SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References
  9. Supporting Information

Weekly PCb (paclitaxel + carboplatin) in neoadjuvant chemotherapy (NCT) for breast cancer has a high pathological complete remission (pCR) rate. The present study was to identify pCR predictive biomarkers and to test whether integrating candidate molecular biomarkers can improve the pCR predictive accuracy. Ninety-one breast cancer patients treated with weekly PCb NCT were retrospectively analyzed. Eleven candidate molecular biomarkers (Tau, β-tubulin III, PTEN, MAP4, thioredoxin, multidrug resistance-1, Ki67, p53, Bcl-2, BAX, and ERCC1) were detected by immunohistochemistry in pre-NCT core needle biopsy specimens. We analyzed the relationship between these biomarkers and pCR. Univariate analysis showed that estrogen receptor, progesterone receptor, molecular classification (clinicopathological markers), and Tau, β-tubulin III, p53, Bcl-2, ERCC1 (candidate molecular biomarkers) expression were associated with pCR rate; however, multivariate analysis revealed that only β-tubulin III, Bcl-2, and ERCC1 were independent pCR predictive factors. Patients with β-tubulin III negative, Bcl-2 negative, or ERCC1 negative tumors were associated with higher pCR rate, with OR (odds ratios) 6.03 (95% confidence interval [CI], 1.44–25.24, = 0.014), 7.54 (95% CI, 1.52–37.40, = 0.013), and 4.09 (95% CI, 1.17–14.30, = 0.028), respectively. To compare different logistic regression models, built with different combinations of these variables, we found that the model integrating routine clinical and pathological variables, as well as the β-tubulin III, Bcl-2, ERCC1 molecular biomarkers had the highest pCR predictive power. The area under the ROC curve for this model was 0.900 (95% CI, 0.831–0.968), indicating that it deserves further investigation. Trial name: Weekly Paclitaxel Plus Carboplatin in Preoperative Treatment of Breast Cancer. (Cancer Sci 2012; 103: 262–268)

Neoadjuvant chemotherapy is the standard treatment for patients with locally advanced breast cancer. In early breast cancer, NCT is associated with significantly higher breast conservation rate compared with those receiving adjuvant chemotherapy.(1) Numerous phase III randomized clinical trials have shown that patients who achieved pathological complete remission after NCT had better prognosis than those who did not.(2,3) Furthermore, NCT can test in vivo tumor chemosensitivity. It offers great opportunities to carry out biological studies on primary tumor, and to better understand the mechanisms of tumor response and chemoresistance, thus improving the efficacy of breast cancer treatment.

Paclitaxel, knowing its antitumor activity through tubulin stabilization and cell cycle arrest, has revolutionized breast cancer therapy.(4) The Early Breast Cancer Trialists’ Collaborative Group has revealed that taxane-based adjuvant chemotherapy can significantly improve the outcome of patients compared to those treated with anthracyclin-based therapy. The mechanism of carboplatin action involves covalent binding to purine DNA bases, which primarily leads to cellular apoptosis.(5) Paclitaxel combined with carboplatin has shown great activity in ovarian and non-small-cell lung cancer treatment. In NCT, we have previously reported that a weekly PCb regimen had great antitumor activity and tolerability in breast cancer patients, with the pCR rate as high as 19.4%.(6)

Not all patients who receive NCT can achieve pCR, with pCR rates in early studies reported to be 6% and 28%.(7) Similarly, the pCR rate was 19.4% in patients treated with PCb. So we need to identify the patients who will respond, thus allowing a tailored NCT to be formulated.(8) However, at present, we still lack reliable markers to predict the efficacy of PCb in breast cancer treatment.

Breast cancer is a heterogeneous disease. Patients with histologically similar tumors may have different prognosis and treatment responses. In NCT, several factors like HR negativity, high histological grade, and Ki67 levels may correlate with a high pCR rate, whereas the predictive value of other biomarkers, like HER2 expression, and Bcl-2 and p53 status, are still unclear.(8) Based on microarray and real-time PCR technology, we have identified several molecular classifications of breast cancer by different gene expression profiling, which can better predict prognosis and efficacy compared to routine clinicopathological parameters.(9–16) In addition, models based on gene expression profiling have great pCR predictive accuracy.(13–16) A nomogram that includes several clinicopathological factors can more accurately predict pCR than a single clinicopathological marker.(17) With the development of pharmacogenomics, we can predict the efficacy of NCT from the genetic level, using candidate gene or whole genome methods.(18) Several molecular biomarkers, such as Tau, MAP4, β-tubulin III, PTEN, thioredoxin, MDR-1, Ki67, p53, Bcl-2, BAX, ER, and HER2, could be associated with paclitaxel efficacy.(19,20) Apoptosis-related genes (p53, Bcl-2, and BAX) and DNA repair genes (ERCC1) may correlate with carboplatin activity.(21) Candidate molecular biomarker status can be detected by IHC rather than RT-PCR assay, with the following advantages: (i) it eliminates the stromal and inflammatory cell contamination problem by observing only staining cancer cells; (ii) it enables retrospective analysis by using paraffin sections; and (iii) it represents more reliable gene function than mRNA expression.(22)

Based on these factors, we analyzed these candidate molecular biomarkers in breast cancer patients treated with weekly PCb to identify pCR predictive markers. Furthermore, constructed different models using various factors to evaluate whether models integrating candidate molecular biomarkers can improve pCR predictive accuracy.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References
  9. Supporting Information

Patients and treatments.  This study was carried out retrospectively in patients who participated in a weekly PCb NCT phase II trial between December 2007 and December 2008 (Cancer Hospital, Fudan University, Shanghai, China). A total of 107 patients were recruited with pCR rate 19.4%.(6) All of these patients underwent unltrasound-guided core needle biopsy to confirm invasive breast cancer and received four cycles of PCb with the dose of 80 mg/m2 paclitaxel and carboplatin AUC = 2, given on days 1, 8, and 15 out of every 4 weeks. Detailed physical examination, mammography, ultrasound, and MRI were carried out before treatment. Regional lymph node status was confirmed by fine needle aspiration for patients with palpable node. Disease stage was evaluated according to the AJCC guidelines (6th edition). Ninety-one breast cancer patients had adequate samples for molecular biomarker analysis. Age, menstrual status, height, weight, histopathology type, primary tumor size, AJCC staging, ER, PR, and HER2 status were prospectively recorded before NCT.

Immunohistochemistry.  Expression of candidate molecular biomarkers, including Tau, β-tubulin III, PTEN, MAP4, thioredoxin, MDR-1, Ki67, p53, Bcl-2, BAX, and ERCC1 in pre-NCT specimens was detected by IHC according to the manufacturer’s protocol. Characteristics of antibodies used in this study are summarized in Table S1.

The IHC results were evaluated by two senior pathologists in a blind procedure without knowing patients’ outcome. β-tubulin III, Tau, thioredoxin, MAP4, MDR-1, Bcl-2, and BAX status was recorded by the percentage of positive tumor cells with any cytoplasm staining; Ki67, p53, PTEN, and ERCC1 status was determined by the percentage of positive tumor cells with any nuclear staining. A minimum of 200 cells were counted, and IHC scores were assigned to the variables except Ki67 (IHC −, ≤10%; IHC 1+, 11–29%; IHC 2+, 30–49%; and IHC 3+, ≥50%). Distributions of IHC scores for each marker are summarized in Table S2. The expression for each biomarker was analyzed as a categorical variable, and each was evaluated as high versus low. Tumors with low expression were considered to be biomarker negative and cases with high expression were considered to be biomarker positive. The cut-off value for Ki67 was calculated by the mean percentage of positive tumor cells in this population. Thresholds defined for each variable are summarized in Table S1, which were mostly according to previous reports.(19,20,22) In addition, we adjusted the cut-off values for β-tubulin III, MAP4, and thioredoxin variables from 50% to 30% which was optimized for response prediction.(19,20)

Eestrogen receptor, PR, and HER2 status before NCT were assessed by IHC, which was routinely carried out in the Department of Pathology (Cancer Hospital). The methods and cut-off values of ER, PR, and HER2 evaluation were described in our previous study.(6) Hormone receptor positivity was defined as ER/PR positivity. Breast cancer was then classified into four molecular subtypes as follows: luminal A (HR+/HER2−); luminal B (HR+/HER2+); triple-negative (HR−/HER2−); and HER2 positive (HR−/HER2+).

Response evaluation after NCT.  Standard Response Evaluation Criteria in Solid Tumors guidelines were used to evaluate response. Pathological complete remission was defined as non-invasive tumor cell in breast and axillary samples. Of three cases with clinical progression disease, one patient received further chemotherapy and did not undergo surgery, and the remainder were treated with surgery after two and four cycles of NCT. These three patients were classified as pathological progression disease.

Statistical methods.  Clinicopathological characteristics and pathological response were described by tabulation. Clinicopathological factors and candidate molecular biomarkers expression were compared between patients with pCR and non-pCR by Fisher’s exact test. Spearman correlation analysis was used to calculate the correlation among these variables. Multivariate logistic regression analysis was then applied to examine the independent pCR predictive factor. The pCR predictive accuracy was evaluated among different logistic regression models, which were established by integrating various clinicopathological, molecular classification, or candidate molecular biomarkers. All statistical tests were two-sided at the 5% level of significance and were carried out using spss statistical software version 13.0 (SPSS, Chicago, IL, USA) and Stata software version 8.0 (Statacorp, College Station, TX, USA). Odds ratios are presented with their 95% CI. The study was reviewed and approved by the independent ethical committee/institutional review board, and all patients gave their written informed consent before inclusion in this study.

Results

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References
  9. Supporting Information

Clinicopathological characteristics and treatment response.  Of the 91 patients with adequate specimens to process IHC analysis, 18 patients achieved pCR, with a pCR rate 19.8%. Characteristics of patients are summarized in Table 1. The median age was 52 (32–70) years, 85.7% patients were diagnosed with stage III disease. The proportion of ER and PR positivity was 59.3% and 57.1%, respectively.

Table 1.   Clinicopathological characteristics of 91 patients with breast cancer before neoadjuvant chemotherapy (NCT)
CharacteristicNo.%
Age (mean 52 [range, 32–70] years
 ≤503336.3
 >505863.7
Body mass index
 <245054.9
 ≥244145.1
Menopausal status
 Pre/peri-menopause4448.4
 Post-menopause4751.6
Pathological type
 Invasive ductal carcinoma5661.5
 Invasive carcinoma3235.2
 Specific type of breast cancer33.3
Pre-NCT tumor stage
 0–23336.3
 33033.0
 42830.8
Pre-NCT lymph node stage
 01213.2
 12224.2
 25054.9
 377.7
Pre-NCT TNM classification
 II1314.3
 IIIA4751.6
 IIIB2426.4
 IIIC77.7
Estrogen receptor status
 Negative3740.7
 Positive5459.3
Progesterone receptor status
 Negative3942.9
 Positive5257.1
HER2 status
 Negative7076.9
 Positive2123.1

Age, body mass index (weight/height2), menopausal status, tumor size, histologic type, lymph node status, and HER2 status before NCT showed no significant correlation with pCR. Patients with ER negative tumors had a significantly higher pCR rate (35.1%) than those with ER positive disease (9.3%; P = 0.003), and the pCR rate among patients with PR negative disease was 33.3%, which was much higher than those with PR positive disease (9.6%; P = 0.007). The pCR rate differed significantly among these four molecular classes of breast cancer, patients with triple-negative or HER2 positive subtype were associated with the highest pCR rates, 38.9% and 40.0%, respectively; luminal A tumors had the lowest pCR rate of 7.7% (= 0.003). In addtion, the pCR rate in luminal B subtype (16.7%) was not significantly different compared with triple-negative or HER2 positive subtypes, with P = 0.334 and = 0.322, respectively.

Relationship between candidate molecular biomarkers and pCR to NCT.  There was no significant correlation between MAP4, PTEN, thioredoxin, MDR-1, Ki67, or BAX expression and pCR rate. However, patients with negative Tau, β-tubulin III, p53, Bcl-2, or ERCC1 disease showed a higher pCR rate than those with positive expression (Fig. 1). Patients with negative β-tubulin III disease had a significantly higher pCR rate (30.0%) than those with positive disease (7.32%; P = 0.008). Spearman analysis showed that Tau status was significantly associated with β-tubulin III, MDR1, and Bcl-2 expression. In addition, β-tubulin III expression was significantly related with p53 status, whereas ERCC1 status was significantly related with PTEN, MDR1, and Ki67 expression (Table S3). In luminal A disease, patients had a higher rate of Bcl-2 positive (65.4%) or Tau positive (65.4%) tumors than those with other subtypes (P < 0.001). However, ERCC1 and β-tubulin III expression had no significant difference within these subtypes, with P-values of 0.519 and 0.368, respectively.

image

Figure 1.  Results of immunohistochemical analyses of candidate molecular biomarkers in breast cancer.

Download figure to PowerPoint

Next, we included ER, PR, and the five candidate molecular biomarkers Tau, β-tubulin III, p53, Bcl-2, and ERCC1, which were associated with a higher pCR rate in univariate analysis, to examine the independent pCR predictive factor. In multivariate analysis, β-tubulin III, Bcl-2, and ERCC1 status were identified as independent variables associated with pCR. Patients with negative β-tubulin III, negative Bcl-2, or negative ERCC1 tumors had significantly higher pCR rates, with OR values of 6.03 (95% CI, 1.44–25.24, P = 0.014), 7.54 (95% CI, 1.52–37.40, P = 0.013), and 4.09 (95% CI, 1.17–14.30, P = 0.028), respectively (Table 2). In addition, if molecular subtypes, Tau, β-tubulin III, p53, Bcl-2, and ERCC1 were included in multivariate analysis, independent pCR predictive factors were still β-tubulin III, Bcl-2, and ERCC1. We compared these candidate biomarkers within four molecular subtypes, and we found that these factors were no longer associated with treatment response (data not shown), which was partly due to the small number patients in the separate subtypes.

Table 2.   Multivariate logistic regression analysis of pathological complete remission independent predictive factors in breast cancer
 Odds ratio95% CIP
  1. Bold indicates that these results met statistical significant difference. CI, confidence interval; ERCC1, excision repair cross complementing gene 1.

Estrogen receptor status
 Negative2.410.62–9.380.204
 Positive11 
Progesterone receptor status
 Negative0.780.10–6.290.814
 Positive11 
Tau status
 Negative1.890.37–9.810.448
 Positive11 
β-tubulin III status
 Negative6.031.44–25.240.014
 Positive11 
p53 status
 Negative2.380.66–8.610.187
 Positive11 
Bcl-2 status
 Negative7.541.52–37.400.013
 Positive11 
ERCC1 status
 Negative4.091.17–14.300.028
 Positive11 

Comparison of different pCR predictive models.  To evaluate whether knowledge of these candidate molecular biomarkers can improve pCR predictive probability compared with routine clinicopathological factors, we constructed four different pCR predictive logistic regression models including routine clinical variables (age, tumor, node stage), pathologic variables (ER and HER2 status), molecular classification of breast cancer, and candidate molecular biomarkers (β-tubulin III, Bcl-2, ERCC1) in various combinations (Tables 3 and 4). Estrogen receptor status was an independent pCR predictive marker in model 1 (combining routine clinical and pathological variables), and molecular classification of breast cancer was the only factor associated with pCR in model 2 (combining routine clinical and molecular classification), with P-values of 0.018 and 0.016, respectively (Table 3). However, in model 4, which included routine clinicopathologcial variables and candidate molecular biomarkers, independent pCR predictive variables were still β-tubulin III, Bcl-2, and ERCC1, with P-values of 0.010, 0.016, and 0.026, respectively (Table 4).

Table 3.   Pathological complete remission predictive logistic regression models 1 and 2 for breast cancer
VariablesModel 1: Clinical and pathological variablesModel 2: Clinical variables and molecular classification
OR (95% CI)POR (95% CI)P
  1. Bold indicates that these results met statistical significant difference. CI, confidence interval; ER, estrogen receptor; LN, lymph node; NA, not applicable; NCT, neoadjuvant chemotherapy; OR, odds ratio; T, tumor.

Age, years
 ≤503.18 (0.89–11.39)0.0763.59 (0.96–13.52)0.059
 >5011
Pre-NCT T stage
 0–22.69 (0.81–8.87)0.1052.23 (0.65–7.65)0.203
 3–411
Pre-NCT LN stage
 0–11.47 (0.43–4.98)0.5401.54 (0.44–5.36)0.498
 2–311
ER status
 Negative4.77 (1.31–17.44)0.018NANA
 Positive1NA
HER2 status
 Negative0.48 (0.13–1.77)0.271NANA
 Positive1NA
Molecular classification
 Luminal ANANA1.000.016
 Luminal BNA2.74 (0.21–36.20)0.444
 Triple-negativeNA8.08 (1.76–37.04)0.007
 HER2 positiveNA11.75 (2.28–60.60)0.003
Table 4.   Pathological complete remission predictive logistic regression models 3 and 4 for breast cancer
VariablesModel 3: Candidate molecular biomarkersModel 4: Clinicopathologcial and candidate molecular biomarkers
OR (95% CI)POR (95% CI)P
  1. Bold indicates that these results met statistical significant difference. CI, confidence interval; ER, estrogen receptor; ERCC1, excision repair cross complementing gene 1; LN, lymph node; NCT, neoadjuvant chemotherapy; OR, odds ratio; T, tumor.

Age, years
 ≤50NA 4.54 (0.95–21.78)0.059
 >50NA1
Pre-NCT T stage
 0–2NA 4.06 (0.92–17.91)0.064
 3–4NA1
Pre-NCT LN stage
 0–1NA 3.27 (0.71–15.11)0.129
 2–3NA1
ER status
 NegativeNA 1.75 (0.38–8.04)0.472
 PositiveNA1
HER2 status
 NegativeNA 0.53 (0.11–2.43)0.411
 PositiveNA1
β-tubulin III status
 Negative6.03 (1.44–25.24)0.0147.96 (1.41–44.85)0.019
 Positive11
Bcl-2 status
 Negative7.54 (1.52–37.40)0.01311.23 (1.56–81.06)0.016
 Positive11
ERCC1 status
 Negative4.09 (1.17–14.30)0.0285.32 (1.22–23.13)0.026
 Positive11

Furthermore, we applied ROC curves to measure the predictive accuracy of these four models. Model 3 was built only with β-Tubulin III, Bcl-2, and ERCC1. The AUC values of these models were 0.783 (95% CI, 0.664–0.902) for model 1, 0.801 (95% CI, 0.686–0.916) for model 2, 0.826 (95% CI, 0.732–0.920) for model 3, and 0.900 (95% CI, 0.831–0.968) for model 4, which showed significantly different pCR predictive power (P = 0.002, Fig. 2). To further compare the AUC values one by one, we found that model 4 had a significantly higher AUC value than model 1 and model 2, with P-values of 0.006 and 0.018, respectively.

image

Figure 2.  Receiver operating characteristic curve of pathological complete remission predictive logistic regression models in breast cancer.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References
  9. Supporting Information

The aim of our present study was to evaluate the pCR predictive factors for breast cancer patients treated with weekly PCb NCT, and we found patients with negative β-tubulin III, negative Bcl-2, or negative ERCC1 tumors had a significantly higher pCR rate than those with positive disease. Furthermore, the pCR predictive model integrating routine clinical and pathological variables and β-tubulin III, Bcl-2, ERCC1 status had the highest accuracy to predict the pCR possibility, which indicated that, on the basis of clinicopathological variables, measuring β-tubulin III, Bcl-2, and ERCC1 can improve pCR predictive accuracy in patients treated with weekly PCb NCT.

Taxanes are microtubule-stabilizing agents that function primarily by interfering with spindle microtubule dynamics, causing cell cycle arrest and apoptosis.(4,20) Microtubules are composed of α- and β-tubulin heterodimers, and the cellular target for paclitaxel is on the β-tubulin site.(20)β-tubulin III is an isotype of the β-tubulin family, which is less stable with an increased tendency toward depolymerization compared to other β-tubulin isotypes, indicating that β-tubulin III expression is related with taxane efficacy.(23) Several in vivo studies reported that β-tubulin III status could be considered as a predictive biomarker of taxane treatment response, with high expression of β-tubulin III significantly associated with taxane resistance.(24,25) In a group of 70 patients with advanced breast cancer treated with paclitaxel-based chemotherapy, β-tubulin III status could predict the treatment efficacy; only 2% of patients with low β-tubulin III expression progressed after paclitaxel chemotherapy compared with 38% of those with high β-tubulin III expression (P < 0.001).(26) In our study, we found that patients with β-tubulin III negative tumors had a significantly higher pCR rate than those with positive disease. In multivariate analysis, β-tubulin III status was an independent pCR predictive factor, which indicated that β-tubulin III status was associated with paclitaxel-containing treatment efficacy. Furthermore, Dumontet et al.(27) reported that β-tubulin III status was a prognostic factor for node-positive breast cancer patients treated with docetaxel-based regimen in the BCIRG 001 trial. Long follow-up is required to evaluate the relationship between β-tubulin III expression and outcome in our study.

Bcl-2 is a protein that can inhibit the process of cell apoptosis. Taxane treatment can cause Bcl-2 phosphorylation through various signaling pathways, leading to the inactivation of Bcl-2 and induction of apoptosis.(20) In breast cancer treatment, the efficacy of paclitaxel given every 3 weeks was inferior to a weekly regimen, which may inhibit tumor regrowth between cycles and limit the emergence of malignant cell populations resistant to chemotherapy by enhancing its apoptotic and antiangiogenic effects.(28) In addition, carboplatin can also cause apoptotic effects by causing DNA damage. In an in vivo study, estrogen could increase the intracellular Bcl-2 level and inhibit paclitaxel-induced apoptosis of human breast cancer MCF-7 cells. Blocking the estrogen by tamoxifen could decrease the expression of Bcl-2, which resumed the sensitivity of paclitaxel.(29) In the current study, we found that Bcl-2 was an independent pCR predictive marker; patients with negative Bcl-2 tumors had a higher pCR rate than those with positive disease. However, Poelman et al.(30) had shown that there was no significant association between Bcl-2 expression and response or outcome in metastatic breast cancer patients treated with single paclitaxel. The main explanations for this difference are that patients in our series were treated with weekly paclitaxel and concurrently with weekly carboplatin, which might have greater anti-apoptosis effects than single paclitaxel given with every 3 weeks.(28)

Platinum–DNA adducts can cause distortions in DNA platinum drugs (cisplatin and carboplatin), thus leading to cellular apoptosis. There are several mechanisms involved in DNA repair pathways, including NER, base-excision repair, mismatch repair, and double-strand-break repair. However, NER is considered as the major pathway to remove platinum drug lesions from DNA. As an NER endonuclease protein, ERCC1 plays an important role in the NER pathway and may be related with platinum drug response.(21) In several solid tumors, such as ovarian cancer, lung cancer, and bladder cancer, patients with low ERCC1 level disease had a significantly higher response rate and better outcome than those with high ERCC1 expression disease, and may get more benefit from platinum-based therapy.(31–35) However, there are relatively few data on ERCC1 expression and platinum-based treatment efficacy in breast cancer. Among our cohort of 91 breast cancer patients treated with weekly PCb, we found that patients with ERCC1 negative breast cancer had a much higher response rate than those with positive disease, which was similar to previous reports in other solid tumors.

Previous studies have indicated that ER,(36,37) HER2,(38) Tau,(39) MAP4,(40) MDR-1,(41,42) Ki67,(43) PTEN,(44,45) p53,(46) thioredoxin,(16,22) and BAX(21) status might be associated with taxane or carboplatin response. In univariate analysis, we found that patients with Tau negative tumors had a significantly higher pCR rate than those with positive tumors, which was similar to patients treated with neoadjuvant paclitaxel followed by FAC.(39) However, in multivariate analysis, Tau expression was no longer an independent pCR predictive factor, which may due to the significant correlation with another microtubulin-associated protein, β-tubulin III. In addition, patients with ER or p53 negative disease had a higher chance to achieve pCR, but again in multivariate analysis, these were not independent pCR predictive factors. The absence of an independent relationship between ER status and pCR when adjusting for all markers was surprising, but the direct relationship with Bcl2 expression could provide a possible explanation (Spearman correlation analysis, r = 0.491, P < 0.001).

Many studies have been aimed at predicting the response of NCT in breast cancer, based on gene expression signatures of variably expressed genes detected by gene-chip assay, and have shown that these classifications were superior compared with routine clinical and pathological variables.(13–16) In patients treated with paclitaxel followed by FAC NCT, molecular classification determined by a “breast intrinsic” gene set could predict the pCR rate, and patients with basal-like and HER2 subtypes were more sensitive to NCT than those with other tumors.(47) In our current study, breast cancer molecular classification, constructed by ER, PR, and HER2 status, was an independent pCR predictive factor among routine clinical and pathological variables. Furthermore, Rouzier et al.(47)built three logistic regression models, including various combinations of clinical and histopathological variables, and the molecular class yielded similar AUC values, which indicated that the molecular class alone can replace histopathologic characteristics to predict pCR. However, another report indicated that a model combining clinical and genomic information was nominally best, and may further improve the pCR prediction performance.(14) A logistic regression model, incorporating routine clinicopathological variables and these three candidate molecular biomarkers, had the highest AUC value (0.900, 95% CI, 0.831–0.968) and was much higher than models only including routine clinical factors plus pathological variables or molecular classification, which also indicated that integrating these new candidate molecular biomarkers could improve the pCR predictive performance. In addition, Tau protein expression was inversely associated with the pCR rate in patients receiving NCT.(39) However, high Tau protein expression was associated with better prognosis in patients treated with adjuvant anthracycline and paclitaxel chemotherapy and endocrine therapy in the NSABP-B 28 trial.(48) So, with this relatively short period of follow-up, we are not able to show the relationship between the pCR predictive model and patient survival, which may be due to the heterogeneity of luminal A disease and the effects of adjuvant endocrine therapy after chemotherapy.

There are also several potential limitations to this study. We lack an independent cohort of patients to validate this pCR predictive model due to the retrospective nature of the study and the limited number of patients enrolled. Therefore, our results are still exploratory, and further study is needed to confirm them, especially in the context of other neoadjuvant trials that combine weekly paclitaxel with carboplatin. In addition, selection bias existed in our study by applying the candidate biomarker approach. Finally, breast cancer is a highly heterogeneous disease, which includes many subtypes, such as ER positive, triple-negative, or HER2 positive, each with varying treatment responses and prognosis.(16,47) So, it would be better to study the pCR predictive factors in certain breast cancer subtypes to test whether this predictive model would be useful for all phenotypic subsets of disease or just within some certain phenotypes, like ER negative, high histological grade, or high Ki67 expression disease, thereby reducing the influence of disease heterogeneity.

In summary, β-tubulin III, Bcl-2, and ERCC1 can predict the pCR rate among patients treated with weekly PCb NCT. Patients with β-tubulin III negative, Bcl-2 negative, or ERCC1 negative disease had a higher pCR rate than those with positive disease. A model incorporating routine clinical and pathological variables as well as β-tubulin III, Bcl-2, and ERCC1 improved the pCR predictive power. Further investigation is needed to confirm the clinical significance of these biomarkers in a large number of patients.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References
  9. Supporting Information

This research was supported in part by grants from the Leading Academic Discipline Project of Shanghai Municipal Education Commission (Grant No. J50208), the Cancer Foundation of China (Grant No. 0901), and the Science and Technology Commission of Shanghai Municipality (Grant No. 09411961400).

Abbreviations
AJCC

American Joint Committee on Cancer

AUC

area under the ROC curve

CI

confidence interval

ER

estrogen receptor

ERCC1

excision repair cross complementing gene 1

FAC

fluorouracil–doxorubicin–cyclophosphamide

HR

hormone receptor

IHC

immunohistochemistry

MDR-1

multidrug resistance-1

NCT

neoadjuvant chemotherapy

NER

nucleotide-excision repair

OR

odds ratio

PCb

paclitaxel + carboplatin

pCR

pathological complete remission

PR

progesterone receptor

ROC

receiver operating characteristic

References

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Disclosure Statement
  8. References
  9. Supporting Information

Table S1. Characteristics of antibodies for candidate molecular biomarkers in breast cancer tumors, and their positivity cut-off values.

Table S2. Distributions of immunohistochemical (IHC) scores for each candidate molecular biomarker in breast cancer tumors.

Table S3. Spearman correlation analysis among candidate molecular biomarkers in breast cancer tumors.

FilenameFormatSizeDescription
CAS_2135_sm_tS1-3.doc169KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.