- © 2003 by American Society of Clinical Oncology
HER2 Overexpression and Doxorubicin in Adjuvant Chemotherapy for Resectable Breast Cancer
- Angela Moliterni,
- Sylvie Ménard,
- Pinuccia Valagussa,
- Elia Biganzoli,
- Patrizia Boracchi,
- Andrea Balsari,
- Patrizia Casalini,
- Gorana Tomasic,
- Ettore Marubini,
- Silvana Pilotti and
- Gianni Bonadonna
- From the Medical Oncology Unit, the Molecular Targeting Unit, Department of Experimental Oncology, the Department of Pathology, Scientific Direction, Istituto Nazionale Tumori; Institute of Medical Statistics and Biometry; and Institute of Pathology, Università degli Studi di Milano, Milan, Italy.
- Address reprint requests to Sylvie Ménard, PhD, Molecular Targeting Unit, Department of Experimental Oncology, Istituto Nazionale Tumori, Via Venezian 1, 20133 Milano, Italy; email: menard{at}istitutotumori.mi.it.
Abstract
Purpose: Human epidermal growth factor receptor 2 (HER2) overexpression was found to predict a good response in breast carcinoma patients treated with doxorubicin (Adriamycin [ADM]). Evidence from our recent study indicates that node-positive patients respond to cyclophosphamide, methotrexate, and fluorouracil (CMF) regardless of HER2 status. We address the issue of whether therapy regimens including CMF and ADM versus CMF alone have the same therapeutic effect in patients with HER2+ and HER2− tumors in terms of relapse-free survival (RFS) and overall survival (OS).
Methods: Archival specimens of the primary tumors from 506 patients in a prospective clinical trial were stained with the anti-HER2 monoclonal antibody CB11. Originally, patients were randomly allocated to receive either 12 courses of intravenous CMF or eight courses of the same regimen followed by four cycles of ADM. RFS and OS were analyzed by a Cox model taking into account treatment, HER2 status, and the interaction between treatment and HER2 status, adjusting for the effect of other known clinical and biopathologic factors.
Results: Analysis of survival rates indicates a possible differential effect of treatment in the patients grouped according to HER2 status. Improved RFS and OS were observed in the HER2+ subgroup after treatment with CMF plus ADM versus CMF alone. With a median follow-up of 15 years, the hazard ratio (HR) for RFS was 0.83 in HER2+ tumors and 1.22 in HER2− tumors. The effect of treatment was more evident on OS in HER2+ patients (HR = 0.61; CI, 0.32 to 1.16) than in HER2− patients (HR = 1.26).
Conclusion: Our data indicate that adding ADM to CMF might be beneficial for patients with HER2+ tumors.
THERE IS considerable interest in the ability of biologic markers to predict the response of cancer patients to therapy. Human epidermal growth factor receptor 2 (HER2) overexpression has been shown to be a potential indicator of responsiveness to doxorubicin (Adriamycin [ADM]). Indeed, the study from the Cancer and Acute Leukemia Group B of node-positive patients randomly allocated to three dose levels of cyclophosphamide, doxorubicin, and fluorouracil (CAF) indicated a similar treatment outcome for all three dose levels in HER2− tumors, but showed a significant dose relationship in HER2+ tumors. The authors concluded that overexpression of this protein could be a useful marker in identifying those patients most likely to benefit from a high-dose CAF regimen.1–,3 In addition, Paik et al4 from the National Surgical Adjuvant Breast and Bowel Project examined the effect of doxorubicin in node-positive, hormone-receptor–negative breast cancer patients randomly allocated to receive l-phenylalanine mustard plus fluorouracil with or without the addition of the anthracycline. They reported that the administration of doxorubicin exhibited a statistically significant benefit in tumors that were HER2+, but no effect of the anthracycline was detected in HER2− tumors.
Recently, we have shown that HER2+ tumors, in opposition to indications by other studies, also benefited from cyclophosphamide, methotrexate, and fluorouracil (CMF) treatment.1 Indeed, in patients from the clinical trial including node-positive patients randomly assigned to receive either CMF for 12 cycles or no adjuvant treatment, CMF was found to induce a clinical benefit in both HER2+ and HER2− subgroups. The poor prognosis associated with HER2 overexpression in the untreated group could be completely overcome by the chemotherapy treatment.
In light of these previous studies, which indicate that HER2+ tumors respond to regimens including ADM as well as to CMF alone, the question arises whether these tumors are more responsive to one therapy than another. To address this issue, we analyzed the outcome of the disease in patients enrolled in the clinical trial, and examined the role of ADM given in addition to CMF, in which one arm was designed to receive CMF alone for 12 cycles and the other arm received CMF for eight cycles followed by ADM for four cycles. HER2 status was retrospectively analyzed by immunohistochemistry to evaluate its effect on disease-free survival and overall survival (OS). Notwithstanding the intrinsic limitations of the study design, the results indicated a potential benefit for patients with HER2+ tumors with the addition of ADM to the chemotherapy regimen.
METHODS
Patients
The study group consisted of patients enrolled between November 1981 and July 1990 in a prospective randomized trial carried out at the Istituto Nazionale Tumori in Milan, Italy.5 Women 70 years of age or younger who had had a radical mastectomy or a quadrantectomy plus full axillary dissection for unilateral carcinoma of the breast, and who had histologic evidence of involvement of one to three axillary nodes, were considered for inclusion in the study. The protocol design was approved by the members of the institute’s research and ethics committees.
After stratification by menopausal status, patients were randomly assigned to receive either CMF (cyclophosphamide 600 mg/m2 intravenously [IV] day 1; methotrexate 40 mg/m2 IV day 1; fluorouracil 600 mg/m2 IV day 1) for 12 cycles every 3 weeks or the CMF → ADM regimen, consisting of the same CMF schedule as described for eight cycles followed by ADM (75 mg/m2 every 3 weeks) for four cycles. No additional therapy was planned beyond that allowed in the protocol without documented evidence of treatment failure. In particular, no adjuvant endocrine therapy was administered and irradiation was limited to the remaining portions of the breast after quadrantectomy.
Before surgery, all patients underwent a complete physical examination, x-ray study of the chest and skeleton (skull, spine, pelvis, and upper third of femurs), bone scan, bilateral mammography, ECG, a differential blood count with platelet count, and biochemical tests. In the absence of symptoms, physical examination was performed every 3 weeks during chemotherapy treatment, then every 6 months for the first 5 years, and every 12 months thereafter. Biochemical tests, chest roentgenography, and bone roentgenography or bone scanning were performed every 6 to 8 months during the first 5 years and every 12 months thereafter. Mammography (bilateral in women subjected to breast-conserving surgery) was planned once a year. After the 15th year of follow-up, the patients were examined every 12 to 18 months. In patients with suspicious or controversial radiologic findings, examinations were performed more often. Liver ultrasonography was performed only if there were suspicious clinical or biochemical findings.
In determining relapse-free survival (RFS), treatment failure was considered to have occurred with the first documented evidence of new manifestations of disease in locoregional areas (including homolateral supraclavicular adenopathy), distant sites, the contralateral breast, or any combination of these sites. Among the total case series of 552 women with a median follow-up time of 178 months, 263 patients experienced new manifestation of disease: 68 in locoregional areas, 126 in distant sites, 33 in distant plus locoregional, and 36 in contralateral breast. Twenty-five observed second primary cancers were not considered treatment failures, and, for RFS estimation, follow-up time was censored at the date of the last visit if no cancer events occurred or at the occurrence of a second primary cancer.
OS was estimated by considering death from all causes. In a total of 198 deaths, 170 patients died of breast cancer and 28 patients died of other causes—10 of tumors in other sites and 18 of nonneoplastic causes. Causes of death were ascertained through medical records, death certificates, family doctors, and, when available, autopsy records.
Starting from the original case series, complete information on all the considered variables was available for 475 patients (242 in CMF and 233 in CMF → ADM arms); 506 cases had complete HER2 evaluations. The baseline characteristics of the study population (552 patients) and the selected subset were found to be similar, indicating no evident bias of selection. Indeed, both series consisted of about 65% of tumors smaller than 2.1 cm, and 69% (CMF) versus 67% (CMF → ADM) of patients younger than 51 years.
Immunohistochemistry
A panel of immunocytochemical stains was performed on paraffin-embedded tissue. The following panel of monoclonal antibodies was applied: anti–c-erbB-2 CB11 (1:100 diluted, Ylem, Avezzano, Italy), anti-p53 MAb DO7 (1:500 diluted, Novocastra, Newcastle upon Tyne, United Kingdom), antiestrogen receptor (ER; clone 1D5; 1:200 diluted, DBA, Segrate, Milan, Italy), antiprogesterone receptor MAb1A6 (1:100 diluted, DBA, Segrate). Immunostaining was performed by a sensitive peroxidase-streptavidin method on Bouin-fixed, paraffin-embedded material as described.1 Immunostaining was performed using an automated immunostainer (TechMate 1000, Dako, Copenhagen, Denmark). As a negative control, the primary antibody was replaced with a nonimmune serum from the same species in which the primary antibody was produced. Appropriate cases with known reactivity for each antibody applied were used as positive controls. Sections were scored positive when more than approximately 10% of the tumor cells were labeled, except for HER2, which was scored as positive when strong membrane labeling was observed. With this scoring system, Bouin-fixed, CB11+ cases were found to correspond to formalin-fixed tumors scoring 3+ with HercepTest (Dako)6 as evaluated on the same primary breast carcinomas.
Statistical Analysis
To evaluate marginal treatment effects on RFS and OS according to HER2 status, RFS and OS probability curves for the subgroups of patients identified by treatment and HER2 status were estimated by means of the Kaplan-Meier method.
The differential effect of treatment according to HER2 status was formally tested by means of a semiparametric Cox model including treatment, HER2 status, and the interaction treatment × HER2 status term. Moreover, other recognized prognostic factors were included in the model to provide adjusted estimates, which were categorized according to conventional clinical criteria, including tumor size (≤ 2 cm/> 2 cm), age (≤ 50 years/50 years), estrogen receptors (−/+), progesterone receptors (−/+), and p53 (−/+). The proportional hazard (PH) assumption was investigated by means of Schoenfeld residual plots7 and tested according to the method proposed by Grambsch and Therneau.8 If PH was not tenable, a stratified analysis was conducted on the basis of the covariates that led to PH violation. Hazard ratios (HRs) and corresponding 95% confidence intervals (95% CI) were estimated from the regression models.
Retrospective analyses of subgroups or interactions can be explored without affecting the original trial’s conclusions. However, the credibility of these analyses is improved if they are confined to a few predefined effects on the basis of biologically plausible hypotheses9 or if specific statistical procedures are adopted in the absence of prior knowledge.1 The true statistical power for detecting the effects of interest, because the study was not originally planned to account for them, can be called into question. In this study, prior information was available from literature supporting treatment by HER2 status interaction effect, but no similar evidence was available for the other covariates, and therefore additive effects were considered for the latter.
To verify the robustness of the results on the interaction with respect to standard assumptions underlying conventional hypothesis testing, a bootstrap procedure was applied in 200 samples. The significance level of the interaction effect has been reassessed on the basis of the bootstrap inclusion fraction of the interaction term into the model; for each probability value obtained by testing the null hypothesis of no interaction, an expected cutoff on bootstrap inclusion fraction was calculated and compared with the observed inclusion fraction. For example, a significance level of 5% in the original data can be referred to a cutoff value of 50% for the bootstrap inclusion fraction using a selection level of 5% in each replication.10 Statistical analysis was performed by S-PLUS (Insightful, Seattle, WA). The libraries written by Harrel et al11 were applied for some of the steps of the model-building procedure.
RESULTS
Table 1⇓ reports the distribution of the considered tumor markers according to the treatment. The observed and missing values are evenly divided among the treatment arms. Moreover, patients with missing data are randomly distributed over the recruitment period (data not shown).
Tumor Marker Values of Patients Enrolled in the Trial According to the Treatment
The RFS and the OS curves for the combination of treatment and HER2 status (CMF/HER2−: 208 patients, 94 relapses, 65 deaths; CMF → ADM/HER2−: 203 patients, 104 relapses, 77 deaths; CMF/HER2+: 50 patients, 28 relapses, 25 deaths; CMF → ADM/HER2+: 45 patients, 20 relapses, 17 deaths from breast cancer) are provided in Figs 1⇓ and 2⇓. The estimated RFS represented in Fig 1⇓ suggests a possible differential effect of treatment in the subgroups of patients according to HER2 status. For HER2− patients, CMF followed by ADM showed a slightly worse outcome, especially after 5 years of follow-up. RFS probability at 10 years for CMF/HER2− was 0.59; 95% confidence limit (CL), −0.66 to −0.52; for CMF → ADM/HER2−, it was 0.56; 95% CL, −0.62 to −0.49. For HER2+ patients, the effect seemed to be reversed. RFS probability at 10 years for CMF/HER2+ was 0.46; 95% CL, −0.59 to −0.32; for CMF → ADM/HER2+ it was 0.55; 95% CL, −0.69 to −0.40). Regarding OS (Fig 2⇓), a difference between the two treatments was less evident in HER2− patients, although it was consistent with that observed for RFS (OS probability at 10 years for CMF/HER2−: 0.77; 95% CL, −0.82 to −0.71; for CMF → ADM/HER2−, 0.76; 95% CL, −0.82 to −0.70). A benefit of CMF followed by ADM compared with CMF alone was much more evident with respect to RFS in the HER2+ group (OS probability at 10 years for CMF/HER2+: 0.54; 95% CL, −0.67 to −0.39; for CMF → ADM/HER2+: 0.64; 95% CL, −0.76 to −0.49).
Relapse-free survival of breast carcinoma patients according to treatment and HER2 status.
Overall survival of breast carcinoma patients according to treatment and HER2 status.
The results of Cox multiple regression models for RFS are reported in Table 2⇓. Because a marked violation of the proportional hazard assumption was apparent for progesterone receptors only, the model was stratified for this variable. The presence of the treatment-HER2 interaction effect was not supported (P = .251). However, consistent with the estimated RFS probability curves, the estimated HRs for the treatment within HER2 subgroups were 0.83 (95% CI, 0.46 to 1.49) in HER2+ patients and 1.22 (95% CI, 0.91 to 1.64) in HER2− patients.
Cox Model Results for Relapse Free Survival and Overall Survival
Cox multiple regression model results for OS are reported in Table 2⇑. Neither the formal tests on each independent variable nor Schoenfeld residual plots show time-dependent effects. Therefore, no stratification was adopted in the model. Evidence for the interaction effect (P = .052) was much more relevant for OS than for RFS. The estimated HRs for the treatment within HER2 subgroups were 0.61 (95% CI, 0.32 to 1.16) in HER2+ patients and 1.26 (95% CI, 0.89 to 1.79) in HER2− patients, indicating no difference in treatment effects for HER2− patients, whereas HER2+ patients treated with CMF → ADM seemed to show improvement in OS in comparison to those treated with CMF alone.
The bootstrap procedure confirmed the significance levels of the conventional statistical tests on the interaction for both end points. In fact, the interaction seemed to be significant in 23% and 47.5% of the bootstrap samples for RFS and OS, respectively. These values are comparable with the expected inclusion fractions of 21% and 49%, corresponding to the original respective P values of .251 and .052.
DISCUSSION
These results suggest a therapeutic benefit of ADM treatment in patients with HER2+ breast carcinoma. This observation further confirms the conclusions of other reports suggesting that ADM activity is restricted to the subset of breast carcinomas overexpressing the HER2 oncogene.2–,4 The responsiveness of HER2+ tumors to CMF has been clearly shown in our study1 and those of others,12,13 although in some studies no significant benefit from CMF-based chemotherapy was observed.14–,17 This report indicates that the subset of HER2+ tumors seems to be more responsive to ADM plus CMF than to CMF alone. This advantage was not observed in patients with HER2− tumors. In this subset, ADM treatment did not increase the therapeutic benefit achieved with CMF alone.
Our data are in keeping with previous reports indicating the advantage of regimens containing ADM in patients with HER2+ tumors, whereas either an ADM or CMF regimen may be considered for patients with HER2− tumors.4 The association between HER2 positivity and response to ADM given in an adjuvant setting has been well documented in another large series.2,3 It is noteworthy that in metastatic disease, no direct association between HER2 overexpression and response to chemotherapy has been observed to date,18–,20 although a trend toward the chemoresistance of HER2+ tumors has been reported.21 In addition, in locally advanced disease, Geisler et al22 reported that expression of HER2 predicted chemoresistance.
These data suggest a different role of HER2 during tumor progression and indicate the importance of clarifying the biomolecular mechanism at the base of a possible association between HER2 and response to chemotherapy in an adjuvant setting.
It is noteworthy that the Cox model shows a statistically significant interaction for OS but not for RFS, the opposite of what was expected considering the number of events. This suggests that ADM treatment affects more the aggressiveness of the recurrences than their incidence, as expected if the drug (which is active on cells with high HER2 overexpression) selects tumor cells with lower HER2 expression, which are less aggressive.
As in previous studies, these results come from a controlled clinical trial that was not originally designed to address the issue of prediction of response to therapy by biomolecular markers, which have been retrospectively determined. In addition to study design, the conclusions suffer because of the limited number of HER2+ cases represented in this series—17% of the tumors. This frequency is somewhat lower than that reported by others, but it is consistent with the frequency observed in our institution in many other studies, even with other reagents,14,23 indicating that, with archival specimens, some HER2 positivity may be lost and that only the strongly overexpressing tumors maintain a positive score. Indeed, when analyzed in parallel on the same tumors, the CB11+ cases were found to correspond to the cases scoring 3+ with the standardized method used today to evaluate to HER2 status, indicating that cases with intermediate HercepTest staining were lost. The conclusions of the study are restricted, therefore, to the group of tumors with a high level of overexpression corresponding to the cases selected for anti-HER2 immunotherapy.24
The overall result of the clinical trial was that no additional therapeutic benefit can be obtained by addition of ADM to CMF regimen. This conclusion does not seem to be applicable to the HER2+ tumor subset. Indeed, subsets of low frequency such as HER2+ tumors do not add enough weight to the final results to change the results on the overall series. Our previous data identify the HER2+ breast carcinoma as a particular group of tumors with peculiar clinical and pathologic behaviors.6 Overall, these observations suggest the need for re-evaluation of conclusions from clinical trials on which clinical management of breast carcinoma is based, inasmuch as some of these conclusions may not be applicable to HER2+ tumors.
Footnotes
-
Supported in part by the Associazione Italiana Ricerca sul Cancro, Milan, Italy.
- Received April 2, 2002.
- Accepted October 15, 2002.
