Contrast-enhanced magnetic resonance mammography: does it affect surgical decision-making in patients with breast cancer?
- First Online:
- Cite this article as:
- Pediconi, F., Catalano, C., Padula, S. et al. Breast Cancer Res Treat (2007) 106: 65. doi:10.1007/s10549-006-9472-9
- 95 Views
Diagnostic imaging in women with suspected breast cancer should accurately detect and diagnose malignant tumors and facilitate the correct choice of therapy. Contrast-enhanced magnetic resonance mammography (CE-MRM) is potentially the imaging modality of choice for accurate patient management decisions.
A total of 164 women with suspected breast cancer based on clinical examination, conventional mammography and/or ultrasound each underwent preoperative bilateral CE-MRM using an axial 3D dynamic T1-weighted gradient-echo sequence and gadobenate dimeglumine as contrast agent. Images were evaluated by two readers in consensus. Histological evaluation of detected lesions was performed on samples from core biopsy or surgery. Determinations were made of the sensitivity, accuracy and positive predictive value of CE-MRM compared to mammography/ultrasound for the detection of malignant lesions and of the impact of CE-MRM for surgical decision-making.
Conventional mammography/ultrasound detected 175 lesions in the 164 evaluated patients. CE-MRM revealed 51 additional lesions in 34/164 patients; multifocal and multicentric cancer was detected in 7 and 4 additional patients, respectively, contralateral foci in 21 additional patients and pectoral muscle infiltration in 2 additional patients. CE-MRM also confirmed the absence or benignity of 3 and 1 lesions suspected of malignancy on mammography/ultrasound. The sensitivity and accuracy for malignant lesion detection and identification was 100% and 93.4%, respectively, for CE-MRM compared to 77.3% and 72.1% for mammography/ultrasound, respectively. Patient management was altered for 32/164 (19.5%) patients as a result of CE-MRM.
CE-MRM positively impacts patient management decisions and should be performed in all women with suspected breast cancer based on clinical examination, mammography and/or ultrasound.
KeywordsBreastBreast neoplasms MRCancerContrast mediaMagnetic resonance (MR)
The primary objective of any diagnostic imaging modality is to accurately define the presence, type and extent of disease in order to optimize patient management decisions and best plan therapeutic and surgical interventions. In women with suspected breast cancer the aim of diagnostic imaging is to detect and accurately diagnose malignant tumors and to facilitate the correct choice of therapy, be it mastectomy or breast-conserving surgery (e.g., lumpectomy) with or without pre-operative neoadjuvant chemotherapy. The choice between breast-conserving surgery and mastectomy depends on numerous factors including tumor size, location and grade, the ratio of tumor size to breast volume, multifocality or multicentricity of the tumor, and patient preference [1, 2]. Whereas mastectomy may be the only viable option for women with advanced or aggressive breast cancer, breast-conserving surgery is likely to be the preferred option for women with early breast cancer (stage I or II) given the equivalent survival times for patients undergoing the two approaches [3–11]. The clinical need therefore is to accurately differentiate those patients for whom breast-conserving surgery is a realistic option from those patients requiring mastectomy. A further need among women with the option of breast-conserving surgery is to identify those patients requiring lumpectomy from those requiring more extensive surgery (e.g., quadrantectomy).
Currently, conventional mammography and ultrasound are standard imaging techniques for detection and evaluation of breast disease and are the primary imaging modalities used in screening of women considered to be at high risk for breast cancer. However, these techniques are associated with limited sensitivity and specificity for the detection and diagnosis of breast lesions [1, 12–18], particularly in women with dense breast parenchyma [12, 13, 18]. Clearly, there is increased likelihood of inappropriate patient management if therapeutic decisions are made solely on the basis of mammography and ultrasound findings.
In recent years contrast-enhanced magnetic resonance mammography (CE-MRM) has emerged as the most sensitive imaging modality for the detection and diagnosis of breast lesions [1, 16–22]. Numerous studies have confirmed the superior diagnostic performance of CE-MRM compared to conventional mammography and ultrasound [16–18] and the greater potential of the method for effective cancer screening in women with high familial risk of breast cancer [16, 23, 24]. Studies to evaluate the impact of CE-MRM on patient management decisions have similarly revealed superiority compared to conventional mammography and ultrasound [1, 2, 25–27]. The present analysis was aimed at further evaluating the impact of CE-MRM on surgical decision-making relative to decisions taken solely on the basis of clinical examination, conventional mammography and ultrasound. At variance with previous studies, a gadolinium contrast agent (gadobenate dimeglumine, Gd-BOPTA, MultiHance®: Bracco Imaging SpA, Milan, Italy) with high relaxivity was used since studies have shown this agent to significantly improve the detection and characterization of breast lesions compared to conventional gadolinium contrast agents [28–31].
Materials and methods
Between October 2003 and February 2005 a total of 164 consecutive women (mean age: 53.4 ± 12.0 years; age range: 30–80 years) who underwent clinical examination, conventional bilateral mammography and whole breast ultrasound and who had one or more breast lesions detected by one or more of these procedures which was considered suspicious or highly suspicious (BI-RADS IV–V) for malignancy was evaluated. Based on findings from conventional mammography and whole breast ultrasound all 164 women subsequently underwent a preoperative CE-MRM examination. Histological evaluation of all detected lesions was performed on samples obtained from core biopsy or surgery.
All patients provided written informed consent to participate in the study. Women unable to undergo CE-MRM because of a pacemaker, aneurysm clip, or metallic foreign body in or near the eye, or who were otherwise contraindicated for MR imaging were excluded from the study. The study had institutional review board approval and was conducted according to the principals of Good Clinical Practice (GCP).
Contrast-enhanced magnetic resonance mammography
CE-MRM was performed on a 1.5 T magnet (Siemens Vision plus, Erlangen, Germany) using a bilateral breast surface coil with the patient in the prone position. An axial 3D dynamic T1-weighted gradient-echo sequence was employed with images acquired before contrast agent administration (pre-contrast unenhanced images) and at 0, 2, 4, 6, and 8 min after the administration of contrast agent (post-contrast enhanced images). The pre- and post-contrast 3D T1-weighted gradient-echo dynamic images were acquired with TR/TE = 8.1/4 ms; flip angle = 30°; 1 excitation; matrix = 396 × 512 pixels; rectangular field of view (FOV) of ≤36 cm; and slice thickness = 2 mm with no interslice gap. The total acquisition time for the 3D T1-weighted gradient-echo dynamic sequence was no more 120 s. No fat-suppression sequences were used; image subtraction (post-contrast–pre-contrast) was performed.
Post-contrast 3D T1-weighted gradient-echo dynamic images were acquired after administration of 0.1 mmol/kg bodyweight of gadobenate dimeglumine (Gd-BOPTA, MultiHance®; Bracco Imaging SpA, Milan, Italy) through an 18G needle-cannula positioned in an antecubital vein. Gadobenate dimeglumine was administered using an automatic injector (Spectris, Medrad, Indianola, PA, USA) at a rate of 2 ml/s and was followed by 10 ml of saline solution at the same rate. Data acquisition started after contrast agent injection, at the end of the saline flush.
Evaluation of images was performed in consensus by two observers (F.P. , S.P.) with 8 and 6 years’ experience, respectively, in interpretation of conventional mammography and breast ultrasound images and 6 and 5 years’ specific experience, respectively, in CE-MRM interpretation. Ultrasound images and conventional mammograms were reviewed at the time of the CE-MRM interpretation, and all clinical findings were available. Conventional mammograms and sonograms were evaluated for tumor detection and size. Where discrepant findings between these techniques were noted in terms of lesion size, the largest measurement was taken for subsequent analyses.
Mammograms, sonograms and histology were thereafter re-evaluated to identify patients theoretically eligible for conservative surgery compared with candidates for radical surgery, biopsy, or neoadjuvant chemotherapy. Mammograms were also reviewed to determine breast density according to the 4-point (1–4) scale of the Breast Imaging Reporting and Data System (BI-RADS) classification .
The potential impact of CE-MRM on surgical decision-making was thereafter evaluated for each patient. CE-MRM was considered to accurately suggest the appropriateness of breast conservation in cases in which the CE-MRM images clearly demonstrated the resectability of the lesion and in which CE-MRM was the only imaging modality able to do so. CE-MRM was considered to accurately suggest the necessity of a mastectomy in cases in which the CE-MRM images clearly showed more extensive disease than otherwise suspected from conventional mammography or ultrasound.
Evaluation of tumor extent
Approaches to treatment were based on the tumor classification (TNM) system in which tumors classified as PT1, PT2 and PT3 are defined as invasive tumors that are ≤2 cm, 2–5 cm and ≥5 cm, respectively, and PT4 is a defined as tumor infiltration of the chest wall and/or skin and an inflammatory carcinoma. In accordance with the TNM system, multifocal and multicentric tumors were classified according to the size of the largest single lesion. Two or more foci of cancer associated with one ductal network were defined as multifocal whereas two or more foci in separate lobes or segments (i.e., tumors in different breast quadrants) were classified as multicentric. All patients with unifocal carcinomas smaller than 3 cm underwent lumpectomy. Conversely, patients with multifocality underwent quadrantectomy. Exceptions to this approach were patients with very small breasts who underwent mastectomy even if the tumor was smaller than 3 cm in size. Most patients with a tumor larger than 3 cm and all patients with multicentric cancer underwent mastectomy. Usually, this approach to patient management was not influenced by the age of the patient, the location or grade of the tumor, the histologic findings, or the presence of immunohistologically-defined prognostic factors .
Tumor size on CE-MRM was determined directly from the computer workstation as the greatest dimension of macroscopic disease across the three orthogonal dimensions. Conventional mammography, ultrasound and CE-MRM images were considered concordant in terms of tumor extent if image measurements were within 50–150% of the macroscopic pathology measurement in the greatest dimension.
All women for whom histopathologic assessment revealed an invasive breast carcinoma were referred for axillary lymph node dissection. Likewise, patients with inflammatory carcinoma or with a tumor larger than 3 cm underwent neoadjuvant chemotherapy. If histopathologic evaluation revealed a benign lesion, a conservative approach to therapy was adopted.
Assessment of the impact of CE-MRM on surgical decision-making
Images were evaluated initially for accuracy in identifying malignant lesions relative to findings at histology. For this assessment, lesions detected and correctly classified as malignant were deemed true positive (TP) lesions whereas lesions detected but incorrectly classified as malignant were deemed false positive (FP) lesions. Lesions that were detected on mammography and/or ultrasound and classified as malignant but which were considered benign or non-existent on both initial and 18-month follow-up CE-MRM were deemed FP for mammography/ultrasound but true negative (TN) lesions for CE-MRM. Finally, lesions missed at initial mammography and/or ultrasound but detected on subsequent CE-MRM were deemed false negative (FN) lesions for mammography/ultrasound.
Due to the inclusion criteria for the study no patients were evaluated who were considered to definitely or possibly have non-malignant breast lesions (BI-RADS II or III) on the basis of clinical investigation, mammography or ultrasound.
Based on lesion detection and characterization relative to histology and patient follow-up, determinations were made of the sensitivity, accuracy, and positive predictive value (PPV) for the identification of malignant breast lesions on mammography/ultrasound and CE-MRM. Because the inclusion criteria considered only patients that were suspicious or highly suspicious for malignant breast cancer, no determinations were performed for specificity or negative predictive value (NPV).
A final assessment of the impact of CE-MRM on surgical decision-making was made based on the number of patients whose management was altered because of additional information available on CE-MRM relative to mammography and/or ultrasound.
Tumor detection on CE-MRM
Histopathology of 226 lesions in 164 patients
Number of lesions
Ductal carcinoma in situ
Lobular carcinoma in situ
Invasive ductal carcinoma
Invasive lobular carcinoma
Intraductal papilloma with DCIS
Impact of CE-MRM on the detection of breast lesions
Patients with lesions suspected on mammography and/or ultrasound
Final diagnosis based on mammography /ultrasound and CE-MRM
Patients with additional lesions detected only on CE-MRM
Number of additional lesions
Pectoral muscle infiltration
A review of the conventional mammograms of the 32 women with multifocal, multicentric or contralateral breast lesions detected on the basis of CE-MRM alone revealed either heterogeneously dense breast parenchyma (BI-RADS 3; 15 patients) or extremely dense breast parenchyma (BI-RADS 4; 17 patients). Similarly, the breast parenchyma of the 3 patients later shown not to have lesions was either BI-RADS 3 (2 patients) or BI-RADS 4 (1 patient).
Accuracy of imaging procedures for identification of malignant lesions
Of the 175 lesions detected on mammography/ultrasound and considered suspicious or highly suspicious for malignancy, 3 were subsequently shown on CE-MRM to be non-existent while 12 lesions (6 fibroadenomas, 2 cases of focal adenosis, 2 papillomas, 1 case of fibrocystic disease and 1 case of dysplasia) were subsequently determined to be non-malignant following histological examination. These 15 lesions were considered FP lesions for mammography/ultrasound. The remaining 160 lesions detected and correctly classified as malignant on mammography/ultrasound were considered TP lesions. Because of the inclusion criteria of the study, no lesion detected on mammography/ultrasound was considered a TN lesion.
Of the 226 lesions detected on CE-MRM, 15 (5 fibroadenomas, 4 cases of focal adenosis, 3 papillomas, 2 cases of fibrocystic disease, and 1 case of dysplasia) were determined to be non-malignant following histologic evaluation. These 15 lesions comprised 11 lesions (5 fibroadenomas, 2 cases of focal adenosis, 2 papillomas, 1 case of fibrocystic disease, and 1 case of dysplasia) detected and misclassified as malignant on mammography/ultrasound and 4/51 lesions detected only on CE-MRM (2 cases of focal adenosis, 1 papilloma, and 1 case of fibrocystic disease). These 15 lesions were considered FP lesions for CE-MRM. In each case the patients harboring FP lesions were classified as BI-RADS IV rather than BI-RADS V for level of suspicion of malignancy. The FP lesions on CE-MRM were noted in 12 (7.3%) patients overall; in 3/12 patients the lesions were in the ipsilateral breast while in the remaining 9/12 patients they were in the contralateral breast.
The 51 lesions missed on mammography/ultrasound compared to CE-MRM comprised 47 that were histologically confirmed to be malignant and 4 that were confirmed to be non-malignant. In terms of malignant lesion identification, the 47 confirmed malignant lesions were considered TP lesions for CE-MRM but FN lesions for mammography/ultrasound. Concerning the 4 non-malignant lesions detected on CE-MRM but not on mammography/ultrasound, these were included among the 15 FP lesions for CE-MRM. The values for sensitivity, accuracy and PPV of mammography/ultrasound for the identification of malignant lesions were therefore 77.3% (160/[160 + 47]), 72.1% (160/[160 + 47 + 15]), and 91.4% (160/[160 + 15]), respectively. In comparison, the availability of CE-MRM findings resulted in the detection and correct classification of 211/226 lesions as malignant (TP lesions). The remaining 15 lesions were FP lesions. No lesions detected on CE-MRM were incorrectly classified as non-malignant (FN). The values for sensitivity, accuracy and PPV of CE-MRM for the identification of malignant lesions were therefore 100% (211/[211 + 0]), 93.4% (/[211 + 15 + 0]), 93.4% (211/[211 + 15]), respectively.
Impact of CE-MRM on surgical decision-making
The improved diagnostic performance of CE-MRM resulted in a change to the proposed therapeutic procedure indicated by clinical evaluation, mammography and/or ultrasound for 32/164 (19.5%) patients (28/34 patients with additional lesions on CE-MRM and 4 patients for whom CE-MRM confirmed the absence of malignant lesions). The 7/12 patients with multifocal lesions seen on CE-MRM alone for whom lumpectomy had originally been planned because of a suspected solitary tumor based on mammography and/or ultrasound findings were subsequently referred for quadrantectomy (5 patients) or mastectomy (2 patients) on the basis of the additional CE-MRM findings. Similarly, mastectomy instead of lumpectomy or quadrantectomy was performed for 3 of the 4/8 patients with multicentric carcinoma detected on CE-MRM alone (Fig. 1). The therapeutic approach (mastectomy) to the remaining patient with multicentric carcinoma detected on CE-MRM alone was unchanged from that planned on the basis of mammography/ultrasound because the primary tumor was larger than 3 cm in size.
The remaining 22 patients for whom an altered approach to therapy was indicated on the basis of CE-MRM comprised 12 patients with malignant contralateral breast lesions detected on CE-MRM alone, 2 patients with pectoral muscle infiltration detected on CE-MRM alone, 4 patients with a larger tumor size detected on CE-MRM, 1 patient for whom CE-MRM correctly indicated a benign fibroadenoma, and 3 patients for whom CE-MRM (and subsequent 18–24 month follow-up CE-MRM) confirmed the absence of any breast lesion. The altered approach to management in the 18 patients with confirmed malignant lesions involved an additional surgical intervention (lumpectomy) in the 12 patients with malignant contralateral tumors, additional chemotherapy before surgery in the 2 patients with pectoral muscle infiltration, and quadrantectomy with (2 patients) or without (2 patients) associated chemotherapy in the 4 patients with a larger tumor size on CE-MRM (Fig. 3). Management of other patients with lesions determined to be larger on CE-MRM compared with mammography/ultrasound was unaltered apart from the need for wider excision margins. Management of the 3 patients with no lesions on initial or follow-up CE-MRM was altered from a proposed lumpectomy on the basis of mammography/ultrasound to no treatment. Finally, the 1 patient with benign fibroadenoma correctly diagnosed on CE-MRM (Fig. 4), as well as the 9 patients with additional contralateral breast lesions determined to be non-malignant (FP) following percutaneous biopsy, were all managed conservatively.
Accurate evaluation of tumor extent and examination of the whole breast is vital for optimal treatment planning of patients with confirmed breast cancer. Whereas the preferred option is invariably breast-conserving surgery, this approach to patient management requires careful preoperative assessment to ensure the full extent of disease is limited to a restricted breast volume. The principal contraindication to breast-conserving surgery is multicentric cancer for which mastectomy is the only realistic treatment option [1, 2, 10, 11, 26, 27, 33]. Whereas multifocal cancer sometimes offers more opportunity for partial breast conservation, the comparatively high frequency of local recurrence [2, 4, 25, 34] often necessitates wide excision margins which may also preclude effective breast conservation if the affected breast is small. The risk in inaccurate pre-operative evaluation of patients with breast cancer is an inappropriate surgical intervention.
Several studies have been performed to evaluate the preoperative use of CE-MRM as a supplemental examination to clinical examination, mammography and/or ultrasound to assess the extent of disease within the breast [1, 2, 17, 26, 27]. These studies have shown unequivocally that CE-MRM permits the detection of additional malignant foci not seen on mammography and that detection of these additional foci results in a change in patient management in 11–20% of patients. Our findings confirm the results of these previous studies in showing that CE-MRM permitted the detection of 51 additional lesions not seen on conventional mammography or ultrasound. Moreover, 3 of 4 further patients suspected of having breast cancer on mammography/ultrasound were subsequently shown not to have any lesions at all while the fourth patient was confirmed to have only a solitary benign fibroadenoma based on CE-MRM findings. In terms of the impact on patient management, our CE-MRM findings resulted in a change of diagnosis for 38/164 (23.2%) patients overall and an altered approach to patient management for 32/164 (19.5%) patients. The altered approach to management involved more extensive surgery in 28/164 (17.1%) patients because of additional lesions or a larger lesion size, and cancellation of a proposed lumpectomy in 4/164 (2.4%) patients.
Concerning the detection and identification of malignant lesions CE-MRM was markedly superior to mammography/ultrasound (sensitivity for detection: 100% vs. 77.3%, accuracy for malignant lesion identification: 93.4% vs. 72.1%). In large part this can be ascribed to the breast density of the patients concerned: conventional mammography and ultrasound are known to be limited in patients with dense breast parenchyma [13, 15, 22, 35]. The 32 women with multifocal, multicentric or contralateral breast lesions detected on CE-MRM alone in our study had either heterogeneously dense (15 patients) or extremely dense (17 patients) breast parenchyma. Although the positive predictive values were similar for mammography/ultrasound and CE-MRM (91.4% vs. 93.4%, respectively) little can be ascribed to these data since the inclusion criteria for the study required that patients be suspicious or highly suspicious for breast cancer. Similarly, because no patients were enrolled who were known or suspected to have only benign (TN) breast lesions, it was considered inappropriate and potentially misleading to determine values for specificity and NPV based on just four lesions classified as benign or non-existent on CE-MRM.
To note in particular is the very high sensitivity (100%) for the detection of malignant breast lesions on CE-MRM which was due entirely to the absence of FN lesions (i.e., malignant lesions erroneously classified as non-malignant). Previous studies have frequently reported FN lesions on CE-MRM, particularly in the case of DCIS [1, 36–40]. The absence of FN lesions in this study can be ascribed in part to the inclusion criteria of the study which prescribed the enrollment only of patients that were suspicious or highly suspicious for the presence of breast cancer, and in part to the fact that gadobenate dimeglumine was the contrast agent used in the study. This agent has recently been shown to facilitate the identification of malignant breast lesions with significantly higher sensitivity than gadopentetate dimeglumine when administered at an identical dose of 0.1 mmol/kg bodyweight . Moreover, the enhancement of lesions and reader confidence for the characterization of lesions was significantly greater with gadobenate dimeglumine . The greater r1 relaxivity of gadobenate dimeglumine translates into greater T1 relaxation rate enhancement and thus greater signal intensity enhancement on T1-weghted MR images. Although the greater signal intensity enhancement has been shown to result in a greater number of FP findings, particularly at a double dose of 0.2 mmol/kg bodyweight , it is also associated with a smaller number of FN lesions since even hypovascular lesions (e.g., poorly vascularized DCIS) may show some degree of enhancement. In this regard it is clearly preferable to biopsy a detected lesion that is subsequently confirmed to be benign than to miss a malignant breast lesion entirely.
A third reason for the absence of FN lesions in this study was the fact that CE-MRM images were evaluated in conjunction with findings from mammography, ultrasound and clinical examination. Thus, small, aspecific areas of enhancement on CE-MRM, which might otherwise be considered insignificant on the basis of CE-MRM alone, might be interpreted differently if mammography and/or ultrasound also raise suspicions.
Concerning the 15 FP lesions detected on CE-MRM in 12 (7.3%) patients in this study, this number was slightly higher than the rate of FP detection reported elsewhere . However, bearing in mind that 47 additional malignant (TP) lesions were detected in this study solely as a result of the CE-MRM examination and that 4 lesions considered malignant on mammography/ultrasound were correctly classified as non-malignant on non-existent (TN) on CE-MRM, this incidence of FP lesions should be considered acceptable, particularly since the evaluated population comprised patients that were highly suspicious for the presence of malignant lesions who were candidates for biopsy and/or surgical intervention anyway. As noted previously [28, 29] the predominating FP lesion in this study was fibroadenoma.
Although our results clearly support the view that CE-MRM is a valuable imaging modality for accurate patient management decisions, a possible limitation of the study is that only patients considered suspicious or highly suspicious (BI-RADS IV and V) for breast cancer were evaluated. In this regard, further prospective work would be of interest to determine the additional value of CE-MRM in a more diverse group of patients. On the other hand, at many centers a CE-MRM examination is generally only requested for patients presenting with a suspicious lesion on mammography and/or ultrasound; in this context our findings are perhaps an accurate reflection of the diagnostic and pre-operative utility of CE-MRM in the most relevant patient population.
In conclusion the present study shows that CE-MRM impacts extremely positively on patient management decisions in patients suspected of harboring malignant breast lesions based on clinical examination, mammography and/or ultrasound. Although further confirmatory work is needed in a larger patient cohort that includes also patients with known benign lesions, the results of this study in patients with known or suspected breast cancer emphasizes not only the need for careful pre-operative work-up of these patients but also the limitations of conventional breast imaging techniques.