One year of adjuvant tamoxifen compared with chemotherapy and tamoxifen in postmenopausal patients with stage II breast cancer
- Bent Ejlertsena, b, , ,
- Maj-Britt Jensena,
- Johanna Elversangc,
- Birgitte B. Rasmussenc,
- Michael Anderssonb,
- Jørn Andersend,
- Dorte L. Nielsene,
- Søren Coldf,
- Henning T. Mouridsena
- a Danish Breast Cancer Cooperative Group (DBCG), Rigshospitalet, Copenhagen, Denmark
- b Department of Oncology, Rigshospitalet, Copenhagen, Denmark
- c Department of Pathology, Herlev University Hospital, Herlev, Denmark
- d Department of Oncology, Aarhus Hospital, Aarhus University Hospital, Aarhus, Denmark
- e Department of Oncology, Herlev University Hospital, Herlev, Denmark
- f Department of Oncology, Odense University Hospital, Odense, Denmark
- http://dx.doi.org/10.1016/j.ejca.2013.05.006, How to Cite or Link Using DOI
We report the long-term results of a randomised trial comparing tamoxifen with tamoxifen plus cyclophosphamide, methotrexate and fluorouracil (CMF) in postmenopausal high-risk breast cancer patients. In addition, we analyse the prognostic and predictive value of centrally assessed subtypes.
Postmenopausal patients with breast cancer and positive nodes, deep invasion or size exceeding 5 cm were randomly assigned to 1 year of tamoxifen, or cyclophosphamide 600 mg/m2, methotrexate 40 mg/m2 and fluorouracil 600 mg/m2 intravenously on day 1 every 4 weeks for nine cycles plus tamoxifen (CMFT). Tissue microarrays were constructed retrospectively and oestrogen receptor (ER), progesterone receptor (PgR), human epidermal growth factor receptor 2 (HER2), and proliferation-related Ki-67 antigen (Ki67) status were assessed.
From October 1982 through March 1990 we randomised 1445 patients and 969 (67%) were eligible for the biomarker analysis. At 10-years 936 women had suffered a disease-free survival (DFS) event (tamoxifen, 495 events in 686 patients; CMFT, 441 events in 642 patients). The addition of CMF to tamoxifen significantly improved DFS (adjusted hazard ratio 0.82; 95% confidence interval (CI) 0.71–0.93; P = 0.003) but not overall survival (adjusted hazard ratio 0.95; 95% CI 0.85–1.08; P = 0.44). DFS was superior in Luminal A tumours (ER or PgR positive, HER2 negative and Ki67 ⩽14%) when compared to Luminal B or non-luminal (ER and PgR negative) tumours. There was no statistical evidence of heterogeneity by subtype in the benefit from CMF (Pinteraction = 0.45).
CMF added to 1 year of tamoxifen reduces the risk of a DFS event. The benefit from CMF was not significantly different in Luminal A and B subtypes.
- Adjuvant therapy;
- Breast cancer;
Uncertainty persists concerning the incremental benefit of chemotherapy in postmenopausal, hormone receptor positive patients given tamoxifen. Individual trials have reported a beneficial effect on disease-free survival (DFS) but not on overall survival (OS) from adding chemotherapy to tamoxifen in postmenopausal patients with node-positive breast cancer.1, 2, 3 and 4 When combined in the meta-analyses performed by the Early Breast Cancer Trialists’ Collaborative Group (EBCTCG) chemotherapy on average improved disease-free survival and reduced mortality in patients with operable breast cancer.5 In addition, the EBCTCG overview showed the benefit to be independent of nodal status, menopausal status (given age), oestrogen receptor (ER) status, and whether or not tamoxifen had been given. The DBCG 82C trial randomised postmenopausal women with early high-risk breast cancer to tamoxifen (TAM), TAM with radiotherapy (RT) or TAM with cyclophosphamide, methotrexate, fluorouracil (CMF). The results of TAM compared with TAM + RT have previously been published.6, 7 and 8 The predictive value of ER was not firmly established in 1982, and prospective assessment of ER was not required by that time. Here, we report the 25-year results of cyclophosphamide, methotrexate, fluorouracil and tamoxifen (CMFT) compared to TAM alone (N = 1445) according to patient and tumour characteristics including a retrospective central assessment of ER, progesterone receptor (PgR), human epidermal growth factor receptor 2 (HER2) and proliferation-related Ki67 antigen (Ki67).
2. Patients and methods
DBCG 82-C was a randomised, phase III 3-armed trial that nationwide in Denmark recruited women who by mastectomy obtained complete resection of a unilateral invasive carcinoma. Axillary sampling or clearance (level I and part of level II) was required.8 and 9 Eligible were high-risk (node-positive, tumour >5 cm, or invasion of skin or deep fascia) patients without signs of distant metastases as determined by routine examinations. Participants were postmenopausal (last normal menstrual period more than 5 years ago or aged 55 years or older in case of hysterectomy with at least one ovary preserved).
2.1. Central pathology review
Formalin fixed and paraffin embedded tumour blocks were collected and tissue microarrays were constructed centrally without knowledge of treatment assignment or clinical outcome.10 Tumours were assessed for ER (clone 1D5, dilution 1:200) and PgR (clone 1A6, dilution 1:800, Dako, Denmark), and dichotomised as positive (⩾1%) or absent (<1%). HER2 status was positive if HercepTest 3+ or HER2:CEN17 ratio ⩾ 2 in case of 2 + (HER2 FISH pharmDx, Dako, Denmark).11 Ki67 was assessed by immunohistochemistry (IHC) (clone MIB-1, dilution 1:200, Dako, Denmark), and dichotomised as high (>14%) and low (⩽14%).12
ER, PgR, HER2, and Ki67 were used to separate four surrogates for intrinsic subtypes: luminal A (ER and/or PgR present and not HER2 positive or Ki67 high); luminal B (ER and/or PgR present and either HER2 positive and/or Ki67 high); triple negative (TN; ER absent, PgR absent, and HER2 negative); HER2 enriched (ER absent, PgR absent, and HER2 positive).
Eligible patients were assigned to tamoxifen 30 mg daily for 52 weeks, tamoxifen concurrently with nine cycles of CMF (cyclophosphamide 600 mg/m2, methotrexate 40 mg/m2, and fluorouracil 600 mg/m2) given intravenously on day 1 every 4 weeks or tamoxifen concurrently with radiotherapy. Chemotherapy doses were primarily adjusted according to white cell and platelet counts (×109/l) on day 1 of the scheduled cycle as follows: platelets >100 and WBC >3.5, 100%; platelets 75–100 or WBC 2.0–3.5, 50% of all three drugs. If platelets <75 or WBC <2.0, the treatment was delayed for 1 week. Toxicity was recorded using a simplified translation of the World Health Organization (WHO) scale and following Grade 2–3 stomatitis or diarrhoea the doses of methotrexate and fluorouracil were reduced by 25%.
Findings on clinical examination and adverse events were recorded every 4 weeks during the first year, during follow-up every 6 months the following 4 years and then yearly until a first event or for a total of 10 years. Additional biochemical tests and imaging were done when indicated by symptoms or signs. A complete follow-up on vital status and secondary malignancies until 1st June 2010 was obtained for all patients through linkage to the Danish Central Population Registry and the Danish Cancer Registry.
2.4. Statistical analysis
Follow-up time was quantified in terms of a Kaplan–Meier estimate of potential follow-up.13 OS was calculated as the elapsed time from the date of randomisation until death, irrespective of cause of death. DFS was defined as the duration of survival without invasive loco-regional recurrence, distant metastases, contralateral invasive breast cancer, second primary non-breast invasive cancer or death irrespective of cause. OS and DFS were analysed unadjusted by the Kaplan–Meier method and groups were compared using the log-rank test. Hazard ratios [HR] were estimated from the Cox proportional hazards regression model to quantify the effect, and to explore interactions. Interactions between treatment and the covariates were investigated in separate models. The assumptions of proportional hazards were assessed by Schoenfeld residuals, and by including a time-dependent component in the model. Associations between regimen and other characteristics (excluding unknowns) were analysed by chi-square and Fisher exact test.P-Values are two-tailed. The Subpopulation Treatment Effect Pattern Plot (STEPP) statistical method was used to explore the pattern of treatment effect with respect to the magnitude of ER expression.14 The sliding-window approach has been applied. The values on the x-axis show median values of ER for patients in each subpopulation with the number of patients given in brackets. Statistical analyses were done with the statistical software R version 2.11.1 and SAS version 9.1 (SAS Institute, Inc., Cary, NC).
From October 1982 to March 1990, 736 participants were randomised to tamoxifen (TAM) and 709 to CMF plus tamoxifen (CMFT). Complete follow-up for survival was achieved for 1444 patients, while one patient emigrated 7 months after randomisation. A total of 117 patients were without follow-up for recurrence, and 89 patients were excluded in the per protocol analysis (Fig. 1). The randomisation was well balanced with no significant differences between the TAM and the CMFT groups (Table 1).
Fig. 1. Trial profile of the Danish Breast Cancer Cooperative Group 82C Trial.
Table 1. Base-line characteristics of the intention to treat population.
3.1. Study outcome
This analysis was conducted 20 years after closure of recruitment. Median estimated potential follow-up was 10 years for DFS and 24 years for OS. A total of 936 first events were observed (Table 2). Fig. 2A shows the Kaplan–Meier curves for DFS. In the intent to treat (ITT) analysis (N = 1328) the overall unadjusted HR for DFS in the CMFT group compared with the TAM alone group was 0.89; 95% confidence interval [CI] 0.78–1.01; P = 0.08. When adjusting for baseline characteristics, including age, nodal status, tumour size, deep fascia and skin invasion, histological type, grade, ER, PgR, HER2 and Ki67 the benefit of CMF on DFS appeared more marked (adjusted HR 0.82; CI 0.71–0.93; P = 0.003).
Table 2. End-point events at 10-years in the Danish Breast Cancer Cooperative Group 82C Trial.
TAM, tamoxifen; CMFT, cyclophosphamide, methotrexate, fluorouracil, and tamoxifen.
Fig. 2. Kaplan–Meier curves of disease-free survival (Panel A) and overall survival (Panel B). Blue line indicates tamoxifen; gold line indicates cyclophosphamide, methotrexate, fluorouracil and tamoxifen (CMFT) therapy.
Causes of death during the first 10-years are shown in Table 2, and by data cutoff (1st June 2010) 660 and 623 deaths occurred in the TAM and CMFT group, respectively. Fig. 2B illustrates that there was no statistical evidence of a difference in mortality (HR 0.96; CI 0.86–1.08; P = 0.51). OS appeared to be similar irrespective of age, lymph node status, tumour size, histological type and grade (adjusted HR 0.95; 95% CI 0.85–1.08; P = 0.44).
Tumour tissue was available for central assessment from 969 (67%) of the 1445 patients (Table 1). Patients randomised to CMFT more often had assessable material but the assessable cohort had a similar potential follow-up for OS, a similar DFS and a similar relative treatment effect. No significant differences were seen within groups according to ER, PgR or HER2. Median Ki67 was 5%, and patients with available material in the CMFT group significantly more often had high Ki67 (>14%) compared to patients in the TAM group (29% versus 23%, P = 0.02).
STEPP analyses were used to explore the pattern of treatment effect according to ER expression in terms of 10-year DFS and HR for DFS. Fig. 3 shows the distribution of ER expression for the 830 patients who received TAM or CMFT, and the benefit of chemotherapy appeared to be restricted to patients with low (<10%) or very high (⩾90%) ER expression. In patients with moderate ER expression (10–89%) the 10-year DFS does not seem to be improved from adding CMF to tamoxifen (Fig. 5). The corresponding group sizes are also denoted in Fig. 3.
Fig. 3. Subpopulation Treatment Effect Pattern Plot (STEPP) of 10-year disease-free survival (DFS) (Panel A) and hazard ratio (Panel B) according to oestrogen receptor (ER). The values on the x-axis show median values of ER for patients in each subpopulation with the number of patients given in brackets. (Panel A) Blue line indicates tamoxifen; gold line indicates cyclophosphamide, methotrexate, fluorouracil and tamoxifen (CMFT) therapy. (Panel B) Green line indicates hazard ratio of CMFT versus tamoxifen and the dottet lines indicates 95% confidence intervals.
Fig. 5. Forest plot illustrating proportional hazard models for disease-free survival (DFS) according to age, No. of positive nodes, tumour size, and histological type; to single markers oestrogen receptor (ER), and HER2; and to subtype defined by immunohistochemistry (IHC): TN (ER absent, progesterone receptor (PgR) absent, and HER2 negative); HER2 enriched (ER absent, PgR absent, and HER2 positive); Luminal A (ER and/or PgR present and not HER2 positive or Ki67 high); and Luminal B (ER and/or PgR present and either HER2 positive and/or Ki67 high). Hazard ratios refer to adjusted per protocol estimates obtained in the multivariate analysis.
When compared to the luminal A group a qualitative interaction (P < 0.0001) was observed in the non-luminal group between time since randomisation and DFS (Fig. 4). In the first 3-years the non-luminal subset had a significant poorer DFS (HR 3.65; 95% CI 2.90-4.60; P < 0.0001) opposed (Pinteraction < 0.0001) by an improvement in DFS after 3 years (HR 0.52; 95% CI 0.32-0.86; P = 0.02). In contrast, luminal B patients had compared to luminal A patients a significant poorer DFS in the first 3-years (HR 1.99; CI 1.53-2.59;P < 0.0001) as well as after 3-years (HR 1.36; CI 1.00-1.86; P = 0.05) and these two estimates were not significantly different (Pinteraction = 0.07).
Fig. 4. Kaplan–Meier curves of disease-free survival according to biomarker subsets. Panel A by immunohistochemistry (IHC) subtype: green line indicates TN (oestrogen receptor (ER) absent, progesterone receptor (PgR) absent, and HER2 negative); blue line HER2 (ER absent, PgR absent, and HER2 positive); gold line Luminal A (ER and/or PgR present and not HER2 positive or Ki67 high); and red line Luminal B (ER and/or PgR present and either HER2 positive and/or Ki67 high) and (Panel B) unadjusted HRs for disease-free survival (DFS) by time since randomisation with Luminal A as reference group. Confidence interval (CI) indicates confidence interval and P-values indicate test for heterogeneity.
In a multivariate Cox analysis, we examined heterogeneity of treatment effect according to subtype (Fig. 5). DFS seemed to be equally improved in patients with luminal A and luminal B subtypes, and there was no statistically significant interaction between subtype and treatment (Pinteraction = 0.45). CMF appeared to give a more pronounced prolongation of DFS in patients with a non-luminal (HR 0.64; CI 0.45–0.91) compared to a luminal subtype (HR 0.89; CI 0.74–1.07) but there was no statistically significant interaction between luminal and non-luminal subtype and treatment (Pinteraction = 0.10).
We did not find any evidence for heterogeneity of the effect of CMF on DFS according to age, number of positive lymph nodes, tumour size, ER status or HER2 status. The majority of patients (N = 1049; 85%) had an infiltrating ductal carcinoma while 119 (10%) had an infiltrating lobular carcinoma and 71 (6%) had other histological subtypes including 25 patients with medullar carcinomas. A significant reduction in the HR of DFS for CMFT versus TAM was observed for the other than ductal and lobular subtypes (adjusted HR 0.35; CI 0.18–0.67; Pinteraction = 0.01).
No treatment related deaths were reported. The mean of the relative cumulative dose (actual/planned mg/m2) was 0.96 and the mean of the relative dose intensity (actual/planned mg/m2 per time unit) was 0.82 for CMF in the CMFT group. Ninety-six patients (16%) did not complete all cycles, but only 30 (5%) received less than six series of CMF. Among patients who received CMFT 259 (44%) had WHO Grade 2 and 158 (27%) Grade 3 nausea and vomiting, 79 (14%) had Grade 2–3 diarrhoea, 98 (17%) had Grade 2–3 conjunctivitis or stomatitis, and 119 (20%) had complete alopecia.
At 10-years second non-breast malignancies were reported as first event in 19 (3%) patients treated with TAM and eight (1%) treated with CMFT (Table 2). At data cutoff a history of a second non-breast malignancy was reported in 85 patients in CMFT group compared to 95 in the TAM (P = 0.63; Table 3).
Table 3. Second non-breast malignancies.
In this trial, the addition of CMF to tamoxifen significantly improved disease-free but not overall survival in postmenopausal women with high-risk breast cancer. The addition of CMF to tamoxifen resulted primarily in a reduction of local/regional invasive recurrence. In the absence of radiotherapy the extent and quality of surgery may be particularly important and in a previous report from this trial we found that postmastectomy radiotherapy did benefit patients with larger tumours at least to a similar extent as those with smaller tumours.8
Although not statistically significant, the results on mortality of this study are in line with the findings of the EBCTCG overview, with the inclusion of the present study, that demonstrated a significant 11% reduction in overall mortality rate in women aged 50–69 in trials of chemotherapy versus no chemotherapy.5
We performed a retrospective analysis of ER, PgR, HER2, and Ki67 at a single central pathology laboratory in order to explore responsiveness to chemotherapy within subgroups. While patients with non-luminal cancers had a poorer prognosis in the first few years after prognosis a shift to a slightly superior residual prognosis appeared with longer follow-up as previously reported.15 The results do not provide support to the 2009 St. Gallen recommendations as benefits from adding CMF to TAM appeared to be similar in luminal A and luminal B.16 A trend towards a greater magnitude of benefit was shown in HER2 positive and TN cancers but overall there was no statistical evidence of an association between specific subtypes and benefit from CMF. The STEPP analysis suggested that patients whose tumours had intermediate ER levels, contrary to the overall results, did not benefit from treatment with CMF. A close association has been demonstrated between high ER levels and alterations in the ESR1 gene which might be a determinant of chemotherapy responsiveness. 17 and 18 The IBCSG collaborators found a different pattern in their STEEP analysis as the magnitude of benefit from adding CMF to tamoxifen seemed smaller for postmenopausal node-positive patients in IBCSG IX with high or intermediate levels of ER in the primary tumour.19 In SWOG 8814 postmenopausal patients with node-positive breast cancer and ER Allred score seven or eight did not seem to obtain a benefit in DFS from cyclophosphamide, doxorubicin and fluorouracil (CAF) if their tumours were HER2 negative.20 In the most recent EBCTCG meta-analysis of standard CMF versus no chemotherapy the proportional risk reduction appeared independent of ER status and subsets of ER positive.5 Interestingly, a high content of ER (⩾100 fmol/mg) seemed to be a determinant of responsiveness to anthracyclines.
Our study has several important limitations in particular on account of the treatment regimens and our results should be interpreted with appropriate caution. First, we have in a retrospective cross trial comparison among premenopausal patients demonstrated a 30% increase in the risk of recurrence when shifting from classical to the current intravenous and less dose-intensive CMF regimen.21 Dose-reductions have also been associated with decreasing benefit from adjuvant chemotherapy.22 and 23 Regimens including anthracyclines and taxanes are on average considered more efficacious than CMF based regimens.8 and 24However, the incremental benefit observed with anthracyclines and taxanes depends on patient and tumour characteristics and there is evidence to support that postmenopausal patients with ER positive and HER2 negative tumours have the least benefit.25, 26, 27 and 28 Second, in this trial CMF was given concurrently with tamoxifen and the St. Gallen 2009 consensus report recommended postponing endocrine therapy to after completion of chemotherapy.16 and 29
Tamoxifen was in this trial given 30 mg daily for 1 year, and additional benefit has been demonstrated from prolonging endocrine treatment to 5 years in patients with ER positive breast cancer,5 and the use of aromatase inhibitors further reduces the risk of a DFS event.30, 31, 32 and 33 Use of a more effective endocrine therapy could potentially reduce the absolute benefit achieved by CMF, but according to the EBCTCG meta-analysis one would expect the relative reduction in DFS events and mortality to be unchanged.5 Among patients in the TAM and CMFT arms a local or regional first event was experienced by 204 (30%) and 160 (25%) respectively (Table 2). An even larger difference has previously been reported between the TAM and TAM plus Radiotherapy arms, although the patients treated with the combination more often had a first recurrence at a distant site.8 Only 10% of the patients were node negative and more than a third had four or more positive nodes, and in this population a comparison with a CMFT plus radiotherapy arm would have been interesting.
In conclusion, CMF improves outcome in postmenopausal women with node-positive breast cancer when added to 1 year of tamoxifen. Notably the benefit from adding CMF to tamoxifen was not significantly different in luminal A and luminal B subtypes.
Role of the funding source
Conflict of interest statement
The DBCG 82C trial was financed and coordinated by the DBCG. The retrospective collection of tumour blocks and biomarker assays study was supported by IM Daehnfeldt Foundation, Danish Research Council (Grant No. 271060542), and Danish Ministry of Internal Affairs and Health (Grant No. 2006-12103-272).