Standardizing Pathologic Evaluation of Breast Carcinoma After Neoadjuvant Chemotherapy: Recommendations From the I-SPY Pathology Working Group

Accurate identification of the original tumor site in the breast specimen is the most important step for proper classification of treatment response (Figure 1). The general term fibrous tumor bed or tumor bed area refers to the grossly or radiologically identifiable area of breast parenchyma where the tumor was located before initiation of systemic therapy (Figure 2). It is relatively easy to identify residual tumor when there is minimal or no response to NCT. Tumors, particularly smaller ones that show complete or near-complete resolution on clinical examination and imaging studies, are difficult to visualize on gross examination. Even larger tumors with a pronounced response to treatment can be difficult to delineate in a background of dense breast tissue. As the gross findings are typically subtle, radiologic markers and careful palpation of the specimen slices are important to identify tumor bed area(s). Indeed, access to specimen radiography greatly improves the accuracy and confidence in pathologic evaluation after neoadjuvant treatment (Figure 3). Residual calcification patterns can also help to identify tumor location in the specimen radiograph. However, it is important to remember that calcifications are not always associated with invasive carcinoma, and the extent of calcification may not correlate with the disease extent. Lacking these, one must rely on a detailed description of the pretreatment tumor location (eg, quadrant, distance from nipple) or sutures placed by the surgeon or the microscopic proof of the tumor bed to ensure the prior tumor site was removed.

It is important for the grossing pathologist (or pathology assistant) to have the information regarding the number, size, and location of tumor(s) and biopsy clips and any involvement of skin, chest wall, or nipple before treatment, and any documented response to therapy on clinical notes and imaging studies. As with any resection specimen, when the residual tumor or tumor bed is identified, the size of the tumor bed area and any grossly recognizable residual tumor should be documented, at least in 2 greatest dimensions, along with its morphologic characteristics and relationship to the surgical margins.

Although there are no strict guidelines regarding the extent to which tumor bed should be sampled, it is recommended that if there is an obvious tumor bed on gross examination that is large (>3 cm), one section per centimeter of the pretreatment carcinoma size is reasonable. If the residual tumor bed is small (≤3 cm), the entire fibrous area should be submitted for pathologic examination. To determine the dimensions and cellularity of the residual invasive cancer, it is necessary to map the tumor bed and correlate the microscopic sections with gross findings. An important objective is to map out a complete cross section of the tumor bed along its longest axis, because residual carcinoma could be present in any region of the tumor bed (Figure 4). Any additional sampling should be based on other gross and imaging findings in the specimen, as well as pretreatment tumor characteristics and response to treatment from clinical and imaging reports. The RCB system offers a prescriptive protocol with recommendations on tumor sampling on its Web site (http://www.mdanderson.org/breastcancer_RCB).

Microscopically, the tumor bed is characterized by an irregular area of vascularized fibrous stroma that either lacks or has scant normal ducts and lobules (Figure 5). The stroma may exhibit varying degrees of edema, myxoid change, fibrosis with or without elastosis, residual calcification, and inflammatory cells. The inflammatory response typically contains lymphocytes, plasma cells, mast cells, macrophages, and hemosiderin-laden macrophages, sometimes in loose aggregates or large sheets. As with calcifications, extracellular mucin associated with carcinoma does not resolve in the breast or in the metastatic sites after NCT (Figure 6). It is important to confirm the presence of tumor bed microscopically when there is a complete response to therapy. Although some clinicians may find it useful for comparison, the size of the tumor bed is not a reliable indicator of the pretreatment tumor size, as a significant response to treatment shrinks the overall area of a tumor.

Therapy-related changes in residual tumor, if present, are not specific to any agent or regimen. The most common therapy-related change is decrease in cellularity. Additional changes seen in residual carcinoma after NCT include increased cytoplasm, cytoplasmic vacuolization and eosinophilia, pleomorphic and bizarre nuclei, and decreased mitotic activity (Figure 7). Prominent stromal retraction artifacts around the tumor nests are common and should not be misinterpreted as lymphatic or vascular invasion. Residual single tumor cells in cases with near-complete response can be difficult to discern from inflammatory cells on routine hematoxylin-eosin stain. In such cases, immunohistochemical (IHC) stains for cytokeratin AE1/3 can be used to delineate the distribution of tumor cells. Residual ductal carcinoma in situ (DCIS) trapped in fibrous stroma can be mistaken for invasive carcinoma (Figure 8). Occasionally, lymphovascular invasion (LVI) can be the only residual disease after chemotherapy and should not be mistaken for residual DCIS.⁴⁰  Many of the challenges listed above can be resolved by IHC for epithelial, myoepithelial, and endothelial markers.

Determining residual tumor size can be difficult after treatment. It is easier to determine the size of the residual carcinoma when the tumor shrinks concentrically, leaving one contiguous focus, an uncommon scenario after NCT (Figure 9). Most treated tumors with moderate to marked response to therapy leave scattered nodules of residual tumor, multiple clusters, or single tumor cells in the tumor bed (Figure 10, A and B).⁴¹  For a detailed discussion on how to measure residual tumor size for the AJCC and RCB systems, please refer to the corresponding sections.

It may be difficult to identify 10 or more lymph nodes from an axillary dissection because of nodal atrophy and fibrosis induced by chemotherapy.⁴²  Therefore, axillary fat should be carefully searched for lymph nodes, whether it is from an axillary dissection or sentinel lymph node (SLN) biopsy, and all identified nodes without grossly visible tumor should be sliced at 2- to 3-mm intervals perpendicular to the long axis of the node and submitted entirely.⁴³  In instances where very few nodes are found on palpation, submitting fibrotic areas and the tissue around the vessels may reveal additional small lymph nodes.

In most cases, the response in nodal metastases mirrors the response observed in the primary tumor (Figure 1).⁴⁴  Nodal metastases completely eradicated by treatment are replaced by fibrous scars or sheets of macrophages (Figure 11). It is rare to have a large subcapsular scar in axillary lymph nodes of patients who undergo primary surgery for breast carcinoma. Therefore, it is important to document the presence of scars or other findings in negative nodes that most likely represent treated metastasis. It is possible that a prior micrometastasis or smaller metastasis may resolve entirely without leaving a fibrous scar. Residual disease in metastatic nodes with partial response to NCT is often distributed either as single cells or as clusters of tumor cells surrounded by fibrosis. It is recommended that pathologists document any treatment effect in residual metastases. Please see the pertinent section below for measuring the size of metastases after treatment.

Because a large proportion of patients with TNBC and HER2⁺ cancer achieve pCR in both breast and axilla, many centers are performing SLN surgery to avoid the morbidity of axillary dissection.^45–47  It should be emphasized to the clinical team to place a biopsy clip in the lymph node at the time of fine-needle aspiration or image-guided core needle biopsy of a clinically suspicious or positive node (or prior to initiation of systemic therapy if not placed at the time of aspiration or core biopsy). This will ensure removal of the biopsied positive node or suspicious node at the time of definitive surgery. It is important for pathologists to document in the pathology report whether the lymph node with marker clip was removed or not at the time of surgery.

Intraoperative evaluation of SLNs is quite challenging because of fibrosis and treatment-related changes. In an NCT setting, using touch imprints only for intraoperative evaluation of lymph nodes can lead to a higher false-negative rate. Gimbergues et al⁴⁸  reported lower sensitivity and negative predictive value in detecting tumor cells in the SLN using touch imprints after NCT. The authors concluded that the lower sensitivity was related to a higher proportion of cases with micrometastasis and isolated tumor cells (ITCs) after NCT. Although frozen sections may reduce the overall number of false-negative results, the rate may still be significantly high in lobular carcinomas, smaller metastases, and ITCs.

Routine use of cytokeratin to detect metastasis is not currently recommended for SLNs after NCT. Pathologists should use their judgment to determine if a case would benefit from cytokeratin stain. Some of the clinical scenarios where cytokeratin stain may be useful include prior positive node from a lobular carcinoma or HR⁺ tumor and any node with scattered single tumor cells embedded in therapy-related fibro-inflammatory changes or any atypical cell that is difficult to characterize as tumor versus nontumor (Figure 12). Given the challenges of identifying residual lymph node metastasis in the NCT setting, pathologists should have a low threshold for using IHC to decrease the chance of missing metastasis.

Given the high concordance rate of ER, PR, and HER2 status between core biopsies and resections (>90%), most treatment-naïve breast cancers are not retested at excision.⁴⁹  In the NCT setting, however, clinically actionable alterations in ER and HER2 have been reported in 3% to 18% and 1% to 7% of residual tumors, respectively.^50–52  Reasons implicated in discrepant biomarker status between pretreatment and posttreatment tumors include preanalytic factors (eg, cold ischemia time), intratumoral heterogeneity, unsampled tumor, and response to targeted therapy, that is, conversion of HER2⁺ tumor to HER2⁻ status following HER2-targeted therapies (Figure 13).

Current practice data indicate considerable variation in retesting tumor biomarkers after neoadjuvant therapy, with only 65% of surveyed academic pathology practices in the United States reporting routine biomarker retesting.⁵³  Repeat testing of HER2-positive carcinomas is particularly controversial because of the potential impact on patient management with HER2-targeted therapies. Although the current College of American Pathologists (CAP) guidelines for ER/PR testing⁵⁴  and HER2 testing⁵⁵  in breast cancer do not specifically address this issue, other groups such as the Breast International Group and North American Breast Cancer Group have released recommendations for biomarker retesting in the post-NCT setting.⁵⁶  For routine clinical practice, we recommend that retesting of ER, PR, and HER2 be considered based on the pretreatment biopsy testing results, pathologic features of the residual tumor, and whether repeat testing will result in a change in treatment. Example scenarios for consideration of repeat testing include residual tumor(s) with intratumoral heterogeneity or lack of response to targeted therapy; pretreatment tumor biopsy with negative or equivocal biomarker results; heterogeneous, unusual, or unanticipated results; and undersampled or marginally sufficient sample of pretreatment tumor (Figure 14).

IHC assessment of Ki-67 expression is a validated surrogate of proliferation in many tumor types, including breast cancer, and has both prognostic and predictive value. Although intralaboratory reproducibility has been shown to be high,^57,58  concerns about low interlaboratory reproducibility and lack of standard protocol on testing and scoring have prevented universal adoption of Ki-67 testing in breast cancer. Nevertheless, many laboratories have incorporated Ki-67 testing as part of their routine tumor profile testing.

Posttreatment Ki-67 index has been shown to be of prognostic value for both neoadjuvant endocrine therapy and chemotherapy in ER⁺/HER2⁻ disease.⁵⁹  In the NCT setting, change in the Ki-67 index after treatment can be useful to measure partial responses and for consideration of additional adjuvant therapy.⁶⁰  However, not all laboratories currently perform Ki-67 testing of breast cancers before or after NCT. Nonetheless, some clinical trials use centralized Ki-67 testing, including a recently FDA-approved Ki-67 assay as a companion diagnostic test for ER⁺/HER2⁻ breast cancer treated with abemaciclib.⁶¹  While there remains a need for standardization of Ki-67 testing and scoring, pathologists should take the initiative to optimize the assay in their own laboratory or use an automated system for routine clinical use.

Immune profiling has rapidly expanded prognostic and predictive immune-related biomarkers within the tumoral microenvironment. Tumor-infiltrating lymphocytes (TILs) have been shown to be a reproducible biomarker affecting the prognosis and response to treatment in breast cancer.^62–68  In the era of targeted immunotherapy (eg, pembrolizumab for PD-L1⁺ TNBC),^8,69  there is an increasing clinical demand for assessment of TILs and immune-related biomarkers in pretreatment biopsy and posttreatment residual tumors. In a study, the combination of RCB and TILs was a better predictor of outcome than TILs alone.⁷⁰  For example, TILs were strongly prognostic in TNBC only in the RCB-II class.⁷¹  The International Immuno-Oncology Biomarker Working Group on Breast Cancer has issued a guideline for assessing TILs in the residual carcinoma after NCT.⁷²  However, evaluating TILs in breast cancer remains challenging, particularly in pretreatment core biopsies, because of tumor heterogeneity and concerns about interobserver reproducibility.⁷³  Currently, quantitative assessment of TILs in pretreatment tumor biopsy or posttreatment residual tumor is not a standard practice among pathologists.

Multiple methods have been proposed to assess the degree of response to NCT to provide prognostic information, all of which require some degree of residual tumor quantification. Some classification methods, such as those proposed by the National Surgical Adjuvant Breast and Bowel Project B-18,⁴  Penault-Llorca et al,²⁷  Chevallier et al,²⁸  and Pinder et al,²⁹  use broad categories such as complete (or near-complete), partial, or no response. These methods may²⁹  or may not^4,27,28  include lymph node status in the assessment. Although the prognosis of patients with tumors achieving complete or near-complete pathologic response is clearly better than that of nonresponders, it is important to identify assessment methods that can further stratify the prognosis of patients with varying degrees of partial response (Table 1). Methods proposed by Sataloff et al,³²  Smith et al,³³  and Bonadonna et al³⁵  have added additional degrees of response based on the percentage of residual tumor cells. To correct for variation in the cellularity of different pretreatment tumors, the Miller-Payne system³¹  proposes comparison of the posttreatment and pretreatment tumor cellularity, rather than posttreatment cellularity assessment alone. Although this controls for tumor heterogeneity, it does not include lymph node involvement as a prognostic factor in the assessment.

The Residual Disease in Breast and Nodes classification system proposed by Chollet et al³⁴  is one of the most encompassing formulas, including the size of the residual tumor, tumor grade, and nodal involvement to place patients into 4 risk categories. Although most available post-NCT assessment methods are similar to one another and provide prognostic value, RCB is the only Web-based formula to include measurements of tumor in both the breast and lymph nodes with the inclusion of tumor cellularity in the breast to quantify response to NCT as a continuous variable.³⁰  The distributions of RCB have been compared in the randomized treatment arms from clinical trials to evaluate overall treatment efficacy.⁷⁴  Additionally, external validation of the RCB system to demonstrate generalizable prognostic performance within each subtype of breast cancer after NCT is underway.

The RCB system is a method to quantify the extent of residual disease in the resection specimen of patients who received NCT.³⁰  It measures residual disease as a continuous variable, combining the input from the primary tumor bed and axillary lymph nodes. The discussions below are meant to clarify some of the questions that were shared by the working group from the feedback obtained from various pathologists who are asked to use the RCB system or are familiar with the reporting of the RCB system.

For the cross-sectional dimensions of the primary tumor bed, the principle is to provide measurements of the 2 largest dimensions of the tumor bed occupied by residual invasive cancer, irrespective of the cellularity or the distribution of the invasive tumor cells within that area. These dimensions are determined by correlating gross and microscopic examination of the specimen, with microscopic determination of the tumor bed used if they are discordant. Hence, there is considerable value in mapping the specimen indicating where the sections were taken (Figure 4). For patients who have multifocal or multicentric tumors (similar in morphology or in tumor profile), measurements of the largest residual tumor should be used to report RCB. In instances where multiple tumors with different morphology and/or tumor profiles are identified, one should calculate RCB separately for each one.

The overall cancer cellularity in the RCB formula refers to the proportion of cancer in the tumor bed area (as percentage residual cancer cellularity). If the tumor bed spans several tissue blocks, cellularity is estimated for each block in order to estimate the average cellularity for the entire residual tumor area. LVI is interpreted as residual invasive cancer in the RCB system.

Cellularity assessments are typically identified to the nearest 10%, with additional selections of 1% and 5% for low cellularity. But one can enter any value, and for tumors with very low cellularity scattered intermittently over a large area, that value may be less than 1%. The usual misunderstanding is to make estimates only in foci of the tumor bed that contain significant residual tumor. The estimates should represent the average across the entire area of residual invasive cancer, not the most concentrated foci.

One of the challenges in the RCB system is that the calculator requires 2 levels of information (the percentage of the area that is invasive cancer, and the percentage of the cancer that is in situ) to calculate the percentage of the area that is invasive cancer. Many pathologists find this breakdown to be redundant and difficult to ascertain, as they are able to confidently determine in one visual step the proportion of invasive component in an area. Additionally, pathologists are not clear on when not to include the in situ component in the calculation (for example, cases where residual DCIS extends beyond invasive carcinoma and classic lobular carcinoma in situ). Because the inclusion of residual DCIS does not affect the final RCB class or index, the classification of pCR, and, importantly, the outcome of patients, the working group recommends simplifying the RCB calculator to minimize errors (Table 2).

First introduced in 1959, the AJCC staging system, currently in its 8th edition, defines the anatomic TNM staging classification based on primary tumor (T), regional lymph nodes (N), and distant metastases (M).^36,75  In the 8th edition, the AJCC, for the first time, further defined prognostic stage groups by integrating anatomic stage data with other prognostic factors such as histologic grade, and ER, PR, and HER2 status for breast carcinoma. The 8th edition staging manual establishes postneoadjuvant therapy clinical ycTNM classification (defined by clinical and radiographic findings) and pathologic ypTNM classification (determined by microscopic examination of the primary tumor, any lymph nodes, and any metastatic foci).^75,76 

Multiple prospective clinical trials have shown that a pCR is associated with significantly improved survival. Presence of residual in situ carcinoma after NCT in the absence of invasive cancer (classified as ypTis) constitutes pCR by AJCC. Clinical trials using this definition have shown no difference in overall survival whether residual DCIS is present or not in the tumor bed.⁷⁷  When reporting pCR, a grossly identifiable tumor bed confirmed by microscopic examination is sufficient to document pCR. However, the current AJCC staging manual does not provide clear guidance on the pT staging of non-pCR cases. Below we discuss some of the common issues in the current AJCC staging of treated tumors and suggest modifications to consider for the 9th edition to improve clarity and consistency in reporting residual carcinoma after NCT.

Most non-pCR cases have residual invasive carcinoma interspersed by uneven areas of fibrous stroma without tumor cells. Per AJCC staging guidelines, posttherapy tumor size is based on the largest contiguous focus with a modifier “m” if multiple foci are present. We suggest that the modifier m be reserved for clearly multifocal and multicentric tumors (similar to untreated tumors); for example, the modifier m should be reserved when residual tumor is present in each pretreated tumor.

The definition of the pT category is derived from the largest diameter of contiguous invasive cancer by the AJCC system. Although this is certainly reasonable for measuring untreated carcinomas, there are considerable challenges when this approach is applied to variably distributed residual tumors with uneven cell density after NCT. There is subjectivity regarding how strictly to apply the term contiguous, but pathologists should probably be lenient in measuring the largest tumor dimension by including small patches of fibrosis between tumor nests, in addition to describing the total volume of tumor, that is, the number of tumor foci in the tumor bed, as suggested by the AJCC staging manual.⁷⁵ 

The AJCC 8th edition affirms that the presence of lymphatic or vascular tumor emboli without residual invasive carcinoma in the breast precludes posttreatment classification as a pCR.⁷⁵  However, AJCC has not designated any specific ypT classification for this setting. These cases could be classified as ypT(LVI), with a note describing the extent of the lymphatic-vascular channel involvement.

The AJCC ypN staging uses the same categories used for untreated tumors (pN), and is based on the largest contiguous focus of residual tumor in the lymph node (fibrosis between metastatic tumor foci is not included). This method, similar to the T categorization, may underestimate the residual tumor burden in lymph nodes. The method for measuring residual node metastasis should be similar to measuring primary tumor diameter.

Currently, ITCs are not included in the positive lymph node count. Scant or isolated residual tumor cells (less than 0.2 mm or less than 200 cells) in the lymph nodes, whether associated with treatment effect (fibrosis) or not, are currently staged as ypN0(i+), which may be mistaken as a complete nodal response. The presence of ITCs in the NCT setting may not have the same clinical significance compared with untreated tumors.^78,79  Therefore, in order to avoid understaging and undertreatment, nodes with ITCs should not be classified as ypN0 (i+) but as ypN1(i+).

Currently, reporting of extranodal extension is not mandatory in the AJCC system; however, a standardized reporting of extranodal extension should be a part of risk stratification for both adjuvant and neoadjuvant cases, as extranodal extension may be a marker of poor prognosis.^80–82 

Table 3 lists the elements that should be included in the pathology report of resected specimens of treated breast carcinomas. It is mandatory for pathologists to use a synoptic format to report breast carcinomas, both untreated and treated, as recommended by the CAP. In Table 4, we have included the portion of the current CAP protocol⁸³  that is applicable for the reporting of treated breast cancer specimens and proposes revisions to a future version for better categorization of treatment responses. In addition, we suggest that the CAP breast cancer protocol should include a Web-based classification system such as RCB for a standardized assessment of tumor burden, hence allowing comparison of the predictive value afforded by different systems.

One of the future steps that has been tried is the evaluation of treatment response by image-guided core biopsy during NCT to determine whether a specific chemotherapy regimen is effective.⁸⁴  This approach would also allow medical oncologists to modify the therapeutic regimen for nonresponders while informing a response-based decision for surgical treatment. However, the timing of the biopsy will need to be determined by future studies and may vary for different tumor types and treatment regimens.

Currently, a standardized post-NCT breast carcinoma reporting system is not available. The goal should be to maximize the clinical utility of the reported information while keeping the process as efficient as possible, thereby allowing for generalized adoption of the system. This approach would also provide extensive population data available for further studies in a relatively short time frame. To achieve this goal, every pathology practice should implement a standard operating procedure for collecting, handling, and interpreting pathology specimens treated with NST. All pathologists participating in neoadjuvant trials should receive formal training as recommended by the FDA guidance document.⁶ 

Finally, artificial intelligence has recently shown promising utility in applications in medicine and pathology. Incorporating machine learning to evaluate treated breast cancers to standardize reporting and search for prognostic and predictive factors will be an important next step in breast cancer pathology.

In summary, a pressing need is to develop a postneoadjuvant synoptic report format relevant to reporting residual disease after NST. As discussed earlier, the current CAP synoptic report for invasive breast cancer and the RCB system require modifications to address the specific challenges posed by neoadjuvant treatment. To refine a patient's prognostic risk, any future modifications to predictive models, whether the AJCC system or RCB calculator, should be based on pretreatment tumor characteristics and patient outcomes based on the posttreatment response category.

The I-SPY 2 Trial Pathology Working Group comprises pathologists working across national trial sites. I-SPY 2 is a multicenter, adaptive randomization phase 2 neoadjuvant platform trial for early-stage, high-risk breast cancer patients. A subcommittee of the I-SPY 2 Pathology Working Group was formed by a group of pathologists with extensive experience in examining and reporting on breast and lymph node specimens after NCT. This subcommittee met virtually (once a month for a year and quarterly thereafter) to compile and compare reporting practices across sites and discuss the issues and controversies related to therapy before providing recommendations.