Neoadjuvant chemotherapy (NAC) is used increasingly for breast cancer treatment, with two main benefits: it offers the ability to monitor response to treatment, where pathologic complete response (pCR) is prognostic, and it can result in downstaging of tumor and breast conservation treatment (BCT) or eliminate the need for postmastectomy radiation in the setting of pCR.
1‐
6 Many factors influence the choice of surgical procedure after NAC: patient preference, tumor appearance, hormone receptor (HR) and Her2 expression status, and treatment response.
7,
8 Whereas the post-NAC MRI often is used to determine whether BCT is possible, investigators note that the pre-NAC MRI influences surgeons’ recommendations, regardless of tumor appearance after NAC.
9 Because clinicians and patients seek to avoid reexcision, it is important to understand the reliability of the postchemotherapy MRI.
10‐
13
Discussion
Overall, the results in this prospective cohort of patients confirm our previous findings that tumor morphology, captured by MRI phenotype, and tumor subtype affect rates of achieving clinically meaningful tumor reduction after NAC.
26 Patients with well-defined MRI phenotypes and those with Her2+ and Tneg tumors were more likely to have tumor shrinkage to ≤4 cm. We and others have found an improved correlation between post-NAC MRI and surgical pathology in tumors that are well-defined by imaging.
27 Together, these findings suggest that MRI phenotype may be used in conjunction with tumor subtype to set appropriate expectations before undergoing NAC.
MRI phenotype and tumor subtype likely reflect biological differences between tumors. Other phenotypic features are associated with different tumor subtypes. Basal-like breast tumors have distinct histologic and immunotypic properties, with characteristics, such as central scar, tumor necrosis, spindle cells, squamous metaplasia, high mitotic count, and high nuclear to cytoplasmic ratio.
28,
29 Most of these Tneg tumors have mass-like imaging patterns on MRI.
30,
31 Even on mammogram and ultrasound, breast cancer subtypes have imaging characteristics, with Tneg cancers more likely to have smooth margins.
32,
33
MRI phenotype and marker status influenced the likelihood of size discrepancies between imaging and pathology. Similarly, Chen et al.
9,
34 found that MRI was less accurate in tumors that present as non-mass enhancement on MRI, and in another study suggest that MRI can be used more successfully to plan BCT in Her2+ patients. Others also have reported that post-NAC MRI appears to be less accurate in ER+ tumors and most accurate in Tneg or Her2+ tumors and that pre-NAC tumor size and response also impact accuracy.
34‐
37 Although some have reported the highest accuracy of MRI for Her2− disease, not knowing HR status and differences in rates of traztuzumab use could potentially account for the disparate results.
38 Benign proliferative processes can enhance on MRI and are difficult to differentiate from low-grade, ER+ DCIS lesions. False-positive MRI enhancement may reflect a spectrum of change within high-risk tissue, possibly explaining why it is difficult to distinguish residual tumor size in ER-positive patients, especially with diffuse disease.
39,
40
We previously reported differences in response to NAC based on the five MRI phenotypes described. Patients with well-circumscribed masses had the greatest response to NAC.
20 In the current study, the majority of patients (81 %) achieved shrinkage to tumor size ≤4 cm, whereas we previously found only 47 % achieved enough shrinkage to be potentially eligible for BCT.
26 This was likely due to the addition of taxane, which doubles the pCR rate compared with doxorubicin alone.
41 For tumor subtypes, adjusting for pre-NAC tumor size did not change our results. For MRI phenotype, we found that size and phenotype were associated, because diffuse tumors will necessarily occupy a larger space. The larger size of these diffuse tumors could influence the ability to reach the threshold of ≤4 cm, but separating the contribution of phenotype from size is not possible in this study. The tumor response to NAC also can affect MRI accuracy, with good correlation between MRI and pathology noted in tumors with extreme responses (either complete or none), and worse correlation among partial responders.
42,
43
For the MRI phenotypes, the well-defined groups had higher BCT rates, but no difference was seen among tumor subtypes. Whereas patient choice was a factor in 22 % of cases that were potentially candidates for but did not receive BCT, other factors, such as physician recommendations, could play a role in surgical decisions. Knowing the accuracy of MRI could alter these recommendations. Interestingly, post-NAC MRI longest diameter showed a stronger association with surgical procedure than tumor size on surgical pathology or post-NAC mammographic longest diameter among the 175 patients assessed by all three methods (Wilcoxon rank-sum test,
p = 0.001, 0.17, and 0.02 respectively). There was no significant difference in local recurrence or recurrence free survival between subjects who received BCT and those who received mastectomy.
44
More than one third of patients had a size discrepancy ≥2 cm between the post-NAC MRI and surgical pathology. Some have suggested that overestimation on MRI could be a result of taxane causing increased vascular permeability and gadolinium uptake, or related to an inflammatory infiltrate or necrosis.
43,
45 We found more discrepancies in the diffuse tumor phenotypes, which likely reflects increased difficulty in measuring tumor diameter. These discrepancies were particularly notable in the diffuse HR+/Her2− tumors and make it more difficult to set expectations based on post-NAC MRI in these tumor types. However, in the setting of HR− tumors of solid phenotypes, post-NAC imaging did not underestimate residual tumor size. When MRI showed a pCR, the surgical pathology was concordant.
The strengths of this study include central assessment of HR/Her2 status and the consistent timing of MRIs. There were dedicated breast radiologists at each site who underwent centralized training to validate the MR phenotypes. Size assessment can be somewhat subjective, however, especially for diffuse tumors. Adding tumor volume measurements may help to decrease the chance of overestimating the tumor size compared to surgical pathology. Despite these limitations, however, the findings are consistent with those reported in the literature, with the additional finding that MRI size estimates are less likely to correspond well with pathology for HR+ diffuse tumors.
These findings have clinical implications. Whereas the majority did not attain a pCR, most patients attained clinically meaningful tumor reduction. The MRI phenotype and tumor subtype can inform the discussion about the likelihood of achieving enough response to be potentially eligible for BCT. Whether this information would increase rates of receiving BCT is unknown, but increased understanding, particularly of the accuracy of post-NAC MRI, could impact recommendations and patient decisions.
Overall, many more patients have clinically meaningful tumor reductions than have a pCR. Although the reasons for not receiving BCT are complex, there is likely room for improvement in offering BCT to more patients. We are currently developing an algorithm based on biologic and MRI features to help determine the chances of having a clinically meaningful tumor reduction and the likely accuracy of MR post-NAC to guide this decision-making process.
Acknowledgment
The research for Alliance (CALGB) 150007 and 150012 was supported, in part, by grants from the National Cancer Institute (CA31946) to the Alliance for Clinical Trials in Oncology (Monica M. Bertagnolli, M.D., Chair) and to the Alliance Statistical Center (Daniel J. Sargent, Ph.D., CA33601). Additional funding includes: National Cancer Institute Specialized Program of Research Excellence in Breast Cancer (CA58207), American College of Radiology Imaging Network (CA079778 & CA080098), National Cancer Institute Center for Bioinformatics, The Breast Cancer Research Foundation, and Bruce and Martha Atwater. The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute. Dr. Mukhtar received support from the California Breast Cancer Research Program (Post-doctoral fellowship 15FB-0108).