Residual Disease after Neoadjuvant Therapy for Breast Cancer: Can MRI Help?

See also the article by Kim et al in this issue.

Introduction

MRI has established value for defining extent of disease in the breast and is particularly helpful when ductal carcinoma in situ (DCIS) or multifocal or multicentric disease is present. The accuracy of MRI for measuring the extent of disease is influenced by a number of factors, including parameters of the image acquisition and contrast material administration methods, the histologic subtype of the tumor (ductal vs lobular) and the tumor’s hormone receptor (HR) and human epidermal growth factor receptor 2 (HER2) receptor status. When MRI is performed after neoadjuvant chemotherapy (NAC), its accuracy is lower, but it remains effective. After NAC, accuracy can also depend on the initial (pretreatment) tumor morphology and pattern of shrinkage, in addition to the imaging parameters and tumor subtype characteristics. When available, pretreatment MRI can facilitate the assessment of response and residual disease extent by focusing the evaluation to the original tumor bed. However, with or without preceding NAC, it has not yet been established that presurgical MRI leads to improved surgical outcomes or reductions in recurrence rates (1).

A major benefit of neoadjuvant (vs adjuvant) systemic chemotherapy is greater possibility for breast conservation. Accurate presurgical assessment of residual disease extent at MRI can assist in making the decision for breast conservation versus mastectomy and inform the surgical approach. Toward that goal, many published studies have explored the accuracy of MRI for determining residual disease extent after NAC by comparing agreement between size measurements made at MRI and those made at postsurgical histopathologic examination. In two of the larger studies (2,3), agreement was higher in HR-negative/HER2-negative (triple-negative) and in HER2-positive subtypes, whereas it was lower in the HR-positive/HER2-negative subtype. A 2015 meta-analysis by Marinovich and colleagues (4) examined agreement between post-NAC MRI and pathologic tumor size in a pooled group of 300 patients. The performance of MRI was found to be generally superior to that of alternative tests (mammography, US, and clinical examination).

There is special interest in understanding the ability of MRI to confirm a pathologic complete response (pCR) in the breast prior to surgery. A pCR is the most accepted surrogate end point for survival after neoadjuvant treatment of breast cancer. The U.S. Food and Drug Administration (FDA) considers pCR in support of accelerated drug approval for patients with high-risk, early-stage breast cancer (5). Recent improvements in chemotherapy regimens and targeted therapies for breast cancer have led to growing interest in identifying patients prior to surgery who have had exceptional responses and who may be candidates for de-escalated treatment (ie, reduction in the duration of NAC). In some cases, this may include omission of surgery (6). An ongoing multicenter clinical trial is assessing the accuracy of radiologic complete response on the basis of negative findings at three imaging modalities: mammography, US, and MRI (7). The trial uses image-directed biopsy to the tumor bed for assessing accuracy, and patients then proceed to surgery per the standard of care. A separate trial is also underway to test the feasibility of omitting surgery for patients with HER2-positive or triple-negative breast cancer who are deemed to be exceptional responders to NAC (8). Patients in this trial do not undergo surgery but proceed directly to radiation therapy. They are subsequently followed for ipsilateral breast recurrences every 6 months for 5 years by using a physical examination and mammography. In this trial, MRI is used instead of mammography at the discretion of the treating physician.

The immediate goal of MRI after NAC is to provide information on the extent of residual disease in the breast, including both invasive and noninvasive components, to enable surgeons to obtain adequate surgical margins at initial surgery and avoid reexcision. There is also interest in understanding the ability of MRI to predict pCR, as defined for use as an intermediate end point for survival. As a surrogate end point, pCR is most commonly defined as no residual invasive disease in the breast or axilla. Stricter definitions require that neither residual invasive or noninvasive disease (ie, DCIS) remains. The 2014 FDA guidance allows either definition for pCR. Current MRI techniques cannot enable differentiation between invasive disease and DCIS, and mismatches can be expected if pCR is defined on the basis of invasive disease only. MRI also does not adequately evaluate the axilla; however, exclusion of axillary disease is required for a determination of pCR. MRI is not likely to be used alone to confirm pCR presurgically and may need to be used in combination with biopsy, particularly if omission of surgery is being considered.

Appearing in this issue of Radiology, the article by Kim et al (9) is timely in revisiting the question of accuracy in estimating residual disease extent in light of the many issues discussed above. Similar to other recent reports, Kim et al found improved agreement between MRI and histopathologic findings when residual disease size was measured on delayed postinjection images—beyond the 90–120-second delay time that is generally recommended to balance diagnostic sensitivity and specificity. Kim and colleagues found that in the post-NAC setting, MRI led to underestimation of the total tumor size (invasive and DCIS) at 90 seconds after injection and overestimation of the tumor size at 360 and 590 seconds, the two delay times tested in the study. Greater discrepancies were found for cancers of lobular versus ductal histologic nature and for HR-positive/HER2-negative cancers versus other subtypes.

The study by Kim et al has implications for the use of shortened or abbreviated breast MRI protocols that reduce the total postinjection imaging time. Results reported in this and other studies indicate that prolonged delay times result in better agreement with histopathologic estimates of total tumor size, and absence of late enhancement may be important for confirming pCR in the breast (10). Kim and colleagues found comparable agreement with histopathologic findings for the 360- and 590-second delay times, suggesting that there was no benefit to extending delay time beyond 360 seconds for measuring total tumor size in any breast cancer subtype. When tumor size was restricted to the invasive component only, no difference was found between the early and conventional delayed (360 seconds), early and late delayed (590 seconds), or the conventional and late delayed images. These results showed comparable results irrespective of delay, suggesting that for the task of predicting pCR, early postinjection timing may be sufficient and imaging times can be shortened.

The findings of Kim et al suggest that separate guidelines may be needed for performing MRI to evaluate residual disease in the post-NAC setting. This may be especially needed if abbreviated MRI protocols are adopted. Further investigation of the effects of delay times shorter than 360 seconds may shed additional light on the need for separate protocols for treated and untreated breast cancers. It would be interesting for the authors to further explore this question in their data, which included imaging at 180-second delay.

Imaging is likely to be the most effective method to assess response and guide treatment redirection noninvasively in the neoadjuvant setting. As more is learned about the biologic heterogeneity of breast cancer, MRI methods can be further refined for more effective use in personalized treatment strategies.