Prognostic interaction between expression of p53 and estrogen receptor in patients with node-negative breast cancer: results from IBCSG Trials VIII and IX

The designs of IBCSG Trials VIII and IX have been described in detail elsewhere [36, 37]. Briefly, from 1990 to 1999 in Trial VIII, 1,063 pre- and peri-menopausal women with node-negative early breast cancer were randomly assigned to endocrine therapy with 24 months of goserelin alone, six cycles of chemotherapy with classical cyclophosphamide, methotrexate and 5-fluorouracil (CMF) or a sequence of 6 cycles of CMF followed by 18 months of goserelin. Similarly, from 1988 to 1999 in Trial IX, 1,669 eligible postmenopausal women were randomly assigned endocrine therapy with 5 years of tamoxifen 20 mg daily or three cycles of CMF followed by tamoxifen to complete 5 years therapy. In each trial, randomization was stratified according to locally-determined ER status. Institutional review boards reviewed and approved the protocols, and informed consent was required according to the criteria established within the individual countries. Patient follow-up, vital status and date of any relapse or recurrence are recorded in the IBCSG database. Median follow-up from randomization in Trial VIII is 12 years [36] and in Trial IX is 13 years [37]. IBCSG Trials VIII and IX were conducted from 1988 to 1999. Ethical approval was obtained in participating countries according to national regulations. The IBCSG Independent Data Monitoring Committee reviewed the trials periodically. All patients included in the analysis provided consent to participate in the trials. The study was reviewed and approved by the IBCSG Biological Protocols Working Group.

Central review of immunohistochemical expression of ER in whole sections has been documented [38, 39]. The data presented in these published reports is incorporated in a cumulative IBCSG database, which has been used to classify ER status in the present study. Similarly, the central review assessment of tumor size, Bloom and Richardson grade [39], and peri-tumoral vascular invasion, as previously described [40], were used for the present study. HER2 was considered as positive if 3+ by IHC at central assessment or amplified by fluorescent in situ hybridisation (FISH) performed on the tissue microarrays (TMAs) (HER2: C17 ratio > 2.0). There is debate about the threshold at which tumors should be considered positive for ER [41, 42]. Following IBCSG practice [39] and the recommendations of the American Society of Clinical Oncology and the College of American Pathologists [43], and before any data analysis, we defined a cut-point to identify ER as present if at least 1% of cells showed ER staining on immunohistochemistry.

Available tissue blocks from 1,493 patients randomized to IBCSG Trials VIII (593 patients) and IX (900 patients) were sent from the IBCSG Pathology Office to the Cancer Research Program, Garvan Institute of Medical Research, Sydney, Australia for construction of tissue TMAs. The TMAs were produced using the MTA-1 Manual Tissue Arrayer and a 1.0 mm needle to biopsy tumor tissue identified by examination of hematoxylin and eosin (H&E)-stained slides from a standard histological block. Three representative cores were taken from each donor block and deposited in the recipient array block. Each array in Trial VIII comprised 108 (9 ± 12) cores representing about 32 patients. Arrays from Trial IX comprised 96 cores (8 ± 12) representing about 28 patients. An asymmetric template was employed for core orientation. Cores of renal tissue (± 6) were randomly placed within each array to act as orientation markers when scoring. Normal breast cores taken from reduction mammoplasties (± 6) were also placed on each array to allow comparison between immunohistochemical staining in morphological normal breast and invasive breast carcinoma. Details of numbers of patients randomized and those analyzed in the present study are given in the REMARK Diagram (Figure 1) and in Table 1. Briefly, after exclusion of ineligible patients and those from non-compliant institutions, TMAs were prepared from 1,220 patients (Trial VIII, 450; Trial IX, 770). Staining for p53 as described below was successfully performed on TMAs from 1,113 patients (Trial VIII, 417; Trial IX, 696). Comparison of patients assessed for p53 and those not so assessed is presented in Table 1. Centrally reviewed ER was unavailable for 21 patients (Trial VIII, 11; Trial IX, 10) leaving a study cohort for the present analysis of 1,092 patients (Trial VIII, 406; Trial IX, 686).

Four-μm sections were baked in an oven at 63°C for 2 hours followed by rehydration in graded ethanols, followed by water. Antigen retrieval was performed by boiling in a water bath for 30 minutes at pH 9.0 using Dako antigen retrieval solution (S2367, Dako, Denmark). All additional steps were performed on a Dako autostainer: p53 monoclonal antibody (clone DO-7 Dako, Denmark) was incubated at 1:200 dilution for 30 minutes at room temperature, following standard blocking procedures with 3% hydrogen peroxide for 5 minutes. Detection involved Envision labeled polymer-horseradish peroxidise (HRP) anti-mouse antibody (Dako, Denmark) which was added for 30 minutes at room temperature, followed by diaminobenzidine (DAB)+ chromogen (Dako, Denmark) for 10 minutes. Slides were then counterstained with haematoxylin, dehydrated and mounted. Isotype-matched non-specific immunoglobulin was substituted for the primary antibody as the negative control. A p53 positive breast carcinoma was used as the positive control.

All scoring was performed by an experienced breast pathologist blinded to all clinical and outcome information. Maximum nuclear staining > 10% (any intensity) was considered positive, as employed in most studies using this antigen. Data from the scoring assessment of p53 on the TMAs were entered into, the data handling program (Cansto) at the Garvan Institute of Medical Research, Sydney.

Blocks available for this study were those with residual tumor after routine pathology, central review and the conducting of previous translational research studies using whole tumor sections [39, 44, 45]. Thus the 1,113 patients with p53 results available for this project were more likely to be those with more advanced tumors than those not included. This is reflected in the significantly higher proportion in larger primary tumor size categories and with higher Bloom and Richardson grade in patients available from both trials. There was no significant difference in age distribution or the presence of ER and the trends to worse DFS and OS were not statistically significant (Table 1).

Data on p53 status were available in 1,113 patients and for centrally reviewed ER status in 1,092 of these. P53 positivity was more common among patients with absent ER expression (125/212, 59%) than among those expressing ER (171/880, 19%, P < 0.0001). In univariate analyses the overall slight negative impact of p53 expression on DFS was not significant (Figure 2A). In the presence of ER expression, the adverse effect of p53 staining was more marked while in the absence of ER, p53 expression was associated with better DFS (Figure 2B and 3).

A statistically significant interaction was present between p53 status and ER (P = 0.004), as shown in Table 2. The statistical significance of the interaction persisted for both DFS and OS in further models allowing for significant pathological variables, and for treatment allocation and its interaction with ER status (Table 2). We found no evidence of interaction between p53 expression and the relative efficacy of the various endocrine or chemo-endocrine therapies used in the trials (data not shown), indicating that in our studies p53 expression was not a predictive marker for the efficacy of adding CMF chemotherapy to endocrine adjuvant therapy. Martingale residuals reflected no violations of proportionality apart from the ER variable. Tumors lacking ER are known to exhibit higher initial failure rates [46].

The classification of ER presence (if at least 1% of tumor cells showed ER staining on IHC) was defined prior to any data analysis. Exploration of an alternative ER-positive cut-point based on staining of at least 10% of cells [41], resulted in reclassification of 45 patients with ER staining levels of 1 to 9%, and this materially reduced the clarity of the interaction with p53 status (DFS interaction hazard ratio (HR) 1.76, P = 0.04, OS interaction HR 1.68, P = 0.11).

Reflecting the effect of ER expression, a similar dichotomy of impact of p53 expression was seen when patients were divided into luminal and non-luminal subtypes based on IHC expression of ER, progesterone receptor and Ki67, as well as the detection of HER2 by IHC (3+) or FISH [47] (Figure 3).