ultrasound was recommended, an ultrasound correlate was identified in 50%. For nonmasslike lesions measuring 10 mm or less for which targeted ultrasound was recommended, a correlate was found in only 13%. Even for nonmass lesions greater than 15 mm, a 1028 AJR:193, October 2009

5 Ultrasound of the Breast After Abnormal MRI TABLE 4: Tye of Lesion Versus Accurate and Inaccurate Ultrasound Correlation or Missed Malignancy Lesion tye Parameters Accurately Correlated (n = 497) Inaccurately Correlated (n = 22) Missed Cancers (n = 9) Mass 403 (81) 21 (95) 9 (100) Nonmass 94 (19) 1 (5) 0 (0) Size in all lesions (mm) a (7) 4 (18) 0 (0) (55) 12 (55) 8 (89) > (38) 6 (27) 1 (11) Breast density b Fatty 4 (1) 0 (0) 0 (0) Scattered fibroglandular 244 (52) 8 (36) 2 (22) Heterogeneously dense 195 (41) 11 (50) 7 (78) Extremely dense 29 (6) 3 (14) 0 (0) Masses 0 5 mm 36 (9) 4 (19) 0 (0) 6 10 mm 250 (62) 12 (57) 8 (89) > 10 mm 117 (29) 5 (24) 1 (11) Rim enhancement Dark internal setations Heterogeneous Homogeneous Kinetics c Washout 43 (33) 4 (31) 4 (66) Plateau 49 (38) 5 (38) 1 (17) Progressive 37 (29) 4 (31) 1 (17) Nonmasses 6 10 mm 23 (24) 0 (0) 0 (0) > 10 mm 71 (76) 1 (100) 0 (0) Clumed 79 (84) 1 (100) 0 (0) Nonclumed 15 (16) 0 (0) 0 (0) a Based on MRI measurements. b Twenty-five lesions had no recorded breast density. c No kinetic assessment on visual insection was recorded in 274 masses. correlate was found in only 42%. This information may aid radiologists who are deciding whether to recommend targeted ultrasound for susicious lesions seen on MRI. As with other imaging-guided biosies, careful correlation between biosy findings and imaging characteristics and follow-u MRI for benign concordant cases biosied with ultrasound guidance should be erformed to minimize the likelihood of missed cancers. References 1. Liberman L, Morris EA, Dershaw DD, Abramson AF, Tan LK. MR imaging of the isilateral breast in women with ercutaneously roven breast cancer. AJR 2003; 180: Tardivon AA, Guinebretiere JM, Dromain C, Vanel D. Imaging and management of nonalable lesions of the breast. Eur J Radiol 2002; 42: Bartella L, Smith CS, Dershaw DD, Liberman L. Imaging breast cancer. Radiol Clin North Am 2007; 45: Hashimoto BE, Morgan GN, Kramer DJ, Lee M. Systematic aroach to difficult roblems in breast sonograhy. Ultrasound Q 2008; 24: Wiratkaun C, Duke D, Nordmann AS, et al. Indeterminate or susicious breast lesions detected initially with MR imaging: value of MRI-directed breast ultrasound. Acad Radiol 2008; 15: LaTrenta LR, Menell JH, Morris EA, Abramson AF, Dershaw DD, Liberman L. Breast lesions detected with MR imaging: utility and histoathologic imortance of identification with US. Radiology 2003; 227: Sim LS, Hendriks JH, Bult P, Fook-Chong SM. US correlation for MRI-detected breast lesions in women with familial risk of breast cancer. Clin Radiol 2005; 60: Morris EA. Diagnostic breast MR imaging: current status and future directions. Radiol Clin North Am 2007; 45: Orel S. Who should have breast magnetic resonance imaging evaluation? J Clin Oncol 2008; 10: Swayamakula AK, Dillis C, Abraham J. Role of MRI in screening, diagnosis and management of breast cancer. Exert Rev Anticancer Ther 2008; 8: Stavros AT, Thickman D, Ra CL, Dennis MA, Parker SH, Sisney GA. Solid breast nodules: use of sonograhy to distinguish between benign and malignant lesions. Radiology 1995; 196: Raza S, Chikarmane SA, Neilsen SS, Zorn LM, Birdwell RL. BI-RADS 3, 4, and 5 lesions: value of US in management follow-u and outcome. Radiology 2008; 248: Berg WA, Blume JD, Cormack JB, Mendelson EB. Oerator deendence of hysician-erformed whole-breast US: lesion detection and characterization. Radiology 2006; 241: Hashimoto BE. Sonograhic assessment of breast calcifications. Curr Probl Diagn Radiol 2006; 35: Lloyd C, Hart JL, Niewiarowski S, Lim A, Harvey C, Cosgrove D. Ultrasound in breast imaging. Br J Hos Med (Lond) 2008; 69:M6 M9 16. Liberman L, Morris EA, Lee MJ, et al. Breast lesions detected on MR imaging: features and ositive redictive value. AJR 2002; 179: Nunes LW, Schnall MD, Orel SG. Udate of breast MR imaging architectural interretation model. Radiology 2001; 219: Kuhl CK, Mielcareck P, Klaschik S, et al. Dynamic breast MR imaging: are signal intensity time course data useful for differential diagnosis of enhancing lesions? Radiology 1999; 211: Beran L, Liang W, Nims T, Paquelet J, Sickle- Santanello B. Correlation of targeted ultrasound with magnetic resonance imaging abnormalities of the breast. Am J Surg 2005; 190: Lilienstein J, Daniel BL, Ikeda DM. In vivo sonograhy through an oen MRI breast coil to correlate sonograhic and MRI findings. AJR 2005; 184[sul 3]:S49 S Li J, Dershaw DD, Lee CH, Kalan J, Morris E. MRI follow-u after concordant, histologically benign diagnosis of lesions samles by MRI-guided biosy. AJR 2009; 193: AJR:193, October