Answers are grounded in this paper and cite their sources.
It has been more than 10 years since the initial report of an association between amplification of erbB-2 (also known as HER-2/neu and ERBB2) and poor clinical outcome of patients with primary breast cancer (1). Dozens of papers involving thousands of patients have now been published on this topic, and the general conclusion is that erbB-2 abnormalities, either gene amplification or protein overexpression, are associated with worse prognosis of patients with lymph node-positive breast cancer, but the relationship for patients with lymph node-negative disease is weak, at best. Since most of these correlative studies were retrospective and many of the patients were treated before adjuvant therapy was widely administered to patients with lymph node-negative disease, several clinical researchers speculated that these associations were due, in part, to specific adjuvant therapies received by lymph node-positive patients. The first studies that examined the potential role of erbB-2 status for predicting response to adjuvant chemotherapy concentrated on regimens that contained CMF (cyclophosphamide, methotrexate, and 5-fluorouracil) (2,3). The general trend reported in these and subsequent studies is that patients whose tumors have little or no detectable levels of erbB-2 derive considerable benefit from CMF regimens, but patients whose tumors have amplified erbB-2 genes or overexpress erbB-2 protein do not benefit from these therapies. However, it must be stressed that these conclusions are based on retrospective analyses, and they have not been validated in prospective studies with sufficient statistical power to detect interactions between treatment and erbB-2 status. In 1994, Muss et al. (4) published, to our knowledge, the first analysis of an interaction between expression of erbB-2 and adjuvant therapy with doxorubicin-containing regimens. The finding that tumors with high expression of erbB-2 responded better to dose-intensive treatment with CAF (cyclophosphamide, doxorubicin [Adriamycin], and 5-fluorouracil) was contrary to the previous results with the CMF regimens and spurred international efforts to validate or refute the published data and to understand the biologic mechanisms that might explain such results. But such validation studies are difficult to perform, and analyses of interactions between treatments and potential predictive factors can be accomplished only within a randomized clinical trial. Without the protection of randomization, selection biases will result in confounding effects between treatments administered and potential predictive factors that cannot be overcome with the most sophisticated multivariate statistical analyses. A fact not often appreciated is that detection of an interaction effect requires many more patients and events (i.e., relapses or deaths) than detection of treatment effects when no interaction is present. It can take four times the number of failures to detect an interaction effect of the same magnitude as a treatment main effect when no interaction is present at the same level of type 1 and type 2 errors (significance level and power, respectively) (5). The net result is that randomized trials designed to detect treatment main effects have only limited statistical power to detect interaction effects. This problem is compounded if retrieval of archival tissue, required to perform assays of the biomarker under investigation, is incomplete. Large, multicenter, cooperative group trials provide the best opportunities for addressing these questions. In this issue of the Journal, results from retrospective analyses of randomized clinical trials conducted by the CALGB (Cancer and Leukemia Group B) (6) and the NSABP (National Surgical Adjuvant Breast and Bowel Project) (7) are reported regarding relationships between erbB-2 expression and response to doxorubicinbased adjuvant therapy administered to patients with lymph node-positive breast cancer. The CALGB study was designed as a validation of their previous published results (4). The original trial randomly assigned 1572 women to receive one of three regimens of adjuvant CAF to address questions about the efficacy of dose intensity (8). The initial retrospective study that generated the hypothesis about an association between erbB-2 expression and dose response to chemotherapy included 397 (25%) of these patients (4), and the validation study reported in this issue of the Journal (6) included an additional 595 patients (38%). It was disappointing that the hypothesized interaction between erbB-2 expression and dose of CAF was not statistically significant, although the trend was in the same direction as previously reported. Subsequent subset analyses that adjusted for an apparent failure of randomization did produce the hypothesized result. While this study suggests that patients whose tumors overexpress erbB-2 might benefit from dose-intensive CAF, the authors correctly concluded that the hypothesis should be further validated before clinical implementation. The NSABP study was a retrospective analysis of patients enrolled on protocol B-11, a trial originally designed to compare PF (L-phenylalanine mustard plus 5-fluorouracil) with PAF (L-phenylalanine mustard plus doxorubicin plus 5-fluorouracil). Unlike the CALGB study, this trial provided a comparison of a treatment regimen plus or minus doxorubicin and, therefore, a direct test of an interaction between doxorubicin and erbB-2 expression. It is remarkable that archival tissue was obtained for 638 (93.5%) of 682 eligible patients. The results clearly indicate that patients with erbB-2-negative tumors have the same clinical outcomes with or without doxorubicin. Patients with erbB-2- positive tumors who did not receive doxorubicin had a significantly worse prognosis, but the addition of doxorubicin improved clinical outcomes such that they were equivalent to those experienced by patients with erbB-2-negative tumors. Formal tests of interactions between doxorubicin and erbB-2 expression were statistically significant for disease-free survival (P = .02) and distant disease-free survival (P = .02) but not for overall survival (P = .15) or recurrence-free survival (P = .06). These two studies, together with other studies that have recently been presented at national meetings (9) and preclinical data (10), strongly suggest that the hypothesized interaction between erbB-2 expression and adjuvant doxorubicin therapy is real. If indeed this is the case, then perhaps the time has come to standardize our assays and scoring systems for determining erbB-2 status. Many years ago when estrogen receptor (ER) assays were first introduced, the scientific community agreed that a qualitycontrol procedure needed to be established to assure clinicians and patients that ER values would be comparable between laboratories. Today, we need to determine which technique is best for assessing erbB-2 status. Should we report gene amplification, protein expression, or messenger RNA levels? Should gene amplification be determined by Southern blot analysis, differential polymerase chain reaction (PCR), fluorescence in situ hybridization, or some other technique? Should erbB-2 protein expression be evaluated by immunohistochemistry, western blot analysis, enzyme-linked immunosorbent assay, or other techniques? Immunohistochemistry has been the mostly widely used technique, but do all antibodies give the same results? What scoring systems should be used? The primary erbB-2 assays used in each of the studies reported in this issue of the Journal were immunohistochemical, but the antibodies and the scoring systems were different. The CALGB used the CB11 monoclonal antibody and analyzed the results as a continuous variable. To display survival curves for patients with erbB-2-positive and erbB-2-negative tumors, erbB-2 positivity was defined as 50% or more cells showing positive expression. With this definition, 27% of the tumors were erbB-2 positive. In contrast, only 17% of the cases exhibited gene amplification as determined by differential PCR. The correlation between erbB-2 expression and amplification was statistically significant (P<.001) but only moderately strong (r 4 .51). In head-to-head analyses, erbB-2 expression was a better predictor of clinical outcome than was gene amplification. The NSABP used a cocktail of two antibodies, i.e., mAb-1 (a mouse monoclonal antibody) and pAb-1 (a polyclonal rabbit antiserum). Any staining was regarded as positive, yielding an erbB-2-positivity rate of 37.5%. In a recently reported study (9) from the Southwest Oncology Group (SWOG) that compared tamoxifen plus or minus CAF, erbB-2 status was determined by using both CB11 and mAb-1, separately, and each antibody produced a 16% overexpression rate. However, mAb-1 was a better predictor of response to therapy. Thus, there remains uncertainty about the best definition of erbB-2 overexpression. It appears that the proof of principle that response to doxorubicin- based adjuvant therapy depends on erbB-2 status has been achieved. But before we begin to routinely select adjuvant chemotherapy for patients with breast cancer on the basis of erbB-2 status, we need a consensus about the most reliable, most reproducible, and most predictive method to determine erbB-2 status.
No takes yet. Share an insight, caveat, or question.
Gary M. Clark (1998) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: