Pathologic Response and Long-Term Follow-up in Breast Cancer Patients Treated With Neoadjuvant Chemotherapy: A Comparison Between Classifications and Their Practical Application

Sixty-two patients were enrolled between 2000 and 2004 in randomized clinical trial 99-278 (ClinicalTrials.gov identifier NCT00096291) for sequential NACT with doxorubicin (A) and paclitaxel (T).³⁴  Patient consent and approval by the institutional review board had been obtained. Patients were randomized to receive both chemotherapeutic agents doxorubicin (A) and paclitaxel (T) in either the sequence A > T or T > A. At interim analysis for the pathology evaluation, the survival in both groups was not significantly different, and statistical analysis included stratification for treatment randomization. Patients underwent diagnostic core biopsies before NACT, followed by excision or mastectomy after NACT. Mean (median) patient age at diagnosis was 49 (48) years, ranging from 30 to 68 years. For patients without recurrence, the mean and median follow-up time was 93 months.

Tissue samples were fixed in 10% neutrally buffered formalin and paraffin embedded following standard protocols of the histology laboratory. Five-micrometer sections were stained with hematoxylin-eosin for histologic review. If original slides were unavailable at the time of this study, recuts from archival paraffin blocks were prepared. Immunohistochemical stains for E-cadherin, myoepithelial cells (p63, calponin, and smooth muscle myosin), lymphatic endothelial cells (D2-40 and CD31), and estrogen and progesterone receptor proteins were prepared following standard immunoperoxidase protocols. Human epidermal growth factor receptor 2 evaluation was primarily done by immunohistochemistry with scoring as described followed by fluorescence in situ hybridization in cases with equivocal (2+) Human Epidermal Growth Factor Receptor 2 (HER2) immunohistochemical staining.³⁵  Most immunohistochemical and fluorescence in situ hybridization tests were done at time of primary diagnosis.

Study pathologists reviewed hematoxylin and eosin and immunohistochemically stained sections from patients' tumors sampled before (core biopsy) and after treatment (excision or mastectomy). The findings reported at time of initial histopathologic evaluation were updated when necessary. Histologic grading was performed as previously described.^36,37  The macroscopic measurements of tumor bed area and residual tumor mass were supplemented by microscopic measurements. Proportion of ductal carcinoma in situ and overall tumor cellularity was estimated following the instructions of the authors or using the Web-based calculator as applicable.^23,24  The number of positive lymph nodes was grouped for American Joint Committee on Cancer staging and individual classification schemes.^22,26,29 

All patients had magnetic resonance imaging of the breast before NACT. The maximum diameter of the largest nodule evaluated by magnetic resonance imaging served as tumor size pre-NACT, and the gross or histologic evaluations defined tumor size post-NACT.

Distant relapse was defined as histopathologic and/or radiologic documentation of distant metastasis. Distant disease-free survival (DDFS) was defined as the time interval between surgery and the first documented distant relapse, death, or last follow-up. Overall survival (OS) was defined as the time between surgery and death or last follow-up, whichever occurred first. The T and N categories were assigned according to the American Joint Committee on Cancer.²⁹  Survival curves were estimated by Kaplan-Meier product-limit estimates. Log-rank tests were used for unadjusted, univariate analysis of each classification scheme as a predictor of DDFS and OS, stratified by treatment randomization. Linear trend tests were used for ordinal classifications. Cox proportional-hazards regression, adjusting for age and stratified by treatment randomization, was used for age-adjusted univariate and multiple regression models. Number of positive lymph nodes and maximum tumor diameter were Winsorized at their 90th percentiles to avoid undue influence by a few highly leveraged points. P values were increased by Winsorizing in all cases. P values < .05 were considered statistically significant. Hazard ratios (HRs) were given when trends were tested across groups. Statistical analyses were performed using SAS (version 9.2, SAS Institute, Cary, North Carolina). Graphs were generated with MedCalc Software (version 11.2.1, MedCalc Software, Mariakerke, Belgium).

The findings are summarized in Table 1. Mean (median) tumor size was 4.3 (3.5) cm pre-NACT and 1.8 (1.4) cm post-NACT. The majority of tumors were invasive ductal carcinomas (n = 54), with positive estrogen receptor and negative human epidermal growth factor receptor 2 status. Lymph node involvement was associated with survival. No correlations with DDFS or OS were detected between pre-NACT tumor size measured by magnetic resonance imaging and post-NACT tumor size established by pathologic evaluation. Pre-NACT median tumor size in the 5 breast tumors with pCR was 6.5 cm, and was therefore larger than the median tumor size in the overall study population of 3.5 cm. Also, in our study, histologic type (invasive ductal versus invasive lobular carcinoma), histologic grade, presence of lymphovascular invasion, estrogen receptor or Her2 status, and some patient characteristics, such as age or menopausal status, did not correlate with prognosis.

Among the classification schemes evaluated, the age-adjusted trend for the 4 levels of RDBN predicted DDFS but not OS in our cohort (Figure 2, A and B; Table 2). The age-unadjusted trend for the 4 levels of RDBN predicted both DDFS (P = .01; HR, 2.54; 95% confidence interval [CI], 1.36–5.08) and OS (P = .04; HR, 2.25; 95% CI, 1.06–5.16). The continuous RDBN score was also associated with DDFS (P = .01; HR, 1.55; 95% CI, 1.11–2.25) but not OS (P = .08; HR, 1.44; 95% CI, 0.98–2.20) when analyzed by Cox proportional-hazards regression with adjustment for age. The RDBN score remained a significant independent predictor of DDFS even in multiple Cox regression models that adjusted for age and each of the pathologic predictors Sataloff tumor (P = .002; HR, 1.88; 95% CI, 1.29–2.91), Sataloff nodes (P = .02; HR, 2.13; 95% CI, 1.22–4.25), RCB (P = .01; HR, 2.07; 95% CI, 1.23–3.56), pretreatment tumor size (P = .003; HR, 1.79; 95% CI, 1.24–2.71), posttreatment tumor size (P = .001; HR, 2.1; 95% CI, 1.37 to 3.34), NSABP-B18 (P = .01; HR, 1.88; 95% CI, 1.22–3.05), and number of involved lymph nodes (P = .02; HR, 2.36; 95% CI, 1.26–5.51). RDBN remained a strong but not significant independent predictor in models with age and AJCC grouping of lymph nodes (P = .07; HR, 1.83; 95% CI, 1.04–3.94) and the maximum size of lymph node metastases (P = .07; HR, 1.79; 95% CI, 1.02–3.51). The small number of observed distant recurrences (n = 16 among patients with scored samples) precluded evaluation of more complex models.

Median DDFS and OS were shorter among patients with higher RCB scores, but the associations were not significant either by age-unadjusted log-rank test (P = .09 and .08, respectively) or by age-adjusted Cox regression for the continuous RCB score (P = .07 and .14, respectively) (Figure 3, A and B; Table 2).

More positive lymph nodes after NACT correlated with worse outcome (Figure 4, A and B; Table 1). The size of the largest lymph node deposit as measured microscopically also predicted DDFS (P = .01; HR, 1.1; 95% CI, 1.04–1.33) but not OS (P = .17; HR, 1.1; 95% CI, 0.95–1.27). Seventeen patients had been diagnosed with axillary lymph node metastases by cytology (n = 5) or sentinel node excision (n = 12) before NACT; the lymph node status pre-NACT was not associated with survival (data not shown). Lymph node score according to the Sataloff system was associated with DDFS by Kaplan-Meier (Figure 5, A; Table 2) and Cox regression analysis (P = .04 and .04, respectively), but not with OS (P = .11).

29% of all patients (n = 18) developed distant metastases during follow-up, whereas none of the patients with no residual carcinoma in the breast after NACT relapsed (Table 2). Absence of invasive tumor cells in the breast (n = 5) post-NACT was considered pCR with the NSABP-B18, and grade 5 with the Miller-Payne classification. Two patients with pCR in the breast had micrometastatic deposits in the axillary lymph nodes; they were subsequently excluded from the RDBN level 1 (pCR) category (n = 3), which requires absence of invasive tumor in both breast and lymph nodes (Figure 2; Table 2). Pathologic complete response according to the RCB criteria requires absence of invasive carcinoma in both breast and lymph nodes for the most favorable categories (Figure 3; Table 2).

We found no statistically significant association between NSABP-B18 categories and DDFS or OS by log-rank test or Cox regression. Incongruent patterns of the Kaplan-Meier survival curves, that is, better survival of patients with larger tumor or score grades or with less regressive changes over those with smaller tumors, were observed with classifications based exclusively on tumor size or estimates of tumor cellularity such as Miller-Payne grading or Sataloff tumor (Table 2; Figure 5, B through D).

In our study, the composite index RDBN to assess histologic response after NACT was associated with relapse-free survival by age-adjusted univariate and multivariate analysis. Because of the small size of our cohort and rare events in certain groups, statistical evaluation had limitations. Residual disease in breast and nodes and lymph node status for this cohort were statistically significant for DDFS but not for OS when adjusted for age. In age-unadjusted calculations, however, both were significant. The authors think that qualitative interpretation of the survival analysis presented in this study is valuable despite those limitations, and that results should subsequently be validated on larger sample sizes and different cohorts.

Our study confirms that the number of positive lymph nodes is a highly important parameter for the long-term outcome after NACT. We showed that the number of involved lymph nodes as well as the size of metastases is relevant for long-term outcome (Figure 4; Table 1).³⁸  When comparing different classification schemes, those that integrate lymph node involvement, such as RDBN and RCB, demonstrated better correlation with long-term outcome than those based only on breast tumor size or cellularity.

Pathologic complete response indicates an overall favorable survival in breast cancer after NACT, but pCR is defined differently in different classifications and may be redefined for future practice.⁵  Both the RDBN and RCB systems require absence of invasive carcinoma in both breast and lymph nodes.^23,26  Both NSABP-B18 and Miller-Payne grading require absence of invasive carcinoma in the breast but do not include the assessment of the lymph nodes.^6,21  The Sataloff T-A category is defined as “total or near-total therapeutic effect,” which may include a proportion of residual invasive carcinoma, whereas the Sataloff N-A/B category refers to negative lymph nodes.^22,32  The number of patients with pCR in the cohort of this study was small (n = 5), and included patients with no residual tumor in both breast and lymph nodes (n = 3), as well as those with no tumor in the breast but with residual tumor in the lymph nodes (n = 2).

Histologic grade is a well-established prognostic factor for treatment-naïve breast cancer, which also has predictive value for post-NACT breast cancer evaluation.^25,31,36  In RDBN, high tumor grade weighs in as a negative prognostic factor when residual tumor is present, in addition to lymph node involvement and tumor size.²⁶  Other post-NACT response classifications discussed in this study do not consider histologic grade.^{6,21–23,25,36,39}  The favorable survival of low-grade tumors in our cohort coincided with the one of complete responders, and those were high grade.^25,26,31,40  Prior studies emphasized the lack of response to NACT for invasive lobular carcinomas or other low-grade breast cancers.^8,41,42  Overall, high tumor grade was associated with less favorable survival, which is in agreement with other studies whose classifications use histologic grade.^25,26,31,40  Histologic grade was reassessed for tumors in pre-NACT core biopsies and post-NACT excision or mastectomy specimens by histologic slide review; minor differences between pre- and post-NACT grade were observed for low-grade tumors (Table 1). The limiting factor for grading core biopsies was small sample size and poor preservation in 3 cases. The authors conclude that histologic grade on post-NACT specimens is feasible, mostly congruent with the pre-NACT tumor grade, and important to report.

Related to histologic grade and proliferative activity are molecular subtypes of breast cancer, which are increasingly used for prediction of neoadjuvant response.^5,43,44  Independently of their genomic subtype, the post-NACT breast excision or mastectomy specimens still have to be evaluated by a pathologist to assess individual response.

Classifications based on routine staging and grading parameters (eg, tumor size, lymph node involvement, histologic grade) were easier to apply in this retrospective setting than those that require dedicated or prospective measurements to obtain the information in the grossing area or at the microscope (eg, tumor bed size, comparison with prior core biopsy, cellularity estimates). Direct morphologic comparison with the core biopsy is of course desirable, but may not always be possible in practice because some patients have their diagnostic core biopsy at a different hospital. Also, the core biopsy might not necessarily be representative of the entire tumor.

In summary, we confirmed the favorable prognosis of patients with pCR after NACT, none of whom relapsed. Overall, 29% of patients (n = 18) developed distant metastases during follow-up of up to 10 years. Because the majority of breast carcinomas treated with NACT in mixed cohorts do not show pCR, the pathologist needs to determine the extent of residual tumor in breast and lymph nodes. Among the different classifications used to assess pathologic response after NACT, composite classifications that integrate features of primary tumor and lymph node status showed better associations with outcome than those using only a single tumor characteristic, for example, tumor size or cellularity. In this study, RDBN was associated with clinical outcome by univariate and multivariate analysis. However, practical considerations will contribute to the decision which classification, if any, is chosen by an institution to report breast carcinoma response after NACT.