The use of preoperative magnetic resonance imaging (MRI) as a diagnostic and treatment-guiding tool for the management of patients with early-stage breast cancer (ESBC) continues to increase.1 Currently, 65% of all breast cancer patients undergo preoperative MRI during the diagnostic workup.2 The high sensitivity of MRI makes it a useful tool for screening high-risk patients, locating occult cancers not seen with conventional imaging, and documenting treatment response.3 However, MRI has not been associated with a reduction in disease-specific mortality for patients with ESBC.4 For patients with low-risk disease, MRI is associated with an increased mastectomy rate among those who are otherwise lumpectomy-eligible.5,6,7,8

Although the detection of additional disease before surgery is helpful for surgical planning, the addition of MRI to the preoperative workup increases time until surgery, particularly for patients recommended to receive MRI-guided biopsies. Prior studies have found that receipt of MRI is associated with a 22.4-day delay in pre-treatment evaluation, and receipt of MRI biopsy is associated with nearly double the time until surgery in those not receiving neoadjuvant therapy.5,9,10

Bleicher et al.11 found that increases in time until surgery resulted in worse disease-specific and overall survival in early invasive breast cancer. Furthermore, receipt of MRI has not been associated with decreased rates of recurrence, re-operation, or positive margins.6,7,8,12

Therefore, the potential benefit of additional disease detection must be weighed against the prolonged time until definitive treatment for women with ESBC. By analyzing a contemporary cohort of breast cancer patients receiving MRI before surgery, we sought to determine whether those with historically accepted “high-risk” characteristics (young age, dense breasts, significant family history, and pure ductal carcinoma in situ [DCIS] histology) had higher rates of detection of additional disease seen only by MRI, with the hope of improving patient selection for the use of this resource.

Methods

Institutional review board approval was obtained before study initiation. A retrospective cohort study included patients with newly diagnosed, non-metastatic cTis-T2N0-N1 breast cancer treated from January 2016 to December 2018 (New York City) and April 2020 to October 2021 (South Florida). The study excluded those who received neoadjuvant therapy, had MRI before cancer diagnosis, or had bilateral breast cancer diagnosed before MRI (Fig. 1).

Fig. 1
figure 1

Inclusion criteria.

Full size image

Patients were recommended for breast MRI at the discretion of the radiologist and/or the medical or surgical oncologist. Breast MRI examinations were performed on either 1.5 Tesla (Avanto-Siemens Erlangen, Germany) or 3.0 Tesla (Discovery 750, GE Healthcare Chicago, IL, USA or Skyra, Siemens Erlangen, Germany) systems using a 16-channel breast coil (Sentinelle, Invivo Corp, Gainesville, FL, USA). All examinations were acquired with standard multiparametric protocol including axial T1 GRE, inversion recovery, and fat-suppressed dynamic before and after contrast imaging in both axial and sagittal planes. Post-processing was performed with special software (Multiview, Hologic Marlborough, MA and DynaCAD, Philips Andover, MA, USA). Maximum intensity projections, subtractions series, and three-dimensional (3D) deconstructions also were obtained. The images were interpreted by a group of fellowship-trained breast-dedicated radiologists with 2–20 years of experience.

Patient demographics, comorbidities, clinical presentation, and treatment received were collected through review of the electronic medical record. Study data were collected and managed using Research Electronic Data Capture (REDCap) electronic data capture tools hosted at the University of Miami. As a secure, web-based software platform, REDCap is designed to support data capture for research studies, providing an intuitive interface for validated data capture, audit trails for tracking data manipulation and export procedures, automated export procedures for seamless data downloads to common statistical packages, and procedures for data integration and interoperability with external sources.13,14

Demographic and disease characteristics were summarized using descriptive statistics for patients who received MRI, and their respective results were collected and categorized based on necessity of excision. In addition, we also used a multifactor scoring system (Sum3 [age <50 years, dense breasts, high-risk family history] and Sum4 [Sum3 and pure DCIS histology before MRI]) to determine the effect from summation of multiple factors (Tables 1 and 4). A “positive” biopsy result was defined and analyzed in two distinct ways: (1) the rate of biopsies yielding additional insitu or invasive cancer was defined as PositiveCancer, and (2) the rate of biopsies yielding histology often excised (and therefore thought to change the surgical plan) was defined as PositiveSurg, which included cancer, lobular carcinoma in situ (LCIS), atypical ductal hyperplasia (ADH), and atypical lobular hyperplasia (ALH). Benign histology not routinely excised was considered a negative result. Using these definitions, positive predictive value (PPV) was determined using the formula:

$${\text{PPV}} = {\text{TP}}/{\text{TP}} + {\text{FP}}\left( {{\text{TP}} = {\text{true}}\;{\text{positive}},{\text{either}}\;{\text{Positive}}_{{{\text{Cancer}}}} \;{\text{or}}\;{\text{Positive}}_{{{\text{Surg}}}} ;\;{\text{FP}},\;{\text{false - positive}},\;{\text{benign}}\;{\text{result}}} \right)$$
Table 1 Patient and clinical characteristics of patients with and without MRI biopsy
Full size table

Univariate logistic regression determined the association of patient and disease factors with either PositiveCancer or PositiveSurg. Multivariate logistic regression determined whether predetermined risk factors (Sum3 and Sum4) influenced biopsy yield, with adjustment for other relevant covariates. Statistical significance was set at a threshold of p lower than 0.05. All analyses were conducted using SAS version 9.4 for Windows (SAS Institute Inc. Cary, NC, USA).

Results

Between January 2016 and October 2021, 1072 patients underwent surgery for breast cancer. Of these 1072 patients. 457 (43%) received preoperative MRI after cancer diagnosis and did not have neoadjuvant therapy, and 447 underwent preoperative MRI, met the criteria for inclusion, and were analyzed in the current study (Fig. 1).

Patient and clinical characteristics are shown in Table 1. Among the patients who received preoperative MRI, the mean age was 56.9 (± 10.7 years), and 31% were premenopausal. In terms of ethnicity, 71% of the patients were White, 17% were Black, and 37% identified as Hispanic. Of the 447 patients, 58% had heterogeneous or extremely dense breasts, 24% had pure DCIS histology, and 37% underwent mastectomy. Notably, receipt of mastectomy may not have reflected the original plan for surgery.

For 204 (46%) of the patients, MRI-guided biopsy was recommended, and 131 (29%) of the patients underwent 149 total biopsies (96 ipsilateral and 53 contralateral [18 bilateral]). Patients recommended to have MRI-guided biopsy were significantly more likely to have a high-risk family history, dense breasts, higher-grade DCIS (grade 2 or 3), and multifocal disease (Table 1). These patients also were more likely to undergo mastectomy (46% vs. 28%; p < 0.001). When the predetermined characteristics were combined, the patients recommended to undergo MRI biopsy were more likely to have a Sum3 of 2 or higher and a Sum4 of 3 or higher. The groups recommended and not recommended for MRI biopsy were similar with regard to age, menopausal status, race, ethnicity, body mass index (BMI), reason for presentation, histology, grade (invasive), receptors, clinical T stage, N stage, and AJCC stage (Table 1).

Similar to the patients recommended for MRI biopsy, those who underwent the recommended biopsies (excluding those who deferred) were more likely to have dense breasts, pure DCIS, grade 2 or 3 DCIS, higher clinical T stage, multifocal disease, higher Sum3 (≥2), and higher Sum4 (≥3) than the patients who did not receive MRI biopsy. The groups that did and did not receive MRI-guided biopsy were similar with regard to age, menopausal status, race, ethnicity, high-risk family history, and receptor status (Table 1).

Biopsy results are shown in Fig. 2. Those who underwent MRI biopsy had a PositiveCancer rate of 41% and a PositiveSurg rate of 49%. Among the ipsilateral MRI biopsies, the PositiveCancer rate was 54% (21 invasive cancer and 31 DCIS) and 8% were high-risk lesions (6 LCIS and 2 ADH), and combined with biopsies positive for cancer, resulted in a PositiveSurg rate of 63%. Among the contralateral MRI biopsies, the PositiveCancer rate was 17% (2 invasive cancer and 7 DCIS), and 7.5% were high-risk lesions (3 LCIS and 1 ADH), with a combined PositiveSurg rate of 25%. Notably, all the patients with contralateral biopsies resulting in cancer were older than 50 years.

Fig. 2
figure 2

Ipsilateral and contralateral biopsy results.

Full size image

The positive predictive values of ipsilateral and contralateral biopsies are shown in Table 2. Ipsilateral and contralateral values were calculated overall and by predetermined relevant covariates of age, menopausal status, family history, breast density, and histology. Overall, the PPVCancer of ipsilateral biopsies was 54%, and the PPVCancer of contralateral biopsies was 17%. The PPVSurg of ipsilateral biopsies was 63%, and the PPVSurg of contralateral biopsies was 25%. Ipsilateral results showed that younger age, premenopausal status, higher breast density, and invasive or mixed histology resulted in numerically higher PPV values for cancer, although the confidence intervals overlapped. Contralateral results showed that postmenopausal status, significant family history, and pure DCIS histology appeared to result in numerically higher PPV. Notably, no patients younger than 50 years had contralateral cancer diagnosed by MRI biopsy, and dense breasts were not associated with higher contralateral PPVCancer or PPVSurg.

Table 2 Positive predictive value of MRI-guided biopsies
Full size table

Uni- and multivariate analyses of ipsilateral MRI biopsy results are shown in Table 3. In the univariate analysis, the patients presenting with an abnormal physical exam were significantly more likely to have an ipsilateral MRI biopsy result of cancer (odds ratio [OR], 2.79; 95% confidence interval [CI] 1.09–7.19; p = 0.033). Positive biopsy results for either definition were highly associated with a diagnosis of multifocality/multicentricity (PositiveCancer [OR, 3.68; 95% CI 1.45–9.3; p = 0.006], PositiveSurg [OR, 4.23; 95% CI 1.54–11.66; p = 0.005]) and the surgical method of mastectomy (PositiveCancer [OR, 15.89; 95% CI 5.41–46.6; p < 0.01], PositiveSurg [OR, 16.38; 95% CI 4.78–56.13; p < 0.01]). Notably, the patients with the predetermined combined factors (Sum3 and Sum4) were not more likely to have an ipsilateral biopsy result of cancer or other high-risk pathology according to the univariate analysis (Table 3). In the multivariate analysis, only surgical method (OR, 13.85; 95% CI 3.82–50.23; p < 0.01) remained significantly associated with a positive ipsilateral biopsy. The patients with the predetermined combined factors (Sum3 and Sum4) also were not more likely to have an ipsilateral biopsy result of cancer or other high-risk pathology according to the multivariate analysis (Table 3).

Table 3 Uni- and multivariate analysis of ipsilateral biopsy results
Full size table

Uni- and multivariate analyses of contralateral MRI biopsy results are shown in Table 4. Both analyses were limited by the small sample of patients who had a contralateral biopsy (n = 53) yielding PositiveCancer (n = 9) and PositiveSurg (n = 13). As shown in the univariate analysis, the patients with pure DCIS histology on their original diagnostic biopsy were significantly more likely to have a contralateral biopsy result of cancer or high-risk surgical pathology (PositiveSurg) (DCIS: OR, 2.79; CI 1.09–7.19; p = 0.033). This association was only a trend when limited to PositiveCancer (DCIS: OR, 4.33; 95% CI 0.99–1.31; p = 0.051). Race, ethnicity, menopausal status, family history, breast density, presentation, grade, multifocality/multicentricity, stage, estrogen receptor (ER) status, and surgical method were not associated with a greater likelihood of a positive contralateral MRI biopsy result (PositiveCancer or PositiveSurg). There was a trend for younger patients to be less likely to have PositiveSurg on contralateral biopsy (age <50 years: OR, 0.06; 95% CI 0.00–1.20; p = 0.066). Our multivariate analysis, limited by sample size, suggested that young patients may be less likely to have PositiveSurg (age <50 years: OR, 0.02; 95% CI 0.00–0.84; p = 0.041). The patients with the predetermined combined factors (Sum3 and Sum4) were not more likely to have a contralateral biopsy result of cancer or other high-risk pathology (Table 4).

Table 4 Uni- and multivariate analysis of contralateral biopsy results
Full size table

The clinical and pathologic characteristics of the patients with a diagnosis of contralateral breast cancer (CBC) by MRI are shown in Table 5. Interestingly, all the patients with contralateral biopsies resulting in cancer were older than 50 years, and all but one of these patients had an ER-positive index tumor, which would have led to a recommendation of hormonal therapy serving as contralateral breast cancer risk reduction. Susnik et al.15 also reported that those found to have CBC via MRI were more likely to be older (p < 0.001) and have ER-positive tumors (p = 0.027).16 Among the nine patients with a diagnosis of CBC, seven were found to have DCIS, and two were found to have small invasive cancers in the contralateral breast. Whereas eight of the nine patients found to have contralateral breast cancer had ER+ index tumors, five of the five patients with known contralateral receptors also had ER+ tumors (Table 5).

Table 5 Clinical and pathologic characteristics of patients with contralateral breast cancer diagnosed by MRI
Full size table

The characteristics of the patients who deferred biopsies are shown in Table 6. For 52 of the patients who deferred biopsies recommended from the MRI, the lesions were not originally seen on conventional diagnostic imaging. The recommended biopsies included 30 non-MRI biopsies only and 22 ipsilateral (n = 17), contralateral (n = 4), or bilateral (n = 1) MRI biopsies. Of 18 patients, 15 (83%) recommended for an ipsilateral MRI biopsy underwent mastectomy, and 11 (73%) of those were older than 50 years. Four (80%) of the five patients recommended for contralateral MRI biopsy underwent bilateral mastectomy, and none had cancer identified on the contralateral risk-reducing mastectomy specimen (Table 6).

Table 6 Characteristics of patients who deferred biopsies
Full size table

Discussion

The purpose of this novel PPV analysis was to inform the creation of a “checklist” of patient or disease characteristics for better selection of patients with ESBC to receive preoperative MRI. Prior guidelines have recommended ordering MRI for patients with lobular carcinoma, aggressive biology, dense breasts, or younger age.17 We did not find that patients with the historically accepted risk factors (young age, dense breasts, significant family history, pure DCIS histology), either alone or combined, had a higher likelihood of yielding additional disease for ipsilateral or contralateral biopsies.

Our study was consistent with a prior larger study of more than 19,000 women, which showed that women with dense breasts had higher biopsy rates without concurrent greater likelihood of cancer detection.16 Even among patients with ILC, radiologically denser breasts has not been shown to lead to an increase in additional findings on MRI in previous work.18 However, the results of our study differed from those of a larger study by Wecsler et al.,19 which showed that higher mammographic density (heterogeneously or extremely dense) and ILC histology were associated with additional biopsy-proven cancer on MRI. Our study did find that MRI is likely to be helpful in detecting additional ipsilateral cancer or high-risk lesions in patients presenting with an abnormal physical exam. This may be related to the greater stage at which these patients presented, which lends itself to an increased risk of multicentric disease.

Our dual-institution analysis showed that more than half of the ipsilateral MRI biopsies exhibited additional cancer, and that fewer than one in five contralateral MRI biopsies showed additional cancer. Of the 447 patients with ESBC who had preoperative MRI, 61 (13.6%) had a PositiveCancer biopsy result, which is lower than the 19% (38 of 199) rate of additional cancer detection previously reported.20

Our study found that only 2% of the patients (9 of 447) evaluated with a preoperative MRI had a CBC not seen on traditional imaging identified, consistent with prior studies reporting a CBC rate of approximately 3%.15,21 However, another study with a larger sample than ours found that the PPV of contralateral lesions detected by breast MRI was 48.8%, about twice our PPV of 24.5%.22 Although our contralateral biopsy analysis was limited by its small sample of contralateral MRI biopsies showing cancer, our preliminary data did trend towards older patients (mean age, 66 years) beinge more likely to have positive contralateral MRI biopsies, and therefore while younger patients were less likely to have PositiveSurg on contralateral biopsy (age <50 years: OR, 0.06; 95% CI 0.00–1.20; p = 0.066).

Compared with other studies of preoperative MRI, our study had several strengths. The novel analysis adds to the growing body of literature arguing against the reflexively ordered MRI for historical accepted patient characteristics, specifically age, used as a surrogate for menopausal status in our study. This evidence is critical to the diagnostic planning for premenopausal patients, who tend to have more biologically aggressive disease and therefore would theoretically be more susceptible to changes in oncologic outcome related to delay.

Additional strengths of this study included the reading of all MRIs by fellowship-trained radiologists who specialize in breast imaging. Furthermore, our study encompassed a racially and socioeconomically diverse patient population from two large tertiary care centers in New York and South Florida, making our findings more generalizable than those of a single-institution study. Finally, the granular data regarding the 9 (2%) patients with a diagnosis of CBC and the 52 (12%) patients who deferred recommended biopsies, many of whom underwent bilateral mastectomy, provide hypothesis-generating information for future analyses.

The retrospective nature of our study created potential for selection bias because we were unable to control for physician or patient preference when allocating patients. For example, breast surgeons and general surgeons are more likely to refer patients with certain risk factors, such as a BRCA mutation, familial or personal breast cancer history, extremely dense breasts, age younger than 40 years, multifocal or multicentric disease, and triple-negative breast cancer, to receive an MRI.23 Additionally, the inclusion of patients from two institutions may have led to slight variability in how images were interpreted by different radiology faculty.

The patients who met the eligibility criteria constituted a reasonably sized sample of patients who underwent MRI-guided biopsy (n = 131). However conclusions with respect to factors associated with MRI-detected CBC are limited by the low number of contralateral MRI biopsies showing cancer (n = 9).

Although cancers of lobular histology may be less easily characterized by traditional breast imaging, which leads to a relatively high referral rate of lobular carcinoma patients to undergo MRI, our cohort included only 11 patients who had ILC, which therefore could not be analyzed as a separate variable in the multivariate analysis. In addition, our analysis did not delineate the impact of in situ versus invasive cancer or ductal versus lobular cancer, which may be helpful variables for future analyses.

In our study, 243 patients (54%) were not recommended for additional biopsies after preoperative MRI, and 62 patients (14%) underwent MRI-guided biopsies with benign results. Therefore, MRI did not provide practice-changing information and possibly did not change the surgical management for these patients, which led to potential delays to surgery. It is therefore important to weigh the risks and benefits of additional imaging and its impact on timely oncologic care. The use of MRI continues to increase without a demonstrable oncologic benefit to patients, and future analyses are necessary to assist with patient selection.4,24 The Alliance (A011104) trial will report locoregional recurrence rates with and without the use of preoperative MRI, and we anticipate that these results will further guide decision-making.

Conclusion

The selection of patients with ESBC for preoperative breast MRI has yet to be standardized or optimized across institutions. The potential disadvantages of MRI should be considered for “low-risk” patients, especially if it is not clear that the detection of additional low-risk disease is critical in the setting of planned adjuvant systemic therapy (7 of 9 CBCs diagnosed in our cohort were DCIS). When viewed in the context of additional disease yield, our analysis did not support the routine use of MRI for young patients with dense breasts and pure DCIS histology. However, MRI still is a useful tool in determining whether patients are candidates for lumpectomy when their physical exam is discordant with their imaging, and in determining whether patients are eligible for neoadjuvant therapy when they have borderline T2 or T3 tumors. As we await the results of the prospective randomized Alliance trial (A011104), further investigations regarding MRI use in ESBC patients should use registries with large sample sizes to account for low contralateral biopsy yield and focus on identifying patients most likely to benefit from the addition of MRI to the preoperative workup.