HER2 expression and efficacy of dose-dense anthracycline-containing adjuvant chemotherapy in breast cancer patients

No data are available on the role of HER2 overexpression in predicting the efficacy of dose-dense anthracycline-containing adjuvant chemotherapy in breast cancer patients. We retrospectively evaluated this role in patients enrolled in a phase III study comparing standard FEC21 (5-fluorouracil, epirubicin, and cyclophosphamide, administered every 3 weeks) vs dose-dense FEC14 (the same regimen repeated every 2 weeks). HER2 status was determined for 731 of 1214 patients. Statistical analyses were performed to test for interaction between treatment and HER2 status with respect to event-free survival (EFS) and overall survival (OS); EFS and OS were compared within each HER2 subgroup and within each treatment arm. Median follow-up was 6.7 years. Among FEC21-treated patients, both EFS (HR=2.07; 95% CI 1.27–3.38) and OS (HR=2.47; 95% CI 1.34–4.57) were significantly worse in HER2 + patients than in HER2 − patients. Among FEC14-treated patients, differences in either EFS (HR=1.21; 95% CI 0.65–2.24) or OS (HR=1.85; 95% CI 0.88–3.89) between HER2 + and HER2 − patients were not statistically significant. Interaction analysis suggested that the use of dose-dense FEC14 might remove the negative prognostic effect of HER2 overexpression on EFS and OS. Our data suggest a potential role of HER-2 overexpression in predicting the efficacy of dose-dense epirubicin-containing chemotherapy and the need to confirm this hypothesis in future prospective studies.

HER-2 overexpression or amplification has been widely studied as a prognostic and predictive factor in early breast cancer patients but its role is still controversial (Trock et al, 2000;Ravdin, 2001;Sledge, 2001;Yamauchi et al, 2001;Goldhirsch et al, 2003). In most studies, overexpression or amplification of the HER2 gene has been associated with an adverse clinical outcome (Schnitt, 2001). With regard to HER2 as a predictive factor, retrospective studies have suggested that the benefit from anthracycline-containing chemotherapy might be greater in women whose tumour overexpresses HER2 (Paik et al, 1998(Paik et al, , 2000Di Leo et al, 2002). Another important issue is to evaluate if HER2 overexpression is associated not only with sensitivity to anthracyclines but also to their doseintensity and dose-density effect. Results from a retrospective study suggest that regimens with an increased dose per cycle (dose intensity) of anthracyclines (Thor et al, 1998) are associated with an improved outcome only in women with HER2 overexpression. Recent data showed that anthracycline-containing chemotherapy with increased dose density (obtained by the administration of the drugs with a shortened interval), but with the same dose per cycle and total dose, improves clinical outcome as compared to conventionally scheduled (every 3 weeks) regimens (Citron et al, 2003;Venturini et al, 2003). However, no data are available on the potential role of HER2 overexpression in predicting the efficacy of dose-dense anthracycline-containing chemotherapy. The aim of this study was to evaluate the prognostic and predictive role of HER2 overexpression in early breast cancer patients enrolled in a trial that compared standard vs dose-dense epirubicin-containing adjuvant chemotherapy.

PATIENTS AND METHODS
The patients evaluated in the present study were a subgroup of the study population entered into a prospective clinical trial. HER2 status was centrally evaluated in primary breast cancer samples from patients enrolled in a phase III multicentre study comparing standard vs dose-dense adjuvant chemotherapy (GONO-MIG-1 study, Gruppo Oncologico Nord-Ovest-Mammella Intergruppo). The GONO-MIG-1 study was conducted in 22 Italian centres enrolling 1214 patients from 1992 to 1996. Women (age p70 years) with histologically confirmed breast cancer who had undergone radical mastectomy or breast-conserving surgery plus full ipsilateral axillary node dissection were eligible for enrolment in the study if they had no more than 10 involved axillary nodes or were node negative but had a high risk of recurrence. High risk was defined as the presence of one or more of the following characteristics: age p35 years, negative oestrogen (ER) and progesterone receptor (PgR) status, tumour size X2 cm, poor histological grade, high proliferative rate determined by [ 3 H]thymidine labelling index, or by S-phase determination by flow cytometry. Patients were randomly assigned to receive either six courses of FEC21 (5-fluorouracil 600 mg m À2 , epirubicin 60 mg m À2 , and cyclophosphamide 600 mg m À2 intravenously on day 1, repeated every 3 weeks) or six courses of dose-dense FEC14 (the same drugs at the same doses of FEC21, repeated every 2 weeks), with the support of filgrastim, a granulocyte colonystimulating factor (G-CSF). Granulocyte colony-stimulating factor was subcutaneously self-administered by patients, at a dose of 5 mg kg À1 day À1 , on days 4 -11. Patients with ER-and/or PgRpositive tumours received tamoxifen 20 mg day À1 for 5 years. Postoperative regional radiotherapy limited to the remaining breast was given to patients who had received conservative surgery.

Tumour sample collection
All 22 participating centres were invited to participate in this study on the role of HER2 status (HER2 study) and 18 centres accepted. Each centre was provided with a list of all patients entered into the clinical study. Each centre sent one paraffin-embedded sample of the primary tumour to the Pathology unit of the National Cancer Research Institute of Genoa, Italy. Once samples were received, they were classified and stored at room temperature until the HER2 analysis was carried out.

HER2 evaluation
Immunohistochemical analysis was performed using monoclonal antibody CB-11 (Biogenex, San Ramon, CA, USA). Sections (3-mmthick) from formalin-fixed, paraffin-embedded tissue were cut and mounted on positively charged slides. Tissue sections were deparaffinised and rehydrated in graded alcohol. No antigen retrieval procedure was used. Slides were placed in TBS (0.05 M Tris/HCl; 0.15 M NaCl; pH 7.6) and endogenous peroxidase activity was blocked by a 5 min treatment with 3% hydrogen peroxide solution. Blocking solution was applied for 20 min, followed by incubation with the primary antibody (1 : 10 dilution TBS) for 30 min. Tissue sections received a 5 min TBS rinse before the application of biotinylate secondary antibody (1 : 200 dilution TBS) for 30 min and then a second 5 min TBS rinse. Antibody was localised using the streptavidin biotin immunoperoxidase (Dako, Dakocytomation, Milan, Italy; 1 : 10 dilution TBS), and 3 0 -3 diaminobenzidine was used to visualise the chromogen. Slides were counterstained in Mayer haematoxylin, dehydrated, and mounted. For each run, a composite slide of three formalin-fixed human breast carcinoma cell lines representing different levels of HER2 protein expression (MDA-231 ¼ 0; MDA-175 ¼ 1 þ ; and SKBR ¼ 3 þ ) was used as control. In addition, an overexpressing HER2 tissue section of breast cancer was used as positive and negative control. Negative control was made by substituting the HER2 primary antibody with normal rabbit serum. When both carcinoma in situ and invasive carcinoma were present in the same section, only the invasive component was scored. For the determination of HER2 protein overexpression, only the membrane staining intensity and pattern were evaluated as follows: no staining or membrane staining in less than 10% of the tumour: score 0; a faint/barely perceptible membrane staining in more than 10% of the tumour cells, stained only in part of their membrane: score 1 þ ; a weak to moderate complete membrane staining in more than 10% of the tumour cells: score 2 þ ; a strong complete membrane staining in more than 10% of tumour cells: score 3 þ .
In this study, patients were considered HER2 positive (HER2 þ ), that is, with overexpression of HER2, if it was scored as 3 þ ; patients with HER2 scored as 0, 1 þ , 2 þ were considered HER2 negative (HER2À). The cutoff of 3 þ was chosen before inspection of clinical results in order to consider as positive only tumours with unequivocal HER2 overexpression (Zarbo and Hammond, 2003).
The same pathologist, who was blinded to both treatment assignment and clinical outcome, scored all slides.

Statistical methods
Overall survival (OS) was estimated from the date of randomisation to the date of last contact or death from any cause. Event-free survival (EFS) was defined as the interval that had elapsed between the date of randomisation and the date of local relapse, distant relapse, second breast primary, or death from any cause, whichever occurred first.
All selected patients, regardless of eligibility or of compliance to the assigned treatment, were considered in the arm they were assigned at randomisation in the original study and there were no further exclusions.
The prognostic role of HER2 status on EFS and OS was assessed by fitting a multivariate proportional hazard model to the data. The following variables were initially included in the model:

RESULTS
Results regarding all patients (n ¼ 1214) enrolled in the clinical trial GONO-MIG1 have been recently presented (Venturini et al, 2003). At a median follow-up of 6.7 years, a 19% reduction in the hazard of death was observed in favour of the FEC14 arm.

Collection of tumour samples
Four out of 22 centres involved in the GONO-MIG1 study, which had enrolled 164 cases, chose not to participate in the HER2 study.
The 18 centres participating in the HER2 study enrolled 1050 patients in the GONO-MIG1 study.
A total of 731 paraffin-embedded samples of primary tumours suitable for HER2 analysis were collected between December 1998 and February 2000, corresponding to 60% of the overall trial population (1214 patients) and to 70% of the patients recruited in centres participating in this study (1050 patients). Reasons for the lack of HER2 evaluation in 319 cases were the following: tumour specimen not available (294 cases); tumour specimen inadequate (23 cases); unknown (two cases).

Study population
To explore if the subgroup of patients evaluated in the HER2 study was representative of the whole population entered in the clinical trial, the main patient and tumour characteristics in this subgroup (731 cases) were compared with those of the subgroup without HER2 assessment (483 cases). No major difference was observed between the two subgroups (Table 1). Among the 731 patients evaluated in this study, no differences were observed between the two treatment arms. Tumour characteristics and chemotherapy regimen by HER2 status are shown in Table 2. HER2 was overexpressed, that is, 3 þ (HER2 þ ), in 13.5% of FEC14-treated patients and in 14.7% of FEC21-treated patients. A negative hormone receptor status (i.e. both ER and PgR negative) was observed more frequently in tumours overexpressing HER2 (62%) than in HER2-negative (32.5%) tumours.
A contralateral breast cancer occurred in 13 cases (three cases in HER2 þ patients and 10 cases in HER2À patients), and in two of these cases, it was concurrent with loco-regional relapse. The results of disease-free survival, where the 11 cases of contralateral breast cancer as first event were excluded, closely mirrored those of EFS analyses (data not shown).
Similarly, when the outcomes of patients assigned to FEC14 were compared to those of patients assigned to FEC21, within the subgroup of HER2À patients (n ¼ 628), no difference was seen in EFS (HR ¼ 0.91; 95% CI 0.65 -1.27; P ¼ 0.57), whereas, among the smaller subgroup of HER2 þ patients (n ¼ 103), a reduction in the rate of events was seen in patients assigned to FEC14 (HR ¼ 0.54; 95% CI 0.27 -1.11; P ¼ 0.092), although the difference did not reach statistical significance. When analysing OS, the difference between the effect of the experimental treatment in HER2À and HER2 þ patients was less marked (HR ¼ 0.79, 95% CI 0.49 -1.28, P ¼ 0.34 and HR ¼ 0.59, 95% CI 0.26 -1.37, P ¼ 0.22, in HER2À and HER2 þ patients, respectively). Figure 3 shows Kaplan -Meier  When the potential modifying role of HER2 status on the effect of the experimental treatment was formally assessed by introducing the appropriate interaction term in the two multivariate models (one for EFS, the other for OS), the results were suggestive but not statistically significant for EFS (P for interaction ¼ 0.12), and negative for OS (P for interaction ¼ 0.379).
In both models, however, the estimated coefficients suggested that the use of dose-dense adjuvant chemotherapy might remove the negative prognostic effect of HER2 overexpression on EFS and partially contrasted that on OS (Table 5).

DISCUSSION
We evaluated both the prognostic role of HER2 overexpression and its potential role as a modifier of the effect of adjuvant therapy in a population of early breast cancer patients enrolled in a phase III study comparing two epirubicin-containing regimens with different dose densities. In the overall population, regardless of the FEC regimen received (dose dense or standard), HER2 overexpression was associated with a poor prognosis. Both EFS (HR ¼ 1.64; 95% CI 1.12 -2.40) and OS (HR ¼ 2.2; 95% CI 1.37 -3.53) were significantly shorter in HER2 þ patients than in HER2À patients. Because all patients in our study were treated with FEC chemotherapy, our results indicate that HER2 overexpression retains its prognostic role also in patients receiving anthracyclinecontaining chemotherapy.
The role of HER2 in predicting sensitivity to anthracyclines is still under investigation. Clinical data suggest that women whose tumours overexpress HER2 might derive greater benefit from anthracycline-based than from alkylating agent-based adjuvant  (Paik et al, 1998(Paik et al, , 2000Di Leo et al, 2002). Another clinically important issue is whether HER2 overexpression or amplification is predictive of the benefit deriving from the use of anthracyclines administered at a dose intensity and/or dose density higher than the standard dose. Two retrospective studies addressed the issue of the potential association between HER2 and the increased dose intensity of anthracycline-containing chemotherapy. In both studies, the increase in dose intensity was obtained by increasing the single dose per cycle while intervals between cycles were not reduced, that is, the dose density was not increased (Thor et al, 1998;Di Leo et al, 2002). Results from CALGB 8541 study, in which patients were randomised to three different dose levels of CAF, suggest that the benefit from the higher dose-intensity CAF was confined to women with HER2 overexpression. However, the dose level defined as higher is what we now consider the standard dose, that is, 60 mg m À2 of doxorubicin. The study from Piccart et al (2001) compared two regimens of epirubicin at different dose intensities (60 vs 100 mg m À2 per cycle) with CMF (Piccart et al, 2001). A retrospective analysis carried out in 55% of the clinical trial population showed no difference in EFS between high dose-intensity and standard dose-intensity epirubicin in the small group of HER2-amplified patients. The results of the two above-cited studies indicate that the potential association between HER2 overexpression or amplification and anthracycline dose-intensity benefit remains to be clarified. This study, to the best of our knowledge, is the first to investigate the potential role of HER2 in predicting the efficacy of a dose-dense epirubicin-containing adjuvant chemotherapy. In our study, patients received the same drug doses per cycle and the same total dose in the two arms, and the only difference between the two arms was the interval between cycles, that is, 2-week (dose dense) vs 3-week (standard) schedule. Our results suggest that dose-dense FEC14 may be superior to FEC21 in HER2 þ patients only, even though differences in outcome did not reach statistical significance. In the HER2 þ cohort, the relative risks of failure for FEC14-treated patients as compared to FEC21 were 0.54 (95% CI 0.27 -1.11; P ¼ 0.092) for EFS and 0.59 (95% CI 0.26 -1.37; P ¼ 0.22) for survival. In HER2 þ patients, treatment with FEC14 was associated with 15.2% (95% CI À3 -33.4%) and 14.8% (95% CI 0.4 -29.2%) absolute increase in 5-year EFS and OS, respectively, as compared to HER2 þ patients receiving FEC21. Among patients treated with FEC21 regimen, both EFS and OS were shorter in HER2 þ patients than in HER2À patients, while when patients were treated with dose-dense FEC14, differences in outcome between HER2 þ and HER2À patients were not statistically significant.
The potential higher efficacy of dose-dense FEC14 treatment in HER2 þ patients than in HER2À patients may be biologically explained by data suggesting that HER2 overexpression confers a high proliferative capability to the tumour (Borg et al, 1991) and is associated with an amplification of topoisomerase II alpha gene, a potential marker of anthracycline sensitivity (Di Leo et al, 2002). Because the regrowth of cancer cells between cycles of cytoreduction is likely to be more rapid in HER2 þ tumours, the more frequent administration of anthracycline-containing chemotherapy could be particularly effective in these anthracycline-sensitive tumours.
These results suggest a potential role of HER2 overexpression in predicting the efficacy of dose-dense epirubicin-containing adjuvant chemotherapy, but they must be considered with caution because of some weaknesses of the study. The first one is related to its retrospective nature. Even though no remarkable difference was seen between patients included and those not included in the present study, the possibility of a selection bias cannot be ruled out. The second limitation is due to its statistical power, and specifically to the small number of HER2-positive patients. As a matter of fact, with 170 events and 92 deaths, the study had limited power (o30%) in detecting plausible differences (HR between 1.25 and 1.33) in the risk of relapse and of death, particularly in    HER2 and dose-density chemotherapy L Del Mastro et al interaction analyses. As a consequence, some of the observed differences fail to achieve statistical significance. A third possible limitation is the absence of FISH testing. When our study started, at the end of 1998, FISH testing on paraffin-embedded specimens was not widely available. Then, we performed our analysis using immunohistochemistry. At that time, CB11 was among the most commonly primary reagents used to determine HER2 status. Moreover, the hypothesis of an interaction between anthracycline dose intensity and HER2 overexpression was tested just by using CB11 (Thor et al, 1998). Eventually, as expected, HER2 overexpression was inversely associated with the expression of hormone receptors (Table 2), leading to a more frequent use of endocrine therapy in HER2negative patients than in HER2-positive patients. This different use of endocrine therapy may potentially confound the outcome.
However, our results were obtained from multivariate analysis where the potential confounding effect of hormonal status and the consequent endocrine therapy was accounted for.
Owing to these limitations, our results should be considered exploratory and need to be confirmed in prospective, well-powered studies.