MRI Evaluation of Residual Breast Cancer After Neoadjuvant Chemotherapy: Influence of Patient, Tumor and Chemotherapy Characteristics on the Correlation with Pathological Response

Discussion

The role of MRI in cases of NAC is crucial. During NAC, it allows evaluation of the therapeutic response. Thus, monitoring of the response allows the type of chemotherapy to be changed if the tumor does not shrink, or the chemotherapy to be interrupted if the tumor progresses. At the end of the treatment, the therapeutic response can be described as complete or partial. This response is an important prognostic factor as a complete response is associated with disease-free and overall survival benefit (14, 15). MRI also helps to choose between conservative and radical surgery. Thus, the surgical strategy is often selected according to the MRI data: radical surgery in cases of fragmentation or in the absence of shrinkage of the tumor; conservative surgery in cases of disappearance or concentric regression response on all the axes of the tumor. Thus, the NAC has allowed to increase the proportion of conservative treatment without reducing the survival of the patient (16).

In our population, the overall correlation was poor, with a Pearson correlation factor of 0.52 (p<0.001). This was lower than the correlation found by Segara et al. (r=0.749), with their study of 69 patients, but closer to the correlation found by Moon et al. (r=0.58), in their study of 195 women (6,8). These differences might be attributed to the difference in measurement of residual tumor. Segara and Moon chose to measure only the residual invasive carcinoma. We chose to measure both invasive and in situ carcinoma.

There was a time when patients with a complete clinical response after NAC were exclusively proposed radiotherapy (17). As local recurrences were more frequent among these women, this strategy was abandoned. These local recurrences were related to many false-negative results of clinical examination. Since then, as the correlation between MRI/pathologic response was found to be stronger than the correlation between clinical examination and pathological response, Ring et al. suggested that after a complete response on the MRI, exclusive radiotherapy was possible (17). Thus, with the perspective of trying to avoid surgery for women with a complete response at MRI, the goal of MRI is to evaluate whether or not surgery is necessary after NAC. As DCIS requires surgery, it is crucial that the DCIS component is taken into account. This is why we chose to measure both invasive and in situ carcinoma on the pathological results.

While the correlation for the global population is low, it is better for women over 50 years (r=0.64; p<0.001), postmenopausal women (r=0.61; p<0.001), those with triple-negative tumors (r=0.69; p=0.0002), and for patients who did not receive anthracycline (r=0.70; p=0.08). Special precautions should also be taken when interpreting the MRI of tumors with a DCIS component (r=0.18; p=0.42). Moon et al. found a better correlation for women aged over 45 years (8). This could be explained by a change in the breast architecture at the time of menopause. As shown by Kollias et al., histological type of tumor varies depending on the age of the patient (18). However, in our population, the histological tumor type did not vary significantly with the age of the women. Thus, in our study, the difference of correlation according to the age cannot be attributed to the histologic variations associated with age.

On the HER2 status and accuracy of the MRI, authors do not agree. For Chen et al. in their study of 51 women, MRI was more efficient among those with HER2-positive tumors (9). On the contrary Moon et al. with their study on 195 cases found MRI was less efficient for HER2-positive tumors. In our population, correlation did not significantly vary according the HER2 status (8) As for Moon et al., we found a stronger correlation for triple-negative tumors (0.69 and 0.78) in our population (8).

For tumors with a DCIS component, the authors are unanimous in stating that MRI is less sensitive in detecting DCIS than in detecting invasive carcinoma. For Faverly et al., the detection rate of DCIS was related to the grade of the tumor: high-grade DCIS was detected by MRI in 92% of cases against only 53% of low-grade DCIS (19).

This study was performed at a Center with much experience of breast carcinology and few studies on the correlation between MRI and pathological results concern that many patients (5). Concerning the methodology chosen, we opted for a correlation study and not one with linear regression. Indeed, the use of linear regression presupposes causality between the two factors and this did not seem appropriate for our study. One limitation of this study is its retrospective design with all its inherent bias.

While all patients in our study underwent an MRI, 6% of patients still underwent surgery twice with necessity for radical surgery after conservative procedures. This is not linked, we believe, to the accuracy of the MRI but the difficulty in obtaining tumor-free margins in cases of non-palpable tumors. In our population, 30 patients had a pCR and only 11 were diagnosed on MRI. This shows the high specificity but low sensitivity of MRI in predicting a pCR. While ultrasonography and mammography underestimate the size of the residual tumor, MRI seems to overestimate the size of the residual tumor. For 19 women, the MRI concluded there was residual tumor, while in fact the pathological response was complete. Studies have evaluated the performance of ¹⁸F-fluorodeoxyglucose positron-emission scanning in the evaluation of tumor size after NAC. The accuracy seems worse than that for the MRI due to a high rate of false-negatives (20). Thus, it cannot be recommended for routine practice.