The efficacy of breast MRI in predicting breast conservation therapy

Journal of Surgical Oncology 2006;94:220–225

The Efficacy of Breast MRI in Predicting BreastConservation Therapy

SARAH BLAIR, MD,1* MICHELE MCELROY, MD,1 MICHAEL S. MIDDLETON, MD, PhD,2 CHRIS COMSTOCK, MD,1,2

TANYA WOLFSON, MA,1 MITCH KAMRAVA, MD,3 ANNE WALLACE, MD,1 AND JOANNE MORTIMER, MD1

1Moore’s Cancer Center, University of California San Diego, San Diego, California2Department of Radiology, University of California San Diego, San Diego, California

3Department of Family and Preventive Medicine, Division of Biostatistics, University of California San Diego,San Diego, California

Background: Breast conservation therapy (BCT) has equal efficacy compared tomastectomy in treating breast cancer. Accurate pre-operative measurement of tumorsize can limit re-excision procedures. Breast MRI may improve pre-operativeevaluation of extent of disease.Objective: To examine the correlation of extent of disease on breast MRI withpathologic data to determine the utility of breast MRI in surgical planning of BCT.Methods: We retrospectively reviewed our prospective database of women under-going breast MRI. We identified 115 women with breast cancer who underwent abreast MRI and a surgical resection from 2000 to 2003. We compared patients withhigh-grade tumors (HG, n¼ 40) to patients with low grade (LG, n¼ 75).Results: The size of the tumor on MRI correlated with the pathologic size for HGtumors (HG R¼ 0.76 vs. LG R¼ 0.45, P¼ 0.033). Mastectomy was performed in53 patients. In 10 patients with LG tumors, the MRI findings overestimated theirdisease. In 11 out of 115 patients the primary tumor or a second tumor was only seenby MRI.Conclusion: Breast MRI does change surgical management by detecting additionalmalignancies. Breast MRI is accurate in staging extent of disease in the breast inpatients with HG tumors.J. Surg. Oncol. 2006;94:220–225. � 2006 Wiley-Liss, Inc.

KEY WORDS: breast MRI; BCT; breast cancer

BACKGROUND

The survival for women undergoing breast conserva-tion has been shown to be equivalent to mastectomy.Multiple randomized prospective trials with greater than10-year follow-up have proved that breast conservationtherapy (BCT) has equal efficacy compared to mastect-omy in treating early stage breast cancer. At present BCThas become the standard of care to treat this malignancy[1–3]. Patients with large tumors or multi-centric diseaseare more likely to recur in the breast and should beconsidered for mastectomy. Thus accurate pre-operativemeasurement of tumor size can limit multiple re-excisionprocedures in order to obtain negative margins whilemaintaining a good cosmetic result from BCT.In screening women at high risk for developing breast

cancer, MRI appears to be more sensitive in detecting

breast cancer than mammogram, ultrasound, or clinicalbreast exam [4–6]. Therefore, investigators have exam-ined the use of MRI to improve breast cancer staging ofprimary tumors as well as detect multi-centric andcontralateral disease. In patients with known breastcancers, breast MRI has been reported to have highersensitivity rates particularly compared to mammographywhen patients are studied pre-operatively as well as pre-operatively after neoadjuvant chemotherapy. Studies

*Correspondence to: Sarah Blair, MD, Assistant Professor of Surgery, 3855Health Sciences Dr., San Diego, CA 92093-0987. Fax: 858-822-6366.E-mail: [email protected]

Received 19 August 2005; Accepted 22 February 2006

DOI 10.1002/jso.20561

Published online in Wiley InterScience (www.interscience.wiley.com).

� 2006 Wiley-Liss, Inc.

have reported sensitivity rates of >90% for MRIcompared to 50–60% with mammography [7–10].However, MRI has reported low specificity rates of 30–60%. It often detects secondary lesions with a positivepredictive value of 30–40% [11]. Most previous studiesin breast MRI have been relatively small retrospectivereviews from single institutions studying homogenousgroups of patients [7–14]. The purpose of this study is toexamine our experience using MRI in the pre-operativeassessment of women with newly diagnosed primarybreast cancer. We plan to study the correlation ofextent of disease on MRI with pathologic data todetermine the utility of breast MRI in surgical planningin breast cancer.

METHODS

We reviewed our prospective database of 115 con-secutive patients with known breast cancers who under-went breast MRI as part of their treatment at theUniversity of California at San Diego over the period,2000–2003. After obtaining Institutional Review Boardapproval we retrospectively reviewed their records fordemographic information and patient outcome. BreastMRIs were read by the two radiologists at our institution(C.C., M.M.). All patients had mammograms and theircancers were identified through either routine imagingmammogram or ultrasound or were palpable on physicalexam. Pathologic assessment of all tumor specimens wasreviewed at our institution.

MRI Technique

Breast MRI was performed on a 1.5-T whole-bodyimager using a bilateral breast coil (Siemens Symphony,syngo MR, Malvern, PA). Patients were placed in theprone position and spacers were added as needed toprevent breast motion. T1-weighted gradient-echoimages were acquired through each axilla separatelywith the body coil. Axial fat saturated T2-weightedimages were acquired of each breast separately with thebreast coil. Pre-contrast axial 3D fast spoiled gradientrecalled echo T1-weighted images without fat saturation,were acquired of both breasts simultaneously. Pre-contrast axial 3D SPGR images were obtained of bothbreasts simultaneously with fat saturation, followed byfive identical sequences of 1 min duration each after a20 sec interval during which contrast and saline weregiven. Delayed high contrast axial 3D SPGR imageswere then obtained of both breasts simultaneously withthe breast coil with the center of k-space at the start ofthe sequence axial plane. In pre-menopausal women theexam was performed approximately 14 days after thelast period.

Pathologic Analysis

All breast lumpectomy and mastectomy specimenswere submitted for serial gross pathological examinationand subsequent histological analysis. The longest dia-meter of disease both invasive and pre-invasive and thehistological types were recorded. The Modified BloomRichardson score from 0 to 9 was noted; scores >7 wereconsidered high grade.

Statistics

Spearman correlation coefficients were computedbetween breast MRI measurement of disease andpathological measurement, both overall and by sub-groups. Paired Wilcoxon tests were used to test forsystematic differences between MRI and pathologymeasurements (such a difference would not affect thestrength of the correlation).

RESULTS

Demographic Data

We identified 115 consecutive patients who had aknown diagnosis of breast cancer, underwent bilateralbreast MRI, and who subsequently underwent a definitivesurgical treatment for breast cancer between 2000and 2003 at our institution. The age range for womenincluded in the study was 31–78 years with a median ageof 52 years. The reasons for obtaining a breast MRI inaddition to mammography or ultrasound are shownin Figure 1. Thirty women underwent a prior tumorexcision and had a positive margin. The MRI wasobtained to determine if extensive residual diseasewas present at the surgical site and to look for multi-focal disease in both breasts which would precludeBCT. The study was obtained within 1 month in orderfor patients to complete their cancer care in a timelymanner. Six patients had a MRI for evaluation oflobular carcinoma. Twenty-six patients had a MRI aspart of a protocol for neoadjuvant chemotherapy and53 women had ‘‘indeterminate findings’’ on mammo-graphy including dense breasts which were belowdetection of mammography. Thus, 85 patients had aMRI prior to any surgical management. In 11 patientssynchronous disease was identified only on MRI ineither the same or contralateral breast that changedclinical management. For example in Figure 2, a patientpresented with an axillary mass and a negative mammo-gram. Breast MRI identified the primary tumor in thebreast; therefore, the patient was able to undergo breastconservation. In two patients the primary tumors wereonly palpable and MRI detected second tumors in ninepatients.

Journal of Surgical Oncology DOI 10.1002/jso

Breast MRI and BCT 221

Correlation of Breast MRI and Pathology

Overall there was a significant correlation betweenthe measurement of tumor appearance on breast MRIand pathologic of total amount of tumor (pre-invasiveplus invasive cancer) (Spearman rho¼ 0.51, P-value> 0.001). Similarly, the correlation was significantbetween size by MRI and total tumor size in subsets ofpatients who had a previous chemotherapy or a previousexcision with a positive margin (Spearman rho¼ 0.50, P-value¼ 0.005, Spearman rho¼ 0.53, P-value¼ 0.003,respectively).We also separated the patients into 46 patients with

high-grade (HG) tumors and 69 patients with low-grade(LG) or intermediate-grade tumors. In the patientswith HG tumors the correlation between pre-operativebreast MRI measurement and pathologic measurementof the tumor was considerably higher (Spearmanrho¼ 0.73, P-value< 0.0001, Fig. 3). The majority ofthe patients with HG tumors were younger than 60 years

of age (92%) and were estrogen receptor negative(92%). Only 5 patients over the age 60 and 25 patientswere estrogen receptor negative; 7 patients withHG lesions were HER 2 neu positive (Table I).Additionally, paired Wilcoxon Signed Rank test wasused to compare MRI versus pathology values for allpatients, LG and HG tumor patients. The object of the testwas to assess whether MRI predicted a tumor sizecomparable to that of the pathology evaluation, asopposed to generally higher or lower. All three resultswere not significant, suggesting no systematic differencesbetween the MRI tumor size and pathology tumor size(i.e., MRI does not systematically under- or overestimatethe tumor size).


Fig. 1. Indications for breast MRI. Of the 115 women included inthis study 30 women underwent breast MRI after a previous excision.Six women had lobular carcinomas. Twenty-six women were followedduring neoadjuvant chemotherapy and 53 underwent MRI due toindeterminate findings on breast mammography.

Fig. 2. Example of breast cancer only seen on MRI. A: Bilateralscreening mammogram of patient later found to have invasive cancer inthe right breast: these films were read as BIRADS category 2 due toheterogeneously dense breasts. B: MRI of the same patient: this filmwas read as showing an 11� 17 mm lesion with irregular borders in theright upper outer quadrant suspicious for malignancy.

222 Blair et al.

Clinical Findings

Sixty-two patients underwent breast conservationsurgery and 53 patients were treated with mastectomy.Sentinel node biopsy alone was performed in 52 patientsand 53 underwent a complete axillary node dissection. InTable II we summarize the reasons for mastectomy(Table II). As a result of MRI findings, 21 women who

were initially considered for BCT underwent mastect-omy. A larger volume of disease on MRI (13 patients)multi-centricity (8 patients) or residual disease from priorexcision (3 patients) necessitated removal of the breast.In 10 patients the MRI overestimated the disease andthese patients underwent a mastectomy and in 11 patientsthe MRI correctly diagnosed the indication for mastect-omy; thus giving a positive predictive value of 47%.


TABLE I. Clinicopathologic Features

Clinicopathologic feature Incidence

Age Mean 52 years

Range 31–78

Menopausal status 64 (56%) pre-menopausal

51 (44%) post-menopausal

Grade 40 (35%) high

75 (65%) low

ER/PR 64 (56%) positive

51 (44%) negative

Her 2 neu 11 (10%) positive

104 (90%) negative

Histology 108 (94%) ductal

7 (6%) lobular

Fig. 3. Radiographic and pathologic size correlation for high-grade tumors, low-grade, and all tumors.

TABLE II. Reasons for Mastectomy

Patient reasons N¼ 19

Patient choice 8

Family history 3

Inflammatory breast cancer 2

Cancer recurrence after BCT 4

Positive margins after multiple lumpectomy 2

Imaging reasons N¼ 34

Residual disease on MRI after excision 3

Multi-centric disease on MRI 8

Large area disease on mammogram and MRI 13

Large area disease on MRI alone 10

MRI corrected estimated disease 11

MRI overestimated disease 10


DISCUSSION

This series is one of the largest in the literature toreport unselected collected consecutive cases of breastMRI in breast cancer patients prior to definitive surgicalmanagement. This article reports the reasons cliniciansadd a breast MRI in addition to mammography andultrasound for pre-operative evaluation of tumor. Themost common reason cited was inconclusive mammo-gram because of dense breasts. This series represents aselected population of patients with difficult exams andmammograms that clinicians felt they needed additionalinformation. As the incidence of breast cancer increasesand younger patients present with the disease, mammo-graphy may not be the optimal tool to identify the extentof disease in these patients. In our series, in about 10% ofpatients we identified synchronous disease visualized onMRI only.In contrast to older modalities such as mammography

and ultrasound, MRI uses contrast media and the rate ofuptake of contrast to further distinguish suspicious areasin the breast. This additional information leads to theincrease in detectability of synchronous tumors. Therate of uptake of contrast is related to the vascularity ofthe lesion in the breast in relation to the surroundingbreast tissue. Therefore, patients with proliferativebreasts have whole breast increase in vascularity makingidentification of lesions by all imaging modalities morechallenging [15].In this data set we did include a heterogenous set of

patients. Some of the patients had either previouschemotherapy or excisional biopsies completed whichmake interpretation of their imaging more complicated.For example, some researchers have pointed out thatthere can be a significant overlap in the appearances ofbenign and malignant lesions in determining residualdisease in patients who have had a previous excision[12,13]. Inflammatory changes seen after a previousoperation increase vascularity to that area of the breastcausing increase of contrast uptake. In our series themajority of patients in the previous excision group alsohad LG tumors making interpretation of their MRIimages even more difficult. Although several serieshave reported a high correlation of breast MRI and

pathologic response after neoadjuvant chemotherapy,some researchers have pointed out that the determinationof residual tumor size is unreliable in carcinomasexhibiting significant response to chemotherapy [7].Important factors that contribute to this discrepancyinclude tumor vascularity and the permeability of thevascular wall are responsible for contrast mediumdynamics visualized by breast MRI. Therefore, patientswho have had a significant response to chemotherapymay have less tumor vascularity and may lead tounderestimation of disease on MRI. Factors such asfibrocystic changes, previous excision, and changes invascularity after neoadjuvant chemotherapy may explainthe overall relatively small correlation of pre-operativebreast MRI size and pathologic measurement comparedto homogenous series in the literature [8,10,12,15](Table III). Although for all tumors there was a significantcorrelation between breast MRI measurable disease andfinal pathological size all that the P-value tells you is thatthe correlation coefficient is significantly different fromzero (zero being no correlation at all) given that samplesize. The higher the correlation, the more clinicallyimportant the correlation is. Therefore, we found in HGtumors that there was a higher correlation between tumorsize on pre-operative MRI and pathologic measurement.This fact gives surgeons the confidence to apply BCT inappropriate patients. Breast MRI should be considered inany patient in whom mammography may be under-estimating the extent of disease such as in patients withheterogeneously dense breasts, patients with implants,lobular carcinoma, or unknown breast primary tumor.

Study Limitations

This article presents data from a retrospective reviewof a prospective database; therefore, there are limitationssecondary to a retrospective review. For example, therewas no special attention made to pathologic examinationto correlate MRI finding in patients with false positiveresults and likewise there was no ability to biopsysynchronous disease seen only on MRI. Furthermore, thepower of subset analysis may be limited because of therelatively small numbers in each group. We included aheterogenous group of patients with different indications


TABLE III. Correlation of Pre-Operative MRI Measurement of Tumor and PathologicMeasurement in the Literature

Author N Reason for MRI

Correlation

coefficient P-value

Hata et al. [14] 54 Ductal spread of early stage cancer 0.42 0.001

Partridge et al. [10] 52 Neoadjuvant chemotherapy 0.89 0.001

Thibault et al. [8] 30 Neoadjuvant chemotherapy 0.79 0.01

Blair et al. (present series) 40 High-grade tumors 0.73 0.001

Blair et al. (present series) 115 All tumors 0.51 0.05

224 Blair et al.

for breast MRI including patients with previous excisionand positive margins which are particularly challengingto examine on MRI. However, we feel these data areimportant because they represent the daily experience ofclinicians in treating breast cancer patients.

In conclusion, we did find that MRI changed oursurgical management by identifying additional tumorspre-operatively. We recommend breast MRI in pre-operative planning in patients where mammographymay be sub-optimal regardless of the grade of the tumor.In this series MRI-guided biopsy technology was not yetavailable at our institution. In future, MRI-guidedbiopsies may be used to confirm the extent of diseaseand this practice may improve our ability to accuratelystage breast tumors pre-operatively.

REFERENCES

1. Fisher B, Anderson S, Bryant J, et al.: Twenty-year follow-up of arandomized trial comparing total mastectomy, lumpectomy, andlumpectomy plus irradiation for the treatment of invasive breastcancer. N Engl J Med 2002;347:1233–1241.

2. Poggi MM, Danforth DN, Sciuto LC, et al.: Eighteen-year resultsin the treatment of early breast carcinoma with mastectomy versusbreast conservation therapy: The National Cancer InstituteRandomized Trial. Cancer 2003;98:697–702.

3. Nakamura S, Woo C, Silberman H, et al.: Breast-conservingtherapy for ductal carcinoma in situ: A 20-year experience withexcision plus radiation therapy. Am J Surg 2002;184:403–409.

4. Kriege M, Brekelmans CT, Boetes C, et al.: MRI screening forbreast cancer in women with familial or genetic predisposition:Design of the Dutch National Study (MRISC). Fam Cancer 2001;1:163–168.

5. Hartman AR, Daniel BL, Kurian AW, et al.: Breast magneticresonance image screening and ductal lavage in women at highgenetic risk for breast carcinoma. Cancer 2004;100:479–489.

6. Warner E, Plewes DB, Hill KA, et al.: Surveillance of BRCA1 andBRCA2 mutation carriers with magnetic resonance imaging,ultrasound, mammography, and clinical breast examination.JAMA 2004;292:1317–1325.

7. Rieber A, Brambs HJ, Gabelmann A, et al.: Breast MRI formonitoring response of primary breast cancer to neo-adjuvantchemotherapy. Eur Radiol 2002;12:1711–1719.

8. Thibault F, Nos C, Meunier M, et al.: MRI for surgical planning inpatients with breast cancer who undergo preoperative chemother-apy. AJR Am J Roentgenol 2004;183:1159–1168.

9. Weatherall PT, Evans GF, Metzger GJ, et al.: MRI vs. histologicmeasurement of breast cancer following chemotherapy: Compar-ison with x-ray mammography and palpation. J Magn ResonImaging 2001;13:868–875.

10. Partridge SC, Gibbs JE, Lu Y, et al.: Accuracy of MR imaging forrevealing residual breast cancer in patients who have undergoneneoadjuvant chemotherapy. AJR Am J Roentgenol 2002;179:1193–1199.

11. Harms SE, Flamig DP, Hesley KL, et al.: MR imaging of thebreast with rotating delivery of excitation off resonance: Clinicalexperience with pathologic correlation. Radiology 1993;187:493–501.

12. Lee JM, Orel SG, Czerniecki BJ, et al.: MRI before reexcisionsurgery in patients with breast cancer. AJR Am J Roentgenol2004;182:473–480.

13. Bedrosian I, Mick R, Orel SG, et al.: Changes in the surgicalmanagement of patients with breast carcinoma based onpreoperative magnetic resonance imaging. Cancer 2003;98:468–473.

14. Hata T, Takahashi H, Watanabe K, et al.: Magnetic resonanceimaging for preoperative evaluation of breast cancer: Acomparative study with mammography and ultrasonography.J Am Coll Surg 2004;198:190–197.

15. van den Bosch MA, Daniel BL, Mariano MN, et al.: Magneticresonance imaging characteristics of fibrocystic change of thebreast. Invest Radiol 2005;40:436–441.



The efficacy of breast MRI in predicting breast conservation therapy

Documents

Transcript of The efficacy of breast MRI in predicting breast conservation therapy