We thank Louise Mazur and Karen Baurle for clerical assistance.
Complete pathologic response to neoadjuvant chemotherapy (NACT) is predominantly seen in “ERBB2” and “basal-like” tumors using expression profiling. We hypothesize that a similar response could be predicted using semiquantitative immunohistochemistry for estrogen receptors (ER), progesterone receptors (PR), and human epidermal growth factor receptor 2 (HER2).
ER, PR, and HER2 were used to classify 359 tumors treated with NACT into 6 groups: luminal A (strong ER+, HER2 negative), luminal B (weak to moderate ER+, HER2 negative), triple negative (negative for ER, PR, and HER2), ERBB2 (negative for ER and PR, but HER2+), luminal A-HER2 hybrid (strong ER+ and HER2+), and luminal B-HER2 hybrid (weak to moderate ER+ and HER2+). Complete pathologic response was defined as absence of invasive carcinoma in the breast and regional lymph nodes.
Thirteen percent (48 of 359) demonstrated complete pathologic response. The highest rate of complete pathologic response was seen in ERBB2 (33%; 19 of 57) and triple negative (30%; 24 of 79) tumor classes. Among the ER+ “molecular” group, the highest rate of complete pathologic response was seen among luminal B-HER2 hybrid tumors, 8% (2 of 24). Remainder of ER+ tumors demonstrated a very low rate of complete pathologic response, 1.5% (3 of 198). The 5-year survival for patients achieving complete pathologic response was 96% compared with 75% in patients that failed to achieve complete pathologic response. The overall survival was worse in the ER-negative group (ERBB2 and triple negative) compared with the ER-positive group.
Traditionally, neoadjuvant chemotherapy (NACT) has been used in locally advanced breast cancers that are deemed inoperable.1 More recently, it has been increasingly used for T2 and T3 tumors that are resectable, but the intent is to reduce the tumor size by NACT and subsequently remove a smaller portion of breast tissue than would otherwise be removed at primary surgery.2-4 NACT followed by excision provides a unique opportunity to objectively assess in vivo tumor response to chemotherapeutic drugs. This information can be used for future patient treatment for further consolidation in case of complete response or trigger alternative therapeutic regimen in cases of minimal or no response.
Breast cancer is a heterogeneous disease at the morphologic, immunohistochemical (IHC), and even molecular level. Therefore, some breast carcinomas respond completely to NACT and others show minimal or no response.4, 5 Apart from the breast tumor itself, other factors that may predict response include the number of cycles and type of chemotherapy used,6 tumor tissue microenvironment, and host response. The latter factors are no doubt important, but the current chemotherapeutic protocols are rather limited and use a very similar type of chemotherapy (anthracycline or taxane based) except for human epidermal growth factor receptor 2 (HER2)-positive patients who also receive trastuzumab, and most patients receive 4-8 cycles before surgical excision.7-9 The tumor tissue microenvironment and host response may also be important determinants of chemotherapeutic response but appears to be tightly linked to tumor type itself.
On the basis of decades of morphologic data, it is well-known that only a subset of ductal carcinomas shows pathologic complete response to NACT. Recent gene expression-based studies have shown complete pathologic response in 45% of ERBB2 and basal-like breast carcinomas compared with only 6% in luminal subtype.10
The primary aim of this study was to analyze the NACT database at our institution and identify the molecular classes (using IHC surrogate markers) that predict complete pathologic response to NACT. The molecular classes, tumor grade, and complete pathologic response were correlated with disease-free survival (DFS) and overall survival (OS).
MATERIALS AND METHODS
The institutional review board approved the study. The patient case list was obtained from the hospital tumor registry, which represented consecutive cases treated with NACT at our institution. However, some cases did not meet the criteria for inclusion in the study as explained in Table 1. Clinical information and pathology data was carefully extracted from the clinical charts and reports to include a total of 359 cases for this study (Table 1).
Table 1. Data Acquisition and Reasons for Case Deletion
HER2 indicates human epidermal growth factor receptor 2; FISH, fluorescence in situ hybridization; IHC, immunohistochemical.
Data acquisition: Late 1999 to early 2006
Total patients: 439
Total cases: 451
Cases deleted: 92
Total cases classified and examined for complete pathologic response: 359
Reasons for case deletion
• Not a true neoadjuvant chemotherapy case, ie, excisional biopsy performed that removed a large portion of the tumor
• Receptor study results not available
• Receptor results not semiquantitated
• HER2 results not available
• ERBB2(HER2) FISH not performed on HER2 IHC 2+ case
• Suboptimal pathologic exam to determine complete pathologic response
Pretherapy biopsy slides were available for review on 191 cases (ie, minus 168 cases)
IHC Surrogate Markers of Molecular Classification
Semiquantitative IHC results for hormone receptor and HER2 were obtained from pathology reports to classify tumors into molecular classes (Table 2). By using the criteria mentioned in Table 2, the tumors were classified as luminal A, luminal B, ERBB2, triple negative, luminal A-HER2 hybrid, and luminal B-HER2 hybrid. These criteria correspond to the initial gene expression profiling studies and are similar (but not identical) to the prior studies that used IHC as a surrogate for molecular classification.11-13 The criteria shown in Table 2 primarily correspond to Cheang et al's criteria,14 but we have subdivided the luminal category into luminal A, luminal B, luminal A-HER2 hybrid, and luminal B-HER2 hybrid based on hormone receptor expression level and HER2 positivity. The subdivision was performed to study the impact of semiquantitative hormone receptor analysis and coexpression of hormone receptors and HER2 on chemotherapeutic tumor response and survival. At our institution, estrogen receptor (ER) and progesterone receptor (PR) results are reported using a semiquantitative score (previously described as “H-score”), which details the percentage of positive cells showing none, weak, moderate, or strong staining.15, 16 The score is given as the sum of the percentage staining multiplied by an ordinal value corresponding to the intensity level (0 = none, 1 = weak, 2 = moderate, 3 = strong). With 4 intensity levels, the resulting score ranges from 0 (no staining in the tumor) to 300 (diffuse intense staining of the tumor). HER2 protein was analyzed and scored using either HercepTest (1999-2003) or CB11 antibody (2004-2006) and basic DAB detection on Benchmark XT (Ventana, Tucson, Ariz). HER2 was considered positive with either 3+ immunoreactivity (diffuse strong reactivity in >10% of the tumor cells) or unequivocal amplification by fluorescence in situ hybridization (FISH) (with a ratio of HER2 to chromosome 17 centromeric region >2.2, using PathVysion Vysis dual-color FISH by Vysis Inc., Downers Grove, Ill). All 2+ cases (by immunohistochemistry) included in this study were followed by FISH. Equivocal FISH result (1.8: 2.2 ratio) was considered as negative for HER2 in this study.
Table 2. IHC Criteria Used for Molecular Classification
Triple negative phenotype; core basal if + for CK5/6 or EGFR; 5 negative phenotype if negative for CK5/6 and EGFR
? Luminal C
Luminal A-HER2 hybrid
ER score 200 or higher, HER2 positive
? Luminal C
Luminal B-HER2 hybrid
ER score 11-199 or PR score >10, HER2 positive
A detailed morphologic examination of pretherapy core biopsy specimen was performed on 191 cases. Slides from the remaining 168 specimens were not available for review. Several morphologic parameters were recorded as follows: tumor type (ductal, lobular, other), tumor subtype, tubule formation score (1-3), nuclear pleomorphism score (1-3), mitotic activity score (1-3), total score (3-9), Nottingham grade (1-3), absolute mitoses count/10 high-power fields, tumor borders (infiltrative/pushing), sheet-like growth pattern, lobular-like growth, spindle cells, apocrine differentiation, nucleoli (visibility at ×10), lymphoid infiltrate (10% or less/11%-50%/>50%), geographic necrosis, apoptosis (visibility at ×10), ductal carcinoma in situ (type/grade/comedonecrosis), lobular carcinoma in situ, and lymphovascular space invasion. These criteria were slightly modified from the published criteria by Fulford and colleagues.17
Patient Treatment With NACT
NACT given to patients included compounds that are generally given as adjuvant chemotherapy.2 Because this was not a clinical trial, the patient population was diverse and the treatment received by patients was variable. NACT given to the patients was mainly based on clinical staging and physician discretion. The type of therapy given can be broadly classified as anthracycline-based therapy, ie, doxorubicin (Adriamycin) plus cyclophosphamide, 5-flurouracil, epirubicin, and cyclophosphamide, or taxane-based therapy, ie, taxol plus carboplatinum, and in cases of HER2-positive tumors, such as taxol, carboplatinum, and trastuzumab. In many cases a sequential combination of anthracycline and taxane was given, ie, doxorubicin (Adriamycin) plus cyclophosphamide followed by taxane. The doses given were generally the recommended doses per unit body surface area. The total number of cycles ranged from 4-10 with an average total cycle number of 6. Trastuzumab was part of the NACT regimen in only 6 ERBB2 tumors and 10 non-ERBB2, HER2-positive tumors. Trastuzumab was more often administered as adjuvant therapy in HER2-positive tumors.
Complete pathologic response was defined as absence of invasive carcinoma in the post-therapy resection specimen and within regional lymph nodes. Percentage tumor size reduction was also calculated based on pretherapy clinical tumor size and residual tumor size in the resection specimen.
Disease-free survival (DFS) was calculated as the time from the date of diagnosis of the primary tumor to the date of the development of distant or locoregional recurrence. Overall survival (OS) was calculated as the time from the date of diagnosis of the primary tumor to the date of death from any cause, or the date of last contact. The data were analyzed for DFS and OS separately for patients with complete pathologic response and patients without complete pathologic response. Patient DFS and OS with respect to molecular classes and tumor grade was also analyzed.
The chi-square (Monte Carlo method) test was used to compare tumor stages among molecular classes and NACT given among molecular classes. The Kruskal-Wallis test was used to evaluate possible differences of ages among molecular classes. Univariate logistic regression analyses were performed for each potential predictor variable of complete pathologic response to NACT. Variables were included in the multivariate analysis if the univariate P-value was ≤.25. A stepwise selection procedure was used to predict the probability of achieving complete pathologic response. The survival experience was examined with the 2 methods of time-to-event analyses: DFS and OS. The Kaplan-Meier survival curves were evaluated for difference among groups with the log-rank test. P-values <.05 were considered statistically significant. The data analysis for this paper was generated using SAS software, Version 9.2 of the SAS System for Windows, product of SAS Institute Inc. (Cary, NC).
The 359 tumors treated with NACT comprised 111 (31%) luminal A, 73 (20%) luminal B, 57 (16%) ERBB2, 79 (22%) triple negative, 15 (4%) luminal A-HER2 hybrid, and 24 (7%) luminal B-HER2 hybrid. As these cases were considered for NACT, most were stage II or higher at the time of diagnosis. There was no significant difference in tumor stage among different molecular classes (P = .129). The patients with strong ER+ tumors (luminal A and luminal A-HER2 hybrid with median age of 53 and 54 years) were 5-10 years older than weak/moderate ER-positive and ER-negative tumors (luminal B-HER2 hybrid, luminal B, ERBB2, triple negative with median ages of 44, 47, 48, 49 years, respectively), a difference that was statistically significant (P = .0002). The patient clinical data and tumor characteristics are shown in Table 3.
Table 3. Patient Clinical Data and Tumor Characteristics for Each Molecular Class
Of the 359 cases that could be classified using IHC surrogate markers, 48 cases (13%) demonstrated complete pathologic response. The highest rate of complete pathologic response was seen in ERBB2 and triple-negative tumor classes with 33% (19 of 57) and 30% (24 of 79) showing complete pathologic response, respectively (Table 4). Among the ER+ “molecular” groups, the highest rate of complete pathologic response was seen among luminal B-HER2 hybrid tumors, 8% (2 of 24). The remainder of ER+ tumors demonstrated dismal rate of complete pathologic response, 1.5% (3 of 198). Although complete pathologic response information was available on all 359 cases, tumor size reduction could not be analyzed on 30% of cases because of lack of information about either pretherapy tumor size or ambiguous reporting of the post-therapy tumor size. The precise cellularity of the post-therapy residual tumor specimen was not available; therefore, quantification (such as Miller-Payne scoring)9, 18 of residual tumor was not feasible. Despite these shortcomings, the percentage tumor size reduction (available on 70% of cases) deduced from gross and microscopic description provided a reasonable alternative to compare tumor size reduction in different tumor classes. The mean tumor size reduction was higher in ERBB2 and triple-negative tumors compared with other classes. The average percentage tumor size reduction in descending order was as follows: triple negative (75%), ERBB2 (68%), luminal B-HER2 hybrid (47%), luminal A-HER2 hybrid (33%), luminal B (30%), and luminal A (23%).
Table 4. Molecular Classes and Complete Pathologic Response
Absence of Complete Pathologic Response
Complete Pathologic Response
Luminal A-HER2 hybrid
Luminal B-HER2 hybrid
Type and Amount of Chemotherapy
The patients received anthracycline-based or taxane-based chemotherapy, with most receiving a sequential combination of both (Table 3). However, there were slight differences between chemotherapy given to patients depending on receptor status, HER2 status, clinical stage, and physician preference. Despite this variability, there was no statistical difference for NACT given (for the most effective regimen doxorubicin plus cyclophosphamide followed by T) between patients that achieved complete pathologic response and patients that failed to achieve complete pathologic response (P = .9108). Trastuzumab was part of the NACT regimen in 6 ERBB2 tumors, of which 5 (83%) achieved complete pathologic response. In contrast, 10 patients with non-ERBB2, but HER2-positive tumors (ie, luminal A-HER2 hybrid and luminal B-HER2 hybrid tumors) received trastuzumab, of which none achieved complete pathologic response.
Univariate Analysis of All the Analyzed Variables for Complete Pathologic Response
In a univariate analysis, the following factors predicted complete pathologic response: higher Nottingham tumor grade, ER-negative molecular class, HER2 status, apocrine differentiation, necrosis, apoptosis, and presence of spindle cells. However, it should be noted that with regards to morphologic features, data were available only on 191 cases. If one closely examines each of these morphologic parameters, then the correlation between tumor “molecular” classes, morphology and tumor grade can be identified as shown previously.19
Multivariate Analysis for Complete Pathologic Response
The variables, HER2 status, Nottingham grade, spindle cells, apocrine differentiation, geographic necrosis, apoptosis, stage, and molecular class (ER+ vs ER negative), having P-values ≤.25 in univariate analyses, were used to build a model by using multivariate logistic regression. However, only molecular class was found as a statistically significant predictor for complete pathologic response. The probability of complete pathologic response is lower for patients who have ER+ disease than the patients having ER-negative breast cancer molecular class (OR = 0.06 with 95% confidence interval [CI] 0.02-0.18).
Microscopic examination of 191 cases revealed a correlation between higher tumor grade and complete pathologic response, as complete pathologic response was identified in 1 of 13 (7%) of grade 1, 14 of 125 (11%) grade 2, and 14 of 53 (26%) of grade 3 tumors (P = .0073 for grade 3 vs others). The most high-grade tumors (Nottingham grade 3) were seen predominantly in the triple-negative group (58% cases), less commonly in the ERBB2 (18% cases) and luminal B-HER2 hybrid (11% cases) groups, and least commonly in other molecular groups (ie, 13% of combined luminal A, luminal B, and luminal A-HER2 hybrid cases). However, once the tumors were assigned to one of the molecular classes, further division by Nottingham grade did not add any other information that would have predicted complete pathologic response. In addition, several morphologic features were associated with particular molecular class, such as necrosis and spindle cells with triple-negative tumors, and apocrine differentiation with ERBB2 class. The association of apocrine differentiation with ERBB2 tumor class was quite strong and was identified in 26 of 29 (90%) ERBB2 tumors compared with 35 of 162 (22%) of all other tumors combined (P = .0001).
An average of a 46-month follow-up period was available on all cases. The average follow-up period was similar for cases that showed complete pathologic response (mean, 46 months; median, 42 months; range, 21-99 months) and for cases that did not show complete pathologic response (mean: 46 months; median, 42 months; range, 10-104 months).
Of the 48 patients that showed complete pathologic response, 5 patients recurred at an average of 27 months from initial diagnosis, of which 2 died, both of breast cancer. Forty-six (including the 3 patients that recurred) of 48 patients (ie, 96% patients) that showed complete pathologic response were alive (please see Figs. 1A and 1B for DFS and OS).
Of the 311 patients that failed to show complete pathologic response, 92 recurred at an average of 22 months, 199 had no recurrence, and on 20 cases, recurrence data were not available. In this group, 77 died (ie, 75% patients were alive) in this relatively short follow-up period (please see Figs. 1A and 1B for DFS and OS).
Patient DFS and OS were also analyzed based on tumor grade. There was no difference in DFS among different tumor grades, but OS was the worst for grade 3 tumors (P = .0474).
To test the validity of “triple-negative paradox,” the DFS and OS were compared between ER-negative (ERBB2 and triple negative) and ER+ (luminal A, luminal B, luminal A-HER2 hybrid, and luminal B-HER2 hybrid) molecular classes in patients that did not achieve complete pathologic response. The DFS and OS were worse for ER-negative molecular classes (P = .0007 for DFS and <.0001 for OS; see Figs. 2A and 2B). The DFS and OS for ER+ versus ER-negative tumors was also examined after stratifying by stage and the data remained statistically significant for each stage.
Breast cancer molecular classification using “intrinsic” gene set demonstrated distinct molecular classes with prognostic significance.20-22 As per our current understanding, these classes are luminal A, luminal B, ERBB2, and basal-like.23 The so-called “normal breast-like” group of tumors is now believed to be an artifact of sampling and likely represented poorly sampled tumor tissue rather than a distinct, clinically important group.23, 24 The initially described luminal C category was later included along with luminal B tumors.21, 22 Most IHC studies that followed extrapolated the gene expression findings in less than optimal fashion. In a recent similar study, Carey et al classified the tumors based on IHC as luminal A (hormone receptor+/HER2 negative), luminal B (hormone receptor+/HER2+), HER2 group (hormone receptor negative/HER2+), and basal subtype (hormone receptor negative/HER negative). However, the authors clearly indicated in their study that “hormone receptor+/HER2+ tumors comprise only a minority of luminal B tumors, so this method of subcategorizing the luminal subtype will necessarily misclassify a substantial fraction of luminal B tumors into the luminal A category.”11 We have previously raised similar concerns about this issue.25 Therefore, to faithfully reproduce the gene expression profiling studies, we have used semiquantitative IHC (see Table 2). We believe this distinction is necessary to study the impact of semiquantitative IHC hormone receptor analysis and coexpression of hormone receptors and HER2 on chemotherapeutic tumor response and survival.
By using the simple IHC criteria shown in Table 2, the study findings suggest that routine hormone receptor and HER2 analysis can act as a substitute for expression analysis in predicting complete pathologic response to NACT. In one of the very first studies examining the relationship between breast cancer molecular classes (using gene expression analysis) and NACT tumor response, Rouzier et al demonstrated complete pathologic response in 45% of ERBB2 and “basal-like” breast carcinomas compared with only 6% complete pathologic response in luminal tumors.10 Our results are very similar in the sense that complete pathologic response was predominantly restricted to ERBB2 and triple-negative tumor classes; however, the percentages of cases showing complete pathologic response are slightly lower (30%-33% vs 45%). This difference is likely explained by the larger and more diverse study population in the current study and by slight differences in chemotherapeutic regimen between the 2 studies. However, our findings are very similar to the results obtained by Carey and colleagues,11 and we believe the response rates observed in the current study are more representative of the clinical experience.
Our semiquantitative hormone receptor analysis and IHC criteria for molecular classes were helpful in predicting complete pathologic response and also showed subtle differences in percentage tumor size reduction among these classes. Average tumor size reduction appears to be related to the amount of ER expression and presence of HER2 overexpression. Surgeons and oncologists alike can use this information in making better decisions about NACT. If the intent is to achieve complete pathologic response, only triple-negative and ERBB2 tumors should be considered for NACT. The only other tumor type that may also be considered would be an luminal B-HER2 hybrid tumor. If the intent is to reduce tumor size so that a smaller lumpectomy can be performed, then all tumors may be considered knowing that the tumor size reduction would be lowest in strong ER+, HER2-negative tumors. Because of short follow-up in the current study, the survival differences were not apparent among the ER+ tumors, but were significantly different between ER+ and ER-negative “molecular” classes. Despite the best response to NACT, ERBB2 tumors and triple-negative tumors showed the worst DFS and OS (5-year survival of 65% for stage II and 45% for stage III). This worse survival was seen only in the cases that did not achieve complete pathologic response, as cases that achieved complete pathologic response had excellent survival irrespective of the molecular class (5-year survival of 96%). Our findings are concordant with Carey et al, who called this phenomenon “triple-negative paradox”,11 ie, the good response to NACT, but poor survival rates because of higher relapse among those with residual disease. Because both triple-negative and ERBB2 tumors demonstrate this phenomenon, this triple-negative paradox should rather be called “hormone receptor-negative paradox.”
One unique feature of our study was review of the pretherapy tumor slides and detailed morphologic analysis on 191 cases. This examination showed that tumor grades and molecular classes are not completely independent of each other. It was not surprising to see that 3 tumor grades distinctly showed different response to NACT and were also prognostic. The tumor grades and molecular classes (as identified by IHC) are closely interlinked. Most ERBB2 and triple-negative tumors are Nottingham grade 3, while luminal-type tumors are generally well-differentiated to moderately differentiated tumors. Although, both ERBB2 and triple-negative tumors responds equally to NACT, the morphologic examination suggested that the mechanism of response is probably different. The triple-negative tumors are the most proliferative as was shown in the initial gene-expression profiling study20, 21 and later confirmed with morphologic studies.26, 27 The literature also suggests that ERBB2 tumors are high-grade, but most studies fail to mention that ERBB2 tumors are not as proliferative as triple-negative tumors. In the current study, a mitotic activity score of 3 was identified in 22 of 40 (55%) triple-negative tumors in contrast to only 5 of 29 (17%) ERBB2 tumors. This suggests that effect of NACT in triple-negative tumors is likely related to destruction of rapidly proliferating cells but the effect of NACT in ERBB2 tumors is related to mechanism other than inhibition of cellular proliferation. Rouzier et al also reached at the same conclusion as the genes responsible for complete pathologic response in ERBB2 and “basal-like” tumors in their study were distinct in each group.10
Another instructive feature of our study was the difference in response rates among HER2+ tumors with the inclusion of trastuzumab in the NACT regimen. The use of trastuzumab in metastatic and adjuvant setting is now well-established. However, the use of trastuzumab in the neoadjuvant and adjuvant setting was mainly limited to clinical trials before 2005. This explains why only a total of 16 (of 96) HER2+ patients received trastuzumab in our study. Despite these small numbers, the maximum benefit was seen in ERBB2 (ER+/PR+/HER2+) tumors compared with luminal A-HER2 hybrid (strong ER+/HER2+) and luminal B-HER2 hybrid (weak to moderate ER+/HER2+) tumors. Although none of the luminal A-HER2 hybrid or luminal B-HER2 hybrid tumors that received trastuzumab showed complete pathologic response in this study, the percentage tumor size reduction was far greater in luminal B-HER2 hybrid tumors than in luminal A-HER2 hybrid tumors, indicating extent of ER reactivity by IHC could influence tumor response to trastuzumab. The effect of low ER expression on IHC coupled with HER2 positivity was further exemplified by 8% complete pathologic response rate in luminal B-HER2 hybrid tumors compared with only 1.5% complete pathologic response rate in all other ER+ tumors.
In summary, we have shown that simple IHC-based categorization of breast tumors can help predict extent of tumor response to NACT. The predictive power of IHC criteria appears to be similar to that of gene expression analysis. There are morphologic and immunohistologic correlates to molecular classes. Semiquantitative IHC analyses of hormone receptors helps in better categorization of breast tumors than a mere positive or negative result. This information can be used toward improved therapeutic decisions.