Journal of Surgical Oncology 2011;104:741745

Standardized Pretreatment Breast MRIAccuracy and Inuence on Mastectomy Decisions

MATTHEW F BARCHIE, MD,1* KEVIN S CLIVE, MD,2 JOSHUA A TYLER, MD,2 JOSEPH B SUTCLIFFE, MD,1 AARON D KIRKPATRICK, MD,1 LISA M BELL, MD,1 KEVIN P BANKS, MD,1 SLAVA BELENKIY, MD,2 JEFF S SAENGER, MD,3 AND GEORGE E PEOPLES, MD2

Department of Radiology, San Antonio Uniformed Services Health Education Consortium, Brooke Army Medical Center, Fort Sam Houston, TX 2 Department of Surgery, San Antonio Uniformed Services Health Education Consortium, Brooke Army Medical Center, Fort Sam Houston, TX 3 Department of Pathology, San Antonio Uniformed Services Health Education Consortium, Brooke Army Medical Center, Fort Sam Houston, TX

Background and Objectives: Routine pretreatment breast magnetic resonance imaging in newly diagnosed cancer patients remains controversial. We assess MRI accuracy and inuence on mastectomy decisions after institution of standardized pretreatment MRI. Methods: A prospectively collected database of 74 consecutive new invasive breast cancer patients with pretreatment breast MRI was reviewed for treatment choice, radiologic, and pathologic results. Thirty-eight of 72 patients with available surgical records underwent mastectomy. Mastectomy preoperative and operative electronic records were reviewed for treatment decision analysis. Results: Seventeen of 72 (23.6%) invasive breast cancer patients were likely inuenced to undergo mastectomy by new information from MRI. MRI reported that the multifocal/multicentric (MF/MC) rate was 20 of 72 (27.8%) versus 19 of 72 (26.4%) by surgical pathology. MRI sensitivity for MF/MC disease was 89.5% versus 11.8% for mammography. MRI specicity was 84.2%. All three false positives declined recommended preoperative biopsies. MRI MF/MC diagnosis highly correlated with pathology results, P < 0.001. Conclusions: Increased mastectomy rate from 29 to 52.8% after standardization of pre-treatment breast MRI for invasive cancer is largely due to MRI ndings of increased extent of disease. These MRI ndings correlate well with pathologic ndings and appear to justify the performance of mastectomies in these patients.

J. Surg. Oncol. 2011;104:741745. 2011 Wiley Periodicals, Inc.

KEY WORDS: breast cancer; imaging; treatment choice

INTRODUCTION
Over 40,000 women die from breast cancer annually with the 2009 incidence projected at over 192,000 in the United States [1]. Over the last 40 years, breast cancer treatment has shifted from universal Halsted radical mastectomy to favoring breast conservation therapy (BCT), or more specically, lumpectomy with adjuvant radiation therapy [2]. Over 20 years of follow-up continues to show no survival benet of mastectomy over BCT [2,3]. However, while salvageable, BCT patients do have an increased local recurrence rate, particularly in younger patients [2,4]. Veronesi et al. in the Milan trial suggest the increased local recurrence is in keeping with common multifocal (MF) or multicentric (MC) disease in young women [4]. If multicentric disease is known preoperatively, BCT is contraindicated and mastectomy is performed. Mastectomy is also commonly performed when the extent of disease compared with breast size precludes good aesthetic result from BCT [5]. Typical mastectomy rates in the United States and Great Britain are reported around 30% and 27%, respectively [68]. MRI has been used in recent years to preoperatively assess the extent of disease. Extent of disease is listed as one of the three general indications for breast MRI by the American College of Radiology [9]. The strength of MRI is the increased sensitivity for MF and MC disease compared with mammography, particularly in the broglandular, heterogenously dense, or dense breast. MRI sensitivity for MF/ MC disease has been previously reported at 6789%, improving over the last decade [10]. However, there has been marked debate over the efcacy of MRI due to concerns over lack of improvement in

margin status or decrease in re-excision rates after BCT. Additional areas of concern have included delay of surgical treatment, specicity, and increased mastectomy rates [1113]. In September 2007, the interdisciplinary breast cancer clinic at Brooke Army Medical Center (BAMC), involving surgical oncology, medical oncology, radiology, radiation oncology, and pathology implemented standard pre-surgical bilateral breast MRI for biopsyproven invasive cancer. A prospective database was maintained. Our initial retrospective review of this database addressed the topic of BCT margin status, time to denitive surgical treatment, and mastectomy rates. Compared to consecutive patients from the preceding 9 months, patients with invasive cancers undergoing routine pretreatment MRI did not experience a delay in surgery, had a non-statistically signicant 30% reduction in re-excision rates after BCT, but did experience an increase in mastectomy rate from 29 to 52.8% (Clive K., unpublished data).
Conict of interest: None Disclaimer: All authored materials constitute the personal statements of the authors and are not intended to constitute an endorsement by the US Air Force, US Army, or other federal government entity. *Correspondence to: Matthew F Barchie, MD, Department of Radiology, Brooke Army Medical Center, 3851 Roger Brooke Drive, Fort Sam Houston, TX 78234-6200. Fax: (210) 916-1967. E-mail: mbarchie@satx.rr.com Received 30 November 2010; Accepted 4 April 2011 DOI 10.1002/jso.21960 Published online 25 May 2011 in Wiley Online Library (wileyonlinelibrary.com).

2011 Wiley Periodicals, Inc.

742

Barchie et al.
accuracy was also assessed with area under Receiver Operating Characteristic (ROC) curves. Repeated measures ANOVA test was used to assess differences between modality tumor measurements.

The aim of this study is to determine how much of the increased mastectomy rate is attributable to MRI and evaluate whether MRIs inuence is validated by nal pathology. If the rate of extensive disease on MRI is validated by pathology and is comparable with the rate of patients inuenced by MRI to obtain mastectomy then MRI can conceivably be utilized to assess feasibility of BCT without fear of disproportionate increase in mastectomies.

RESULTS
Of 74 consecutive patients with invasive disease who underwent standard preoperative MRI during the study period, 67 were invasive ductal carcinomas (IDC) and seven were invasive lobular carcinomas (ILC). Of the 74 with invasive disease, two received further care at an outside institution without available records and are excluded from the study. Thirty-eight of the remaining 72 patients (52.8%) received initial or short interval secondary mastectomy (Table I). Twenty of 38 MRIs were suggestive of MF/MC disease. The pathology proven MF/MC rate was 19 of 38 resulting in a sensitivity of 89.5% and positive predictive value of 85%. The MRI MF/MC diagnosis highly correlated with pathology results, P < 0.001 on x2test with continuity correction (Table II). MRI ROC analysis demonstrated strong discrimination with area of 0.853 (95%CI: 0.713 0.992, P < 0.001) (Table III). There were two false negative MRIs. One had an additional focus of DCIS and the other multifocal IDC. Specicity was 84.2%. There were three patients listed as false positives. All declined recommended further workup (i.e., targeted ultrasound biopsy). Two of the three had decided on mastectomy prior to MRI. The third was not eligible for BCT due to prior breast cancer and radiation and declined biopsy of the contralateral breast (which demonstrated atypical ductal hyperplasia on mastectomy pathology). Ten of 16 true negative MRIs for MF/MC disease yielded other useful positive ndings for staging and surgical planning including three with biopsy proven adenopathy, one signicantly upsized tumor size, and six characterized a previously ill-dened mass or calcication. Mammography reports are not available for three patients. Among the remaining 35 patients, there was no signicant relationship between mammography MF/MC diagnosis and pathology, P > 0.05 on sher exact test, with a sensitivity of 11.8% and specicity of 100% (Table II). Mammography failed to discriminate on ROC curve analysis, area 0.559 (95%CI: 0.3630.754) (Table III). Both MRI and mammography adenopathy diagnoses had statistically signicant relationships with pathology, MRI P 0.039 and mammography P 0.027; however, both modalities failed to discriminate on ROC curve analysis. MRI sensitivity and specicity for adenopathy were 50% and 87.5%, respectively. Mammography sensitivity and specicity were 33% and 100%, respectively (Tables II and III). Primary tumor size was measured to the millimeter on three modalities. Size information was available on 33 MRI, 25 mammogram, and 28 ultrasound reports. If no tumor was seen, the full tumor
TABLE I. Demographics and Disease Characteristics Mastectomy for invasive disease (n 38) Age (mean) Index tumor size mean ( SE) Primary pathology (number) Invasive ductal carcinoma Invasive lobular carcinoma Disease stage I II III IV

MATERIALS AND METHODS

This is a retrospective review of a prospectively collected database of all new invasive breast cancer cases from September 2007 to August 2009 after implementation of standardized preoperative bilateral breast MRI for biopsy-proven invasive cancer. The protocol was approved by the BAMC Institutional Review Board. All consecutive cases of invasive breast cancer receiving preoperative MRI and ultimately mastectomy were considered. MRI, mammography, and ultrasound designation of MF or MC disease, adenopathy, and primary tumor size were compared with the gold standard of mastectomy pathology. Multifocal disease was dened as more than one lesion within the same breast quadrant. Multicentric disease was dened as more than one lesion located in separate quadrants or diffused throughout the breast. Presurgical workup included diagnostic mammography with targeted ultrasound as indicated. Initial biopsy was performed with the most appropriate imaging modality, preferably ultrasound. Bilateral MRI was obtained upon positive biopsy for all invasive disease. Additional biopsies were performed with targeted ultrasound or MRI as appropriate. Screening axillary ultrasound was not standard. Mastectomy was recommended for multicentric disease. Surgical recommendation and ultimate treatment for multifocal disease varied based on extent of disease and residual breast. MF and MC disease are analyzed together in this study as their MRI appearance differs only in location, and MRI can be counted as responsible for the mastectomy decision whether the nding was MF or MC. Neo-adjuvant therapy to increase eligibility for BCT was not performed. MRI was conducted with 1.5 Tesla Marconi and Siemens Esprit magnets with InVivo dedicated seven channel breast coils. Core sequences include axial STIR, 3D axial T1, T2 sagittal fat saturated, pre and post dynamic contrast axial T1 SPGR fat saturated images with and without subtraction. Post-contrast images were repeated to 5 min with 60 sec temporal resolution. Slice thickness was 0.9 mm with 0.89 mm in-plane resolution. Matrix size was 381 448 with eld of view of 34 30cm. Coronal and sagittal reformats were obtained. All studies were performed at BAMC and interpreted by board certied BAMC staff radiologists. The MRI was scheduled at the earliest possible time, regardless of menstrual cycle, to avoid delay of denitive management, accepting possible negative impact on sensitivity and specicity [14]. Contrast administration was automated with a 0.1 mmol/kg gadolinium injection followed by 10 ml saline ush. Standard pathology practice evaluation of regions of interest was conducted by board certied BAMC pathologists. Electronic record review of preoperative and operative notes of mastectomy patients were reviewed for treatment selection decision. Patients whose treatment plan changed from BCT to mastectomy after the MRI or who were offered the option but waited to select mastectomy until after the MRI are listed as likely inuenced by MRI. x2-test and Fischer exact test were utilized by our hospital staff statistician for signicance of radiology to pathology relationship for MF/MC disease and adenopathy. Fischer exact test was used when one or more cells in the contingency table had an expected count of less than ve. Continuity correction was used with the x2-test yielding a more conservative (possibly underestimated signicance) relationship estimate between MRI and pathology. Modality
Journal of Surgical Oncology

58.7 years 2.7 (0.4) cm 34 4 11 (29%) 16 (42%) 9 (24%) 2 (5%)

One with superimposed Pagets. Standard error (SE).

Standardized Breast MRI Program Effects

TABLE II. Modality Performance Data: Multifocal/Multicentric Disease and Adenopathy Multifocal/multicentric diagnosis compared with mastectomy pathology MRI n 38 P < 0.00 TP FP TN FN Sensitivity Specificity PPV NPV Accuracy 17 3 16 2 89.5% 84.2% 85.0% 88.9% 86.8% Mammography n 34 P 0.49 2 0 17 15 11.8% 100% 100% 53.1% 55.9%

743

Adenopathy diagnosis compared with mastectomy pathology MRI n 38 P 0.03 11 2 14 11 50.0% 87.5% 84.6% 56.0% 65.8% Mammography n 35 P 0.027 7 0 14 14 33.3% 100% 100% 50% 60%

Meets statistical signicance criteria. Magnetic resonance imaging (MRI); true positive (TP); false positive (FP); true negative (TN); false negative (FN); positive predictive value (PPV); negative predictive value (NPV).

size was listed as error. Twenty-four patients had measurements or occult nding on all three modalities. Findings are summarized on Table IV with all modalities both over- and underestimating tumor size compared with gold standard surgical pathology. There was no signicant difference in tendency to over- or underestimate tumor size by modality, P > 0.05 with ANOVA. MRI measured two of 33 exactly, underestimated 17, and overestimated 16 by >1 mm. Mean measurement error (absolute value) for 24 comparable patients for MRI was 6.8 mm 1.1 mm standard error, mammography 8.3 1.6 mm, and ultrasound 8.7 1.2 mm. No signicant accuracy difference was found between modality with P > 0.05. MRI was the most accurate or tied for most accurate modality in 14 of 24 cases (58.3%), mammography in 11 of 24 (45.8%), and ultrasound in 6 of 24 (24%). Electronic medical record review of preoperative and operative notes of the 38 mastectomy patients showed 17 of the 38 were likely inuenced toward mastectomy by extent of disease seen on MRI, potentially accounting for the doubling of our mastectomy rate during the study period. Five patients requested mastectomy before the MRI or in spite of reassuring non-MF/MC MRI ndings. Nine were not good candidates for BCT prior to MRI (prior radiation for Hodgkins lymphoma, recurrent disease, male patient, Pagets with invasive disease, widespread metastatic disease, extension to pectoralis on excisional biopsy, and near whole breast disease on mammogram). Six of the patients failed BCT for positive margins. On MRI, four of those six had demonstrated additional regions of nonmasslike

enhancement separate from the primary tumor. The nonmasslike enhancement in three of these cases was suspicious for multifocal disease. The seventh BCT failure appears to have changed intraoperatively, however the reasoning is not well recorded. MRI linear enhancement extending to the pectoralis and surgically proven positive nodes were present in that case. Table V summarizes mastectomy rates and indications.

DISCUSSION
In this study, we have critically assessed the role of standard pretreatment MRI in newly diagnosed invasive breast cancer patients. This practice has increased our institutional mastectomy rate. Based on the analysis presented here, it would appear that the majority of the increased number of mastectomies, above the 29% baseline, were related to MRI ndings of increased extent of disease, predominately MF or MC disease, and correlated well with nal pathology in 87%. Our study further supports multifocal and multicentric diagnosis validity on MRI. MRI MF/MC diagnosis had a highly signicant relationship compared to pathology with a sensitivity of 89.5% and specicity of 84%. Our data are consistent with the 89% MRI sensitivity for invasive disease described by Sardanelli et al., who utilized whole breast pathologic examination [10]. It is comparable to previously published sensitivities for MF disease 82% and multicentric disease 89% [10]. In comparison, mammographic diagnosis of MF/ MC disease was limited by low sensitivity and without signicant relationship to nal pathology. Specicity is critical when evaluating increased mastectomy rates. MRI appears to have performed well. The specicity could have been as high as 100%; however, it was reduced by three false positives all of which refused recommended biopsies of suspicious MRI ndings. Mastectomy may have been avoided with negative biopsies, highlighting the integral role biopsy plays when utilizing breast MR. Turnbull et al. in the COMICE trial similarly demonstrated the importance of additional biopsies if MRI is utilized. Of the 16 (2%) avoidable mastectomies attributed to MRI, only three were known to have had additional biopsies. Six did not have additional biopsies, and data are missing for seven. The background avoidable mastectomy rate for non-MRI patients was approximately 0.5% [13]. In our study, 17 of 72 (23.6%) patients with invasive disease received 18 additional biopsies based on their MRI. Our experience with standardized pretreatment breast MRI has found that this

TABLE III. Modality Performance Data: Receiver Operating Characteristic (ROC) Curve Analysis Multifocal/multicentric diagnosis MRI ROC 95% CI P value 0.853 0.7130.992 P < 0.001 Discriminates Mammography 0.559 0.3630.754 P > 0.05 MRI Adenopathy Mammography

0.690 0.667 0.5130.868 0.4890.844 P > 0.05 P > 0.05 Fails to discriminate

*0.600 < 0.700 Poor; 0.700 < 0.800 Fair; 0.800 < 0.900 Good; 0.900 < 1.0 Excellent test discrimination. Magnetic resonance imaging (MRI), CI (condence interval).

Journal of Surgical Oncology

744

Barchie et al.
TABLE IV. Modality Performance Data: Index Tumor Size Index mass size measurement by modality compared with final pathology MRI Underestimated (# cases) Overestimated (# cases) Exact millimeter (# cases) Millimeters > or < pathology (SE)

Mammography 17 (68%) 7 (28%) 1 (4%) 8.3 (1.6)

Ultrasound 23 (82%) 5 (18%) 0 8.7 (1.2)

16 (48.5%) 15 (45.5%) 2 (6%) 6.8 (1.1)

No signicant difference in accuracy between modalities (P > 0.05). Standard Error (SE).

practice does not delay patient care (Clive K, unpublished data). We recommend early call back and coordination for immediate MRI scheduling after biopsy positive invasive disease. Our preferred method of biopsy is through targeted ultrasound. This method was successful in 65% of MRI masses and 12% of nonmasslike enhancement cases at the University of Chicago [15]. However, MR biopsy capability should be available if MR is to be used. MRI was far superior to mammography in diagnosis of MF/MC disease (Table II). MRI superiority to standard modalities is less denitive, however, for tumor size and adenopathy. MRI was only slightly more accurate than typical modalities, at 6.8 mm average error, from actual tumor size without statistical difference. The magnitude of error is consistent with 6.3 mm found by Onesti et al. which included DCIS [16]. However, unlike Onesti et al., our data did not consistently overestimate. MRI underestimated more than half of reported cases, possibly reecting the exclusion of new DCIS cases, which tends to present as less dened nonmasslike enhancement. The variability we found in under and overestimation may limit the ability of MRI to improve surgical margin/re-excision rates for BCT. In the COMICE trial the MRI predicted management was mastectomy for 144 of 691 patients (21%), yet mastectomy was the initial operation for only 7% of MRI arm patients [13]. Perhaps reliance on tumor size and location, i.e., spatial resolution, and reluctance to convert to mastectomy resulted in the lack of improvement in re-excision rates. MRIs limitations in dening appropriate surgical boundaries are unfortunate but not completely unexpected. The strength of MRI is contrast resolution, i.e., identifying disease, more than spatial resolution. Non-mass like enhancement is a good example. MRI was only slightly more sensitive than mammography for adenopathy, 50% versus 33%. Mixed mathematical signicance was found, with statistical signicance by contingency test, but failure to discriminate by ROC curve analysis. MRI mass size and adenopathy ndings included in the decision process should be considered with these limitations in mind.

TABLE V. Mastectomy Rates and Indications Mastectomy for invasive disease (n 38) Mastectomy rate Prior to standard pre-op breast MR After standard pre-op breast MR Mastectomy indication (number of patients) Extent of disease on MR Failed BCT Patient elected Prior cancer/same side chest radiation Widespread or metastatic disease pre-MR Male with invasive breast cancer Pagets disease 29% 52.8% 17 7 5 3 4 1 1

Our previously reported increase in mastectomy rate from 29% to 52.8% is partially, but not completely, explained by pre-operative MRI. MRI provided new information which inuenced patients toward mastectomy at a rate similar to the pathologically proven rate of MF/MC disease, 23.6% versus 26.4% respectively. MRI ndings which inuenced patientphysician inuence decisions were predominately MF or MC disease but also included tumor size, extension to muscle, and adenopathy. The remaining 29.2% of the 72 patients in the study (21 of the 38 mastectomies) received mastectomy for other reasons as previously outlined. 13.2% of our mastectomies were due solely to patient choice and 18.4% were from conversion from BCT. Not all patients with multifocal disease received mastectomy. In fact MRI indicated multifocal disease in three of the seven BCT failures. The line where mastectomy becomes the preferred treatment option varies by institution, surgeon, and patient. Our mastectomy rate after standardized MRI of 52.8% is above the national average of approximately 30%. Other recent studies have also shown elevated mastectomy rates for invasive cancer. Lee et al. found a mastectomy rate of 47% (when neo-adjuvant chemotherapy to increase eligibility for BCT was excluded). With aggressive neo-adjuvant treatment they found a mastectomy rate of 38% for invasive cancer [5]. The practice of neo-adjuvant therapy to allow BCT is not uniformly practiced nationwide. Recent studies have shown no change in survival at 9 years with a trend toward increased ipsilateral breast tumor recurrence and small increases in long-term mortality by meta analysis [2]. A component of our increased mastectomy rate may also reect generally rising mastectomy rates as reported in regional registries such as the State of California, and reported in single centers such as the Mayo Clinic [17,18]. The Mayo Clinic found mastectomy rates increased from 31% in 2003 to 37% in 2004, and 46% in 2006. Pretreatment MRI and surgical year were independent predictors of mastectomy with rates increasing predominantly within non-MRI patients from 2004 to 2006 [18]. Finally the high mastectomy rate in our study incorporates the above average pathology proven rate of MF/MC disease in our study {i.e., 26.4% compared to the median 16% by meta analysis (range 128%)} [12]. BCT is the standard of care if indicated and desired by the patient. Patient preference for mastectomy is a topic which is currently receiving much attention. Adjuvant radiation therapy is not completely risk free. There is a reported small, but signicant excess incidence of contralateral breast cancer, rate ratio 1.18, with a 1.3% increase in mortality at 15 years from contralateral breast cancer, heart disease, and lung cancer [2,19]. Patients also may want to avoid side effects and inconveniences of BCT adjuvant therapy. Ma et al. report a noncompliance rate for adjuvant radiation of 4% which increased risk of recurrence [20,21]. Noncompliance was 7% for chemotherapy and 37% for Tamoxifen [20]. Similarly 4.6% of BCT patients and 23.4% of mastectomy patients with strong indications for adjuvant radiation did not receive it in a 2010 study utilizing the SEER database. Patient desire to avoid radiation was listed as one of

Journal of Surgical Oncology

Standardized Breast MRI Program Effects

several signicant factors [22]. While requiring further investigation, visualization of the extent of disease on MRI may have a secondary role in encouraging compliance with BCT or mastectomy adjuvant therapy. We acknowledge the limitations of our small sample size and retrospective nature of the study which is prone to selection bias. The standardized administration of pre-treatment MRI in the consecutive cancer cases during our accrual mitigates this to some degree. The small sample size may also lead to a type II error particularly when comparing modality tumor measurement accuracies. Thus the strength of our study is not the magnitude of the mastectomy rates but the demonstrated concordance of MRIs inuence on patient physician mastectomy decisions and nal pathology.

745

CONCLUSIONS
Findings suggest that an increased mastectomy rate after standardization of pre-treatment breast MRI for invasive cancer is largely due to MRI ndings of increased extent of disease. These MRI ndings correlate well with pathologic ndings and appear to justify the performance of mastectomies in these patients.

ACKNOWLEDGMENTS
The authors thank John A. Ward, Ph.D., Research Physiologist, Department of Clinical Investigation, BAMC for providing the statistical analysis of this project.

REFERENCES
1. Ahmedin J, Siegel R, Ward E, et al.: Cancer statistics 2009. CA Cancer J Clin 2009;59:225249. 2. Crago A, Azu M, Tierney S, Morrow M: Randomized clinical trials in breast cancer. Surg Oncol Clin N Am 2010;19:3358. 3. Fischer B, Anderson S, Bryant J, et al.: Twenty-year follow-up of a randomized trial comparing total mastectomy, lumpectomy, and lumpectomy plus irradiation for the treatment of invasive breast cancer. N Engl J Med 2002;347:12331241. 4. Veronesi U, Cascinelli N, Mariani L, et al.: Twenty year followup of a randomized study comparing breast-conserving surgery with radical mastectomy for early breast cancer. N Engl J Med 2002;34:12271232. 5. Lee MC, Rogers K, Grifth K, et al.: Determinants of breast conservation rates: Reasons for mastectomy at a comprehensive cancer center. Breast J 2009;15:3440. 6. Morrow M, Jagsi R, Alderman A, et al.: Surgeon recommendations and receipt of mastectomy for treatment of breast cancer. J Am Med Assoc 2009;302:15511556. 7. Greenwood, A., Exploring the Reasons Women Choose Mastectomies. National Cancer Institute Cancer Bulletin. Available at: http://www.cancer.gov/ncicancerbulletin/111709/page5 [accessed 18 April 2010].

8. Kirby R, Basit A, Manimaran N: Patient choice signicantly affects mastectomy rates in the treatment of breast cancer. Int Semin Surg Oncol 2008;5:20. 9. American College of Radiology (2008). ACR practice guideline for the performance of magnetic resonance imaging (MRI) of the breast. Available at: http://www.acr.org/secondarymainmenucategories/quality_safety/guidelines/breast/mri_breast.aspx [accessed on 5 March 2010]. 10. Sardanelli F, Giuseppetti G, Panizza P, et al.: Sensitivity of MRI versus mammography for detecting foci of multifocal breast cancer in fatty and dense breasts using the whole-breast pathologic examination as a gold standard. Am J Roentgenol 2004; 183:11491157. 11. Bleicher R, Ciocca R, Egleston B, et al.: Association of routine pretreatment magnetic resonance imaging with time to surgery, mastectomy rate, and margin status. J Am Coll Surg 2009;209: 180187. 12. Houssami N, Hayes D: Review of preoperative magnetic resonance imaging (MRI) in breast cancer: Should MRI be performed on all women with newly diagnosed, early stage breast cancer? CA Cancer J Clin 2009. Available at: http://caonline.amcancersoc.org/cgi/content/full/caac.20028v1 [accessed 25 February 2010]. 13. Turnball L, Brown S, Harvey I, et al.: Comparative effectiveness of MRI in breast cancer (COMICE) trial: A randomised controlled trial. Lancet 2010;375:563571. 14. Morris E, Liberman L: Breast MRI: Diagnosis and intervention. New York: Springer Science and Business Media; 2005. 15. Abe H, Schmidt R, Shah R, et al.: MR-directed (secondlook) ultraound examination for breast lesions detected initially on MRI:MR and sonographic ndings. Am J Roentgenol 2010;194:370377. 16. Onesti J, Mangus B, Helmer S, Osland J: Breast cancer tumor size: Correlation between magnetic resonance imaging and pathology results. Am J Surg 2008;196:844850. 17. Gomez S, Lichtenstajn D, Kurian A, et al.: Increasing mastectomy rates for early-stage breast cancer? population-based trends from California. J Clin Oncol 2010; Available at: http:// jco.ascopubs.org/cgi/doi/10.1200/JCO.2009.26.1032 18. Katipamula R, Degnim A, Hoskin T, et al.: Trends in mastectomy rates at the Mayo Clinic Rochester: Effect of surgical year and preoperative magnetic resonance imaging. J Clin Oncol 2009;27:40824088. 19. Clarke M, Collins R, Darby S, et al.: Early Breast Cancer Trialists Collaborative Group (EBCTG). Effects of radiotherapy and of differences in the extent of surgery for early breast cancer on local recurrence and 15-year survival: An overview of the randomized trials. Lancet 2006;366:20872106. 20. Ma AM, Barone J, Wallis AE, et al.: Noncompliance with adjuvant radiation, chemotherapy, or hormonal therapy in breast cancer patients. Am J Surg 2008;196:500504. 21. Huston T, Simmons R: Locally recurrent breast cancer after conservation therapy. Am J Surg 2005;189:229235. 22. Jagsi R, Abrahamse P, Morrow M, et al.: Patterns and correlates of adjuvant radiotherapy receipt after lumpectomy and after mastectomy for breast cancer. J Clin Oncol 2010;28:23962403.

Journal of Surgical Oncology