Breast magnetic resonance imaging for surveillance of women with a personal history of breast cancer: outcomes stratified by interval between definitive surgery and surveillance MR imaging

Background Women with a personal history of breast cancer are at increased risk of future breast cancer events, and may benefit from supplemental screening methods that could enhance early detection of subclinical disease. However, current literature on breast magnetic resonance (MR) imaging surveillance is limited. We investigated outcomes of surveillance breast magnetic resonance (MR) imaging in women with a personal history of breast cancer. Methods We reviewed 1053 consecutive breast MR examinations that were performed for surveillance in 1044 women (median age, 53 years; range, 20–85 years) previously treated for breast cancer between August 2014 and February 2016. All patients had previously received supplemental surveillance with ultrasound. Cancer detection rate (CDR), abnormal interpretation rate and characteristics of MR-detected cancers were assessed, including extramammary cancers. We also calculated the PPV1, PPV3, sensitivity and specificity for MR-detected intramammary lesions. Performance statistics were stratified by interval following initial surgery. Results The CDR for MR-detected cancers was 6.7 per 1000 examinations (7 of 1053) and was 3.8 per 1000 examinations (4 of 1053) for intramammary cancers. The overall abnormal interpretation rate was 8.0%, and the abnormal interpretation rate for intramammary lesions was 7.2%. The PPV1, PPV3, sensitivity and specificity for intramammary lesions was 5.3% (4 of 76), 15.8% (3 of 19), 75.0% (3 of 4) and 98.3% (1031 of 1049), respectively. For MR examinations performed ≤36 months after surgery, the overall CDR was 1.4 per 1000 examinations. For MR examinations performed > 36 months after surgery, the overall CDR was 17.4 per 1000 examinations. Conclusions Surveillance breast MR imaging may be considered in women with a history of breast cancer, considering the low abnormal interpretation rate and its high specificity. However, the cancer detection rate was low and implementation may be more effective after more than 3 years after surgery.


Background
Although women previously treated for breast cancer are at a statistically significant increased risk of future breast cancer events [1,2], annual mammographic screening is currently the only post-treatment imaging modality recommended for breast cancer follow-up by the American Society for Clinical Oncology (ASCO) and the National Comprehensive Cancer Network (NCCN) [3,4].
There has been limited information on breast magnetic resonance (MR) imaging surveillance in this specific patient population. Previous screening trials using MR imaging have focused on high-risk women without a personal history of breast cancer, resulting in increased invasive breast cancer yields at acceptable recall rates and positive predictive values (PPV) of biopsy [5][6][7][8]. Surveillance breast MR imaging may also have potential benefits in women previously treated for breast cancer, primarily by overcoming the decreased sensitivity of mammography in breasts with dense tissue and treatment-related changes [9][10][11]. Despite advances in locoregional and systemic therapy, recurrence or second breast cancer rates are approximately 3% to 5% per year even in early-stage hormone receptor-positive patients [2,12]. Therefore, this patient group would benefit from supplemental screening methods that could enhance early detection of subclinical disease and ultimately improve relative survival [13,14]. Previous studies have reported that breast MR imaging depicted additional cancers even after prior or concurrent negative findings of mammography and ultrasound (US) [11,15,16]. However, due to sparse data on surveillance breast MR imaging, the appropriate interval following surgery for initiation of MRI surveillance has not yet been investigated.
At our institution, surveillance breast MR imaging has recently been implemented as part of the routine posttreatment surveillance protocol for patients previously treated for breast cancer. These patients had previously undergone routine supplemental surveillance with US. As a result, we were able to obtain data from a large group of patients regarding its performance stratified by interval between definitive surgery and implementation of MRI surveillance. The purpose of this study was to investigate the outcomes of surveillance breast MR imaging in women with a personal history of breast cancer.

Study population
This retrospective study was approved by the institutional review board of Yonsei University College of Medicine and the requirement for informed consent was waived. Between August 2014 and February 2016, 1285 breast MR examinations were performed in 1266 women who had been previously treated for breast cancer, either with breast conserving surgery (BCS) (n = 648) or mastectomy (n = 396). Among them, 222 women were excluded either because they underwent MR imaging for reasons other than postoperative surveillance (n = 70); they had BRCA genetic mutations (n = 20); they had undergone screening breast MR imaging prior to the study period (n = 4); or their 12-month imaging followup information was unavailable (n = 128) (Fig. 1)

Post-treatment surveillance
After definitive breast cancer surgery, patients underwent follow-up by clinical examination and breast US every 6 months and with mammography, chest radiography, abdominal US and whole body bone scan every 12 months. After 5 years following initial surgery, patients underwent annual follow-up by breast US and mammography. Breast MR imaging was implemented as part of the routine posttreatment surveillance protocol in 2013, and thereafter patients underwent screening breast MR imaging instead of US at approximately two and five years after surgery. Surveillance breast MR imaging was also performed at the request of clinicians or patients.
The median interval between prior surveillance US and MR examinations was 6.1 months (range, 0-13.9 months). In 19 cases (1.8%), surveillance US and MR imaging were performed on the same day at the Fig. 1 Flowchart of study population selection request of the referring physician. The median interval between prior mammography and MR examinations was 11.5 months (range, 0-65.1 months).

MR imaging technique
Breast MR examinations were performed using two 3 − Tesla MR scanners (Discover 750, GE Medical Systems, Milwaukee, WI, USA; Ingenia, Philips Medical Systems, Best, The Netherlands). Imaging was performed with a dedicated phased array breast coil (8-channel GE or 16channel Philips) with the patient in the prone position. Imaging was performed prior to a rapid bolus injection of contrast agent and six times after injection. Sequences included a three-plane localizing sequence, axial T2weighted fast-spin-echo and T2-stimulated inversion recovery (STIR) sequence, and axial T1-weighted nonfat-suppressed or fat-suppressed sequence before contrast administration. The bolus injection consisted of 0.2 mmol/kg body weight of gadolinium-based contrast agent (Dotarem, Guerbet, Paris, France; Magnevist, Berlex Laboratories, Wayne, NJ, USA; or Gadovist, Bayer Scherming Pharma, AG, Berlin, Germancy) and a 20-mL saline flush delivered at a rate of 2 ml/s. 3D dynamic post-contrast enhanced (DCE) axial images are then performed in the axial plane and a T1-weighted 3D delayed postcontrast sequence is acquired in the sagittal plane. Bilateral examinations were performed for all patients.

MR imaging evaluation
MR images were prospectively interpreted by one of four radiologists with 6-15 years of experience in breast MR imaging interpretation. Computer aided evaluation software (CADstream, Confirma, Kirkland, WA) was used for characterization of lesion kinetics. Of the 1044 women, 89.1% (930 of 1044) previously underwent preoperative MR imaging. Each MR examination was given a Breast Imaging Reporting and Data System (BI-RADS) final assessment category based on the breast lesion morphology and kinetics.
We retrospectively reviewed MR imaging reports, clinical and imaging records. For lesions assessed as BI-RADS 3, follow-up breast MRI or US at 6-12 months was recommended based on the presence of a US-correlate on previous imaging and the period of stability. BI-RADS category 3 was also given to newly found lesions if its findings were probably benign according to the BI-RADS MR lexicon [17]. For breast or chest wall lesions that were assessed as BI-RADS category 4 or 5, targeted US was first performed and US-guided biopsy or MR-guided biopsy was performed accordingly. For suspicious extramammary findings found on breast MR imaging, further evaluation with other imaging modalities was performed with subsequent biopsy when needed.

Statistical analysis
Intramammary cancer was defined as cancer in the ipsilateral breast following BCS or cancer in the contralateral breast. Extramammary cancer was defined as locoregional disease (cancer in the ipsilateral axilla, internal mammary or supraclavicular lymph nodes or in the mastectomy bed) and distant metastasis.
We calculated the overall cancer detection rate and abnormal interpretation rates. The overall cancer detection rate for MRI was defined as the total number of intramammary and extramammary cancers detected at MR imaging per 1000 examinations. The overall abnormal interpretation rate for MRI was defined as the percentage of MR examinations that were given BI-RADS categories 0, 3, 4, 5 or those with findings suspicious for extramammary cancer detected at MR imaging. According to the outcome monitoring section of the BI-RADS 5th edition atlas, we included BI-RADS category 3 in the numerator of the abnormal interpretation rate because further imaging is recommended before the next routine screening [18].
We also calculated outcome measures for MRdetected intramammary lesions. The cancer detection rate for intramammary lesions was defined as the total number of intramammary cancers detected at MR imaging per 1000 examinations. The abnormal interpretation rate for intramammary lesions was defined as the percentage of MR examinations that were given BI-RADS categories 0, 3, 4, or 5. Positive MR examinations were defined as those given BI-RADS categories 4 or 5. Negative MR examinations were defined as those assessed as BI-RADS categories 1, 2 or 3. MR examinations with an initial BI-RADS category 0 assessment were reclassified according to their final assessment. PPV 1 was defined as the percentage of MR examinations with BI-RADS categories 0, 3, 4, 5 that resulted in a tissue diagnosis of cancer. PPV 3 was defined as the percentage of all known breast biopsies performed as a result of positive MR examinations that resulted in a tissue diagnosis of cancer. A true-positive (TP) result was defined as a positive MR examination resulting in a diagnosis of cancer within 1 year. A true-negative (TN) result was defined as a negative MR examination and no detection of cancer within 1 year. A false-negative (FN) result was defined as a negative MR examination with a diagnosis of cancer within 1 year. A false-positive (FP) result was defined as a positive MR examination with no detection of cancer within 1 year.
In addition, we stratified the above performance statistics according to the interval between initial surgery and surveillance MR imaging: (1) for MR examinations performed at or less than a 36-month interval and (2) for examinations performed at more than a 36-month interval following initial surgery. Performance statistics were compared between the two groups using the Fisher exact test. We also compared the intervals between prior surveillance US and MR examinations by using the Student t test. Statistical analyses were performed by using statistical software (SPSS version 23.0; IBM Corp, Armonk, NY.). Table 1

Cancer detection yield for MRI
The overall abnormal interpretation rate for MRI was 8.0% (84 of 1053) and biopsy or further imaging was recommended for 29 examinations (2.7%) with 21 of the 29 exams classified as BI-RADS category 4 and the other 8 exams demonstrating extramammary lesions suspicious for malignancy (Table 2). Of the 21 BI-RADS category 4 lesions, 18 lesions underwent image-guided biopsy (USguided biopsy [n = 16] or MR-guided biopsy [n = 2]) and one lesion underwent surgical excision for a US correlate. Among them, 3 lesions were diagnosed as cancer. All three detected cancers were newly developed contralateral breast cancer, with one cancer detected at a second-round MR examination. Of the 54 lesions that were BI-RADS category 3, one cancer was diagnosed (1.8%). This lesion was an 8-mm enhancing mass at the contralateral breast and moderate background parenchymal enhancement on preoperative MR imaging performed 5 years ago made accurate comparison difficult. Because it was considered to have slightly increased in size, ultrasound correlation was recommended and the final assessment was upgraded to BI-RADS category 4 at US. Subsequent US-guided biopsy yielded invasive ductal carcinoma ( Table 3). All of the four MR-detected intramammary cancers were not detected on prior surveillance US which was performed at a median interval of 5.5 months (range, 4.6-12.4 months).
Among the 8 examinations with suspicious extramammary findings, five were finally considered negative based on image-guided biopsy (n = 3) or further imaging evaluation (PET-CT, whole body bone scan) (n = 2) with no evidence of malignancy for more than 1 year. Of the three examinations with extramammary cancer, two were histologically confirmed by US-guided biopsy (n = 1, chest wall) or surgical excision (n = 1, mediastinal LN). The remaining one patient was diagnosed with sternum metastasis based on imaging alone, which was initially detected on breast MRI and subsequently confirmed by whole body bone scan and PET-CT (Table 4). Therefore, the overall cancer detection rate for MRI was 6.7 per 1000 examinations (7 of 1053).

Cancer detection yield according to interval between initial surgery and MRI
The overall cancer detection rate for MRI was significantly greater in MR examinations performed with more than a 36-month interval following initial surgery than those performed at or less than a 36-month interval (17.4 per 1000 examinations vs. 1.4 per 1000 examinations, p = 0.006). None of the other performance statistics showed a significant difference between the two groups ( Table 2).
The mean interval between prior US and MR examinations was slightly greater in MR examinations performed ≤36 months than those performed > 36 months following initial surgery (6.3 ± 1.0 months vs. 5.9 months ±1.6 months, p < 0.001), but with a mean difference of 0.4 months.

Discussion
With recognition of the increased future breast cancer risk in patients with a personal history of treated breast cancer and the decreased sensitivity of mammography in dense breasts, several studies have recently investigated the performance of surveillance breast MRI examinations [15,[19][20][21][22][23]. Although the patient population and study design differ somewhat between studies, the reported cancer detection rates range from 10.0 to 18.1 per 1000 examinations [15,19,20,22,23], which are higher than the overall cancer detection rate in our study. One possible explanation is that 37.9% of our study population underwent mastectomy, whereas the majority of patients underwent breast conservation surgery in most studies [11,15,22]. Another possible explanation is that the majority (88.1%) of our study population had previously    [19,22]. In addition, our study population had received routine supplemental surveillance US, with a median interval of 6.1 months between prior surveillance US and MR imaging. All of the four MR-detected intramammary cancers in our study were not detected by previous surveillance US performed prior to MR imaging. Therefore, the MR-detected cancers in our study are more likely to represent truly newly developed cancers after treatment of initial breast cancer, which may be difficult to detect with surveillance US. The fact that the cancer detection rate for MR imaging performed at more than 3 years after surgery (17.4 per 1000) was greater than that for examinations performed within 3 years (1.41 per 1000) may provide a basis for establishing guidelines regarding timing of surveillance MR imaging initiation following definitive breast cancer surgery. Our overall cancer detection rate was similar to the incidence cancer detection rate of screening breast MR imaging in average risk women in a recent study (7.5 per 1000 examinations, 13 of 1741) [24]. In another study on women with a history of breast conservation therapy, of whom 91.8% underwent preoperative MR imaging and all had undergone supplemental surveillance US, a more than 24-month interval between initial surgery and MR imaging was an independent factor associated with MR-detected cancers [15]. Similar results have been reported for breast MR imaging screening of women with average risk of breast cancer, with no screening-detected breast cancer diagnoses made until almost 3 years after a negative MR study [24]. This has important implications for the effective implementation of breast MR imaging as a surveillance modality in the future, as breast MR imaging early in the post-treatment surveillance period may have relatively low cancer yields − especially with the increased use of preoperative breast MR imaging.
The overall abnormal interpretation rate (8.0%) in our study was slightly lower than prior studies, which ranged from 10.7% to 19.3% [15,20,22,25]. Although the PPV 1 (5.3%) was slightly lower than that in previous MR imaging studies (approximately 9.4%) [15,20], it was still higher than mammographic screening benchmarks from 2004 to 2008 according to the Breast Cancer Surveillance Consortium (4.3%) [26]. In addition, the PPV 3 in our study for intramammary lesions, 15.8% (3 of 19), was higher than the lower range of reported PPV values of surveillance US in women with a personal history of breast cancer, which ranged from 9.4% to 52.6% [27][28][29][30]. Reported PPV values of surveillance breast MR imaging in women with a personal history of breast cancer have been consistently similar to or higher than that of surveillance US [27][28][29][30]. Furthermore, surveillance MR imaging detected three extramammary rcancers (0.3%, 3 of 1053), which accounted for 42.8% of MRdetected cancers. Therefore, breast MR imaging may be more advantageous compared to US as an adjunctive surveillance tool, considering its low abnormal interpretation rate and ability to detect extramammary malignancy.
Our study had several limitations. First, this was a retrospective study from a single institution. Although our institution recently implemented breast MRI imaging into our post-treatment surveillance protocol to be performed two and five years after surgery, MR imaging was also performed at the request of clinicians or patients and therefore, the intervals between surgery and MR imaging were variable. Second, patients underwent surveillance with mammography and US prior to MR imaging, which could have affected the true cancer yield of MRI. Third, the median interval between initial breast cancer surgery and first-time surveillance MR examination (30.1 months, range, 12.1-240.2 months) was relatively short.

Conclusions
Our data suggest that surveillance breast MR imaging may be considered in women with a history of breast cancer, considering the low abnormal interpretation rate and its high diagnostic performance. However, the cancer detection rate was low and implementation may be more effective after more than 3 years after surgery. Further research on the appropriate timing for surveillance breast MR imaging initiation is required, especially in patients who have undergone preoperative breast MR imaging and supplemental surveillance US.