The use of biologic markers to predict response to neoadjuvant chemotherapy may permit tailoring regimens to achieve maximal tumor response. Taxanes have demonstrated excellent activity in breast carcinoma; however, tumor-specific factors that predict clinical response have not been characterized thoroughly.
The authors performed a historic review evaluating the association of tumor prognostic factors and response to neoadjuvant cyclophosphamide and doxorubicin (AC) with or without docetaxel (D) (AC vs. AC+D) in 121 women who previously were enrolled in a Phase III, randomized, clinical trial. Using pretreatment biopsy materials, immunohistochemical studies were performed for estrogen receptor (ER), progesterone receptor (PR), HER-2/neu, p53, and Ki-67. Outcome variables were pathologic complete response (pCR) and positive clinical response (cPOS), which was defined as a ≥ 50% regression in clinical tumor size prior to surgery.
In a multivariate analysis that controlled for tumor size and lymph node status, improved cPOS rates were observed with the addition of docetaxel in women with HER-2/neu-negative tumors (81% vs. 51%; P < 0.05), yielding an adjusted odds ratio of 3.5 (95% confidence interval, 1.2–13.0) in favor of docetaxel. Women who had HER-2/neu-negative tumors appeared to have a lower response rate with AC alone compared with women who had HER-2/neu-positive tumors (51% vs. 75%; P = 0.06), but response rates were matched when docetaxel was added (81% vs. 78%; P = 0.99). ER, PR, p53, and Ki-67 results were not associated significantly with response rates.
Neoadjuvant chemotherapy for patients with locally advanced breast carcinoma has evolved significantly since its early use in the 1970s. Once used primarily for locally advanced, inoperable breast carcinoma, neoadjuvant chemotherapy now is used commonly for patients with operable breast carcinoma.1, 2 Several investigators have established that different regimens of neoadjuvant chemotherapy result in increased breast conservation.3–8 Although it remains to be determined whether neoadjuvant therapy definitively improves disease-free and overall survival, many investigators have demonstrated complete pathologic response rates from 6.0% to 23% with various chemotherapy regimens.1, 2, 9 Complete pathologic responses have been associated subsequently with increased disease-free and overall survival.10–14
Given the promise of neoadjuvant chemotherapy, the ability to predict therapeutic response offers appealing benefits. Perhaps most important, knowledge regarding a patient's expected response to different types of neoadjuvant chemotherapy regimens would permit a tailored approach in selecting the initial therapy that may yield the best pathologic outcome and subsequent overall survival. More aggressive regimens may be selected earlier based on the biology of the patient's tumor and its expected therapeutic response pattern. Conversely, patients with a low expectation of response to neoadjuvant chemotherapy may be spared delay in surgical treatment and potential toxicities related to chemotherapy.
Taxanes have demonstrated excellent activity in breast carcinoma in both neoadjuvant and adjuvant settings.15–17 To study their potential benefits further, we enrolled patients in a randomized, controlled, Phase III clinical trial that was designed to evaluate whether the addition of docetaxel (Taxotere; Aventis Pharmaceuticals, Bridgewater, NJ) to neoadjuvant doxorubicin and cyclophosphamide (AC) improves clinical and pathologic responses in patients with operable breast carcinoma. In part of this study, patients underwent pretreatment pathologic evaluation for prognostic factors, including estrogen receptor (ER) and progesterone receptor (PR) status, HER-2/neu overexpression, Ki-67 proliferative indices, and p53 overexpression. We reviewed data that were collected prospectively from these patients to determine whether the presence of any of these prognostic factors predicted clinical or pathologic response to neoadjuvant chemotherapy with or without docetaxel.
MATERIALS AND METHODS
A historic review was conducted of patients who received neoadjuvant chemotherapy on study between February, 1996 and August, 2000. Enrolled patients were diagnosed with invasive breast carcinoma with clinical staging T1C–T3, N0, M0 or T1–T3, N1, M0 according to American Joint Committee on Cancer (AJCC) staging protocols. Exclusion criteria included arm edema and advanced axillary adenopathy with fixation to each other or to underlying structures. Patients were randomized to one of three neoadjuvant treatment arms. In the control group, each patient was to receive 4 cycles of preoperative cyclophosphamide and doxorubicin every 21 days at 600 mg/m2 and 60 mg/m2 as well as tamoxifen 20 mg per day for 5 years beginning on the first day of chemotherapy. Patients in the other 2 arms received this backbone regimen with the addition of 4 cycles of docetaxel (AC+D) at 100 mg/m2 every 21 days as either neoadjuvant therapy or adjuvant therapy. Patients who received docetaxel were premedicated with dexamethasone, diphenhydramine, and either cimetidine or ranitidine. Consent for study participation was obtained from each patient, and the protocol was approved by the Institutional Review Board.
Response to therapy was graded based on clinical change in tumor size from the greatest tumor dimension at diagnosis to the greatest tumor dimension at the conclusion of neoadjuvant therapy. Clinical responses were classified as complete response (cCR) for 100% tumor regression, partial response (cPR) for 50% to 99% tumor regression, nonresponse (cNR) for < 50% tumor regression and < 25% growth, and tumor progression (cPROG) for ≥ 25% growth. Pathologic response was determined based on histologic analysis of the resected specimen. Pathologic specimens with no residual invasive carcinoma cells in the original tumor or lymph nodes were classified as having undergone pathologic complete response (pCR). Tumors with residual ductal carcinoma in situ were included in the pCR group.
Pretreatment pathologic specimens were obtained by fine-needle aspiration or core biopsy. If adequate material was available, then immunohistochemical analyses for prognostic factors were performed on the specimens by standard immunohistochemical techniques for ER (6F11; Novocastra, Newcastle, UK), PR (Pgr636; DAKO Cytomation, Carpenteria, CA), HER-2/neu (TAB250; Zymed, South San Francisco, CA), Ki-67 (Mib-1; DAKO Cytomation), and p53 (DO-7; Novocastra). The proliferative rate by Ki-67 was calculated from a count of 200 tumor cells. Immunohistochemistry results were classified strictly as positive or negative, with the exception of Ki-67, which was evaluated as a continuous variable. Fluorescent in situ hybridization analysis was performed on all specimens that had borderline HER-2/neu immunohistochemistry scores.
For the analyses, the control group and the adjuvant docetaxel groups were combined and classified as a group that did not receive docetaxel as neoadjuvant therapy (the AC group). This group was then compared with the patients who received docetaxel preoperatively (the AC+D group). In addition, a category for positive clinical responses (cPOS) was designated by combining the cCR and cPR rates, thereby encompassing all clinical responses with ≥ 50% regression. Associations of response with the presence of prognostic factors were evaluated between the AC group and the AC+D group and for both of these groups as a whole. We compared demographic and clinical characteristics with tumor response using contingency-table analysis. Chi-square analysis was used to test the association between categorical variables and tumor response. Multivariate analyses were conducted using unconditional logistic regression, adjusting for clinical tumor size and lymph node status at diagnosis. The SAS software package (version 8.2; SAS Institute Inc., Cary, NC) was used for all statistical analyses. Preselected, first-order interaction terms were included in the analysis to identify potential effect modification.
Of 144 consecutive women who were enrolled within the study period, 139 women completed all planned cycles of neoadjuvant therapy. Among the patients who completed therapy, one patient in the AC group declined surgical treatment. Of the five patients who did not complete therapy, one patient from the AC group died before planned completion from disease-related causes, and another patient in the AC+D group withdrew herself after a perceived good clinical response. The remaining three patients requested surgery before the completion of chemotherapy due either to therapeutic toxicity (one patient in the AC+D group and one patient in the AC group) or to patient concern about disease progression (one patient in the AC+D group).
Clinical response data were available for 142 patients, including the 141 women who underwent surgery and 1 patient who declined surgery after completing chemotherapy. Patient characteristics are shown in Table 1. The average age was 48 years, and 71% of the analytic population were of Hispanic descent. At the time of enrollment, there were 29 patients with AJCC clinical Stage I disease, 49 patients with Stage IIA disease, 49 patients with Stage IIB disease, and 15 patients with Stage IIIA disease. Clinically positive lymph nodes were observed in 39% of patients. Fifty-five patients ultimately underwent modified radical mastectomy, whereas 86 patients underwent breast-conserving surgery. The randomization scheme resulted in no statistical differences in age or lymph node status between the treatment arms. A greater proportion of patients with Stage IIA disease received docetaxel (51.1% of the AC+D group vs. 26.8% of the AC group; P < 0.01). There were no statistical differences among the patients with disease in other stages.
Table 1. Patient Characteristics
No. of patients
AC: doxorubicin and cyclophosphamide; TMX: tamoxifen; D: docetaxel; AJCC: American Joint Committee on Cancer.
Total patients started on neoadjuvant treatment
Completing neoadjuvant therapy
Total undergoing surgery
AC/TMX and neoadjuvant D
AC/TMX and adjuvant D
Clinical tumor size
≤ 2 cm
> 2 cm, ≤ 5 cm
> 5 cm
Clinical lymph node status
Pretreatment AJCC clinical stage
Modified radical mastectomy
With respect to clinical response, a cCR was reported in 44% of patients (n = 63 women), a cPR was reported in 18% of patients (n = 26 women), a cNR in was reported in 32% of patients (n = 46 women), and a cPROG was reported in 5% of patients (n = 7 women). Among the 141 patients for whom surgical pathologic specimens were available, 39 patients had no macroscopic evidence of residual disease, but 8 patients had microscopic foci of residual invasive carcinoma. Consequently, 31 of 141 patients (22%) were categorized with complete pathologic responses. Among the patients who completed neoadjuvant chemotherapy, 37 of 45 patients (82%) who received AC+D had a cPOS response compared with 53 of 97 patients (55%) who received AC. This difference was accounted for primarily among patients with Stage IIB and IIIA disease (Table 2). Comparatively, a complete pathologic response was more balanced between groups, with a rates of 24% (11 of 45 patients) in the AC+D group and 21% (20 of 96 patient) in the AC group.
Table 2. Positive Clinical Response Rates by Treatment Group and American Joint Committee on Cancer Stage
Adequate tissue was available from 121 biopsy specimens for immunohistochemical analysis. For specimens in which tissue amounts were limited, ER and PR analyses were performed with the highest priority, followed by other tissue assays. No differences were seen with respect to the availability of complete prognostic factor information between the groups.
No significant associations between prognostic factor status and treatment response, clinical or pathologic, were identified for the study group as a whole (Table 3). However, when evaluating individual treatment groups side-by-side, specific associations were identified in univariate analysis of cPOS rates (Fig. 1). ER-positive tumors appeared to demonstrate an increased rate of cPOS responses in the AC+D group compared with the AC group (91% vs. 56% respectively; P < 0.005; chi-square test). Similar findings were observed in PR-positive tumors, with cPOS rates of 87% for the AC+D group and 53% for the AC group (P < 0.005). However, when assessed with multivariate logistic regression analysis that controlled for tumor size and clinical lymph node status, these associations no longer were significant (P = 0.43 for ER-positive tumors; P = 0.65 for PR-positive tumors).
Table 3. Multivariate Analysis of Clinical and Pathologic Response Rates by Prognostic Factor for All Treatment Groups Combined
In the univariate analysis, HER-2/neu status also was associated with response to docetaxel, with HER-2/neu-negative tumors responding better with the addition of docetaxel (81% cPOS in the AC+D group; 51% cPOS in the AC group; P < 0.05) (Fig. 1). In contrast to the hormone receptor results, however, these associations remained statistically significant in the multivariate analysis (P < 0.05), resulting in an adjusted odds ratio of 3.5 (95% confidence interval, 1.2–13.0) in favor of docetaxel in patients with HER-2/neu-negative tumors (Table 4). When evaluating HER-2/neu-positive tumors, there were no observed differences in the cPOS rate between therapies (78% in the AD+C group vs. 75%, in the AC group; P = 0.79) (Fig. 1).
Table 4. Multivariate Analysis of Differences in Clinical Positive Response Rates between Treatment Arms by Prognostic Factor, Controlled for Clinical Tumor Size and Lymph Node Status
To investigate the effect of addition of docetaxel, response rates with respect to HER-2/neu status were compared within treatment arms (Fig. 2). In the AC group, the cPOS rate was decreased among patients who had HER-2/neu-negative tumors compared with patients who had HER-2/neu-positive tumors (51% vs. 75%), a finding that approached statistical significance (P = 0.06; chi-square test). In the AC+D group, however, responses were strong regardless of HER-2/neu status (78% positive vs. 81% negative; P = 0.99).
No associations were identified for cPOS rates between treatment arms based on p53 or Ki-67 in the univariate and multivariate analyses (Fig. 1). In addition, no associations were identified with respect to pCR for any of the prognostic factors (Fig. 3).
Using prognostic factor assessment to guide the selection of therapy has been explored with both adjuvant and neoadjuvant therapy regimens. Results have been discordant due to the heterogeneity of chemotherapy evaluated and the availability of comprehensive prognostic factor information. In our analysis, prognostic factors did not predict a general response to neoadjuvant therapy, but they did identify differential associations across the two arms of therapy.
Our most striking finding was the markedly improved response among patients with HER-2/neu-negative tumors with the addition of docetaxel. Patients with HER-2/neu-negative tumors who did not receive docetaxel had a significantly lower tumor response rate, whereas responses in the other 3 groups nearly were equivalent. These observations suggest that the differential response between HER-2/neu-positive and HER-2/neu-negative tumors who received AC alone appears to be obviated in tumors treated with docetaxel: Response rates in patients with HER-2/neu-negative tumors improved to approximate the rates seen in patients with HER-2/neu-positive tumors.
The impact of HER-2/neu on response to taxanes has been an interesting subject of investigation. Some in vitro studies have suggested that taxanes may exert less activity in tumors that overexpress HER-2/neu.18, 19 Supporting this finding, promising results have been observed in patients who received neoadjuvant trastuzumab in combination with a taxane,20, 21 although full evaluations from randomized studies are pending. One study that evaluated docetaxel alone did not establish a predictive value for HER-2/neu, although it did show high rates of pCRs in patients with HER-2/neu-negative tumors.22 Conversely, a retrospective analysis of patients who received neoadjuvant paclitaxel or docetaxel alone did not identify a correlation between HER-2/neu expression and pCRs, although clinical responses were not addressed in that study.23 However, our current patient sample differed from those studies in the use of anthracycline-based therapy, which may have its own implications on the effect of HER-2/neu to response.
Evidence suggests that HER-2/neu overexpression in anthracycline-based therapies may confer predictive information. In 3 clinical studies that incorporated neoadjuvant anthracyclines, either epirubicin or doxorubicin, HER-2/neu overexpression was associated with a better response to neoadjuvant therapy.6, 24, 25 Di Leo et al. have suggested that these correlations are indirect and are mediated by the observed high association between HER-2/neu overexpression and topoisomerase-IIα aberrations.26, 27 In making this suggestion, they provided a rationale for why HER-2/neu, which is not a known target of anthacyclines, appears to be relevant in therapies that incorporate topoisomerase-IIα inhibitors.
Given the strong evidence for improved response from anthracyclines in HER-2/neu-overexpressing tumors, some have advocated their preferential use in the adjuvant setting, relying on HER-2/neu as a true predictive marker.28 In the context of these collective findings on the role of HER-2/neu in taxane-based and anthracycline-based therapies, our data suggest that the addition of docetaxel to the standard neoadjuvant treatment (AC) may not exert an appreciable benefit on clinical responses in patients with HER-2/neu-overexpressing tumors, for which AC may be adequate therapy. However, when HER-2/neu expression is normal or at low overexpression, docetaxel may “rescue” the response, correcting for relatively decreased potency of AC in low HER-2/neu-expressing tumors. Our analytic approach prevents us from making this observation definitively but suggests that this should be tested in a prospective trial.
With respect to ER and PR status, univariate analysis seemed to suggest an improved response among patients with ER-positive and PR-positive tumors to docetaxel, but this was not supported after adjustment for other factors in multivariate analysis. Historically, positive ER status has been considered to impair response to adjuvant chemotherapy,29, 30 but the impact of ER positivity on response to taxanes has been an issue of some debate. In Cancer and Leukemia Group B Trial 9344, which evaluated the use of adjuvant paclitaxel in early breast carcinoma, Henderson et al. reported a lack of survival benefit in patients with ER-positive tumors when adding paclitaxel, a finding that they qualified as nonsignificant after adjusting for multiple comparisons.31 When it was evaluated in a small trial of neoadjuvant docetaxel alone, ER positivity was noncontributory to tumor response22; however, in the interim analysis of the National Surgical Adjuvant Breast and Bowel Project (NSABP) Protocol B-27, evaluating the addition of docetaxel to neoadjuvant therapy, Bear et al. found that both patients with ER-positive and patients with ER-negative tumors showed increased pathologic responses to docetaxel, with a higher rate of response among patients with ER-negative tumors.32 Finally, in NSABP Protocol B-28, which evaluated the addition of paclitaxel to an adjuvant regimen, Mamounas et al. noted no differences in disease-free survival, recurrence-free survival, or overall survival between patients with ER-positives and patients with ER-negative tumors.33 In our current study, we also observed that positive ER status was not associated with clinical response.
The proliferation rate, as measured by Ki-67, was not predictive in the current study, in contrast to the results reported in 3 previous studies that used a neoadjuvant anthracycline and that identified some predictive benefit. Stearns et al. observed better pathologic responses in tumors with higher baseline proliferative indices when paclitaxel was sequenced with doxorubicin.34 In a larger study, Mauriac et al. observed improved clinical response in tumors with Ki-67 > 40%.6 Finally, in a study of the use of epirubicin, higher Ki-67 rates were correlated with a positive response. No studies reported an attenuated response with low Ki-67 rates; although, in a study that employed docetaxel alone, Ki-67 had no correlation with response.22 Similarly, we did not find a significant correlation, although results from larger studies would be required to confirm this negative finding.
Finally, overexpression of p53 was not associated with response, consistent with the findings of other investigators.6, 25 One study that used DNA analysis correlated certain mutations with increased response to therapy using cyclophosphamide, epirubicin, and 5-fluorouracil.35 We did not perform this type of analysis and, thus, cannot comment on its applicability to our results. Based on results from neoadjuvant therapy studies, p53 expression, as evaluated in immunohistochemical analysis, does not appear to be a promising predictor of response to our regimens of neoadjuvant therapy.
Although it was not the direct objective of the current analysis, we affirmed the finding that the addition of docetaxel to AC and tamoxifen yielded better clinical responses to neoadjuvant therapy, consistent with the recent results of the NSABP B-27 study.32 From those results, Bear et al. reported that the addition of docetaxel increased the cCR rate, the overall clinical response rate, and the pCR rate. This finding has been supported by other studies that cited improved tumor regression with the use of docetaxel in the neoadjuvant setting.36, 37
Our analysis would have been strengthened by the observation of a relation between prognostic factor status and pathologic response rates that mirrored the clinical responses. Our limited sample size may have affected our likelihood of identifying such effects. The exclusion of patients with microscopic disease reduced the number of pathologic responses recorded; however, because the clinical significance of microscopic residual disease had not been elucidated fully, we used this more conservative definition. Pathologic response, by nature, is a more rigorous endpoint compared with clinical assessments of size, because it requires absolute destruction of tumor cells within the treatment period. Consequently, pathologic response rates always will lag behind clinical response rates. This has been demonstrated well in NASBP Protocol B-27, in which cCR rates ranged from 40% to 63.6%, depending on treatment, but cPR rates ranged only from 12.9% to 26.1%.32 Therefore, large sample sizes generally will be required to observe associations with pathologic responses similar to those observed with clinical responses.
Our analysis shares limitations with many of the previously published reports of investigations into the role of prognostic factors in predicting response to therapy. The most significant limitation, as noted above, was our sample size, which, although it was greater than most similar studies, may have affected our ability to detect true differences. Another limitation is that enrollment was not stratified by prognostic factor status and that the treatment trial was not designed originally with this purpose in mind, carrying potential biases in doing a post-hoc analysis. Finally, it is important to recognize that our analysis did not take into account information on longer term outcomes, like survival, and did not analyze outcomes with respect to treatment response. Instead, we chose to focus primarily on treatment response alone, which will affect surgical decision-making more directly. The clinical response findings that we have reported are most relevant in this context.
Despite these limitations, we believe that our data support the potential for HER-2/neu as a predictive marker in neoadjuvant chemotherapeutic regimens, particularly when considering the use of a taxane. Although response to many current standard chemotherapeutic agents will remain difficult to predict, continued exploration of targeted therapies, such as trastuzumab, and a better understanding of the molecular implications of prognostic factors can elevate specific prognostic markers into truly predictive factors that can guide our treatment decisions.