Subjects and Study Design
At our institution, the institutional review board reviewed this study and designated it as exempt under the Department of Health and Human Services regulations. This study is Health Insurance Portability and Accountability Act compliant.
A retrospective cohort study was performed to review the data from all breast MRIs performed at our institution in patients with newly diagnosed breast cancer for evaluation of extent of disease between July 1, 2016 and September 30, 2019. Inclusion criteria were limited to patients with a malignant index mass and mass-associated NME which was defined as the presence of NME extending directly from the mass margins. Patients treated with neoadjuvant chemotherapy were included if they had an MRI-guided biopsy of the mass-associated NME prior to starting treatment. Exclusion criteria were patients with a breast cancer without a dominant mass, no mass-associated NME, malignant index mass not visualized on preoperative MRI, and patients treated with neoadjuvant chemotherapy who did not undergo MRI-guided biopsy of the mass-associated NME prior to initiating treatment.
Current Procedural Terminology codes were used to identify breast MRIs performed during our study time period. Our institution’s Value Analytics team then extracted the corresponding patient medical record numbers and age at the time of examination from our institution’s electronic medical record, Epic (EPIC EMR system, Epic Systems Corporation, Verona, Wisconsin, USA, 2019).
Manual chart review was performed for each patient to record breast cancer characteristics including histological type, molecular subtype and initial treatment performed (lumpectomy, mastectomy, or neoadjuvant chemotherapy). Histologic types were recorded using standard classifications [16]. Benign lesions were defined as benign breast tissue or pseudoangiomatous stromal hyperplasia (PASH). High-risk lesions were defined as atypical ductal hyperplasia (ADH), atypical lobular hyperplasia (ALH), lobular carcinoma in situ (LCIS), radial scar, complex sclerosing lesion, or papillary lesion. Malignant outcomes included ductal carcinoma in situ (DCIS) or invasive carcinoma as both entities managed by surgical excision. Additional grouping classifications are detailed as follows. Micro-invasive DCIS and all subtypes of invasive ductal carcinoma (IDC) such as mucinous or papillary were classified as IDC in our study. Additional cancer features including hormone receptor status, human epidermal growth factor 2 (HER2) status, and Ki-67 level were also extracted from the medical record.
Breast MRI reports were dictated by 11 breast imaging radiologists with experience ranging from 2 to 35 years. Manual chart and imaging review was performed by one of the authors for each patient. This included description of the malignant index mass (size, shape, margins, internal enhancement, and kinetics), mass-associated NME (size, distribution, internal enhancement pattern, and kinetics), and total imaging span of both the malignant index mass and mass-associated NME. The largest size dimension of the malignant index mass, mass-associated NME, and total imaging span were also recorded.
Terminology recorded of mass and NME descriptors were utilized in the Breast Imaging Reporting and Data System (BI-RADS) Atlas 5th edition [17]. Background parenchymal enhancement (BPE) is a subjective classification of a patient’s normal enhancement pattern designed as minimal, mild, moderate or marked. A mass is defined as a 3D space occupying lesion with convex-outward contours. Shape of a mass can be either oval (elliptical or egg-shaped), round (spherical shaped) or irregular (neither round nor oval in shaped). Margins of the mass can be either circumscribed (smooth), irregular (uneven or jagged) or spiculated (lines radiating from the mass). Internal enhancement pattern of the mass can be either homogeneous (uniform), heterogenous (non-uniform), rim-enhancement (enhancement at the borders of the mass), or dark internal septations (dark, non-enhancing lines in the mass). NME is defined as enhancement that cannot be classified as a mass. Distribution of the NME can be focal (enhancement that occupies less than a breast quadrant), linear (enhancement that is in a line), segmental (enhancement that is cone-shaped with apex towards the nipple), regional (enhancement that occupies more than a breast quadrant), or diffuse (enhancement that is randomly distributed throughout the breast). Internal enhancement pattern of NME can be either homogeneous (uniform), heterogenous (non-uniform), clumped (cobblestone enhancement of varying shapes and sizes), or clustered ring (thin rings of enhancement clustered around the ducts). Kinetics of masses and NME can either be Type 1 (continued enhancement over time), Type 2 (enhancement does not change over time after initial uptake), or Type 3 (enhancement decreases by over 10% after its peak from the initial uptake).
The following characteristics were also recorded for each included patient based on retrospective review by one of the authors: the presence and size of calcifications on the diagnostic mammogram immediately prior to preoperative breast MRI that correlated with the mass-associated NME, total imaging span of both the malignant index mass and the calcifications on mammography, the presence and final histologic diagnosis of MRI-guided biopsy of the mass-associated NME, and the presence and final histologic diagnosis of stereotactic-guided biopsy of the calcifications that correlated with the mass-associated NME. The largest size dimension of the calcifications that correlated with the mass-associated NME was recorded. Whether the presence of mass-associated NME on breast MRI changed the initial surgical treatment plan for the patient was also recorded.
Pathology slides of all cases were retrospectively re-evaluated by a breast pathologist for histologic correlation of the mass-associated NME. For each case, the pathologist was provided the size, distribution, and location of the mass-associated NME from the breast MRI examination relative to the index malignant mass to determine if a pathologic correlate existed on the surgical specimen. Total span of disease was defined as the total size of both the malignant index mass and surrounding malignancy in the specimen. For cases without available pathology slides, histologic correlation was obtained from pathology report extraction by the breast pathologist instead (N = 12). These cases as well as the cases with neoadjuvant chemotherapy following MRI-guided biopsy of the mass-associated NME (N = 5) were excluded from span of disease analysis given no direct pathologist evaluation.
Breast MRI Technique
Breast MRI examinations were performed on a 1.5-T Discovery (GE Healthcare) scanner using a 16-channel dedicated breast coil (Sentinelle, In-Vivo), a 3-T Skyra (Siemens Healthineers) scanner using a 16-channel breast coil (Siemens Healthineers), or a 3-T Trio (Siemens Healthineers) scanner using a 16-channel breast coil (Sentinelle, InVivo) with patients in the prone position. Sequences obtained included scout images in 3-planes, an axial T1-weighted sequence without fat saturation, an axial T2-weighted sequence with fat saturation, an axial T1-weighted precontrast sequence with fat saturation, and multiple dynamic axial T1-weighted postcontrast sequences with fat saturation 30 seconds after the administration of 0.1 mmol/kg gadolinium-based contrast agent at a rate of 2 cc/sec followed by a 20 cc saline flush. The T1-weighted pre- and postcontrast sequences with fat saturation are obtained with an in-plane resolution of 0.7 x 0.7 mm and a slice thickness of 2 mm and 1 mm on the 1.5 T and 3.1 T MRI units, respectively. Diagnostic examinations included 3 postcontrast sequences acquired for 120 seconds each. Specific subtraction sequences are obtained by subtracting the precontrast T1-weighted fat-saturated sequence from the postcontrast T1-weighted fat saturated sequence for all 3 postcontrast sequences. Maximum-intensity projection in 360 degree rotation is created based on the first postcontrast T1-weighted fat-saturation subtraction sequence. Dynamic contrast-enhancement kinetics was evaluated using an external software program (DynaCAD, Philips Healthcare) that characterized dynamic enhancement patterns of each pixel above threshold as persistent, plateau, or washout. MultiHance (gadobenate dimeglumine, Bracco) was the gadolinium-based contrast agent used for all examinations before February 2018. After February 2018, the gadolinium-based contrast agent was replaced with Gadavist (gadobutrol, Bayer).