Prognostic effect of estrogen receptor status across age in primary breast cancer

Authors


  • Previously presented as oral presentation at ECCO, Paris, November 3rd, 2005.

Abstract

Estrogen receptor (ER) status is considered as an important prognostic factor as well as a predictive factor for endocrine responsiveness in breast cancer. We analyzed the distribution of ER status across age and estimated variations in the prognostic impact of ER status related to patients' age and time since diagnosis. Overall, 26,944 patients with primary breast cancer diagnosed from 1989 to 2004 were included. The proportion of ER positive tumors increased over age from 51 to 82%. In multivariate analysis of overall survival, ER positive status was found to be a significantly positive prognostic factor over all age groups. This effect was limited to the first 5 years after diagnosis, RR: 2.08 (95% CI: 1.95–2.22, p < 0.0001). Overall survival during the following 5 years was slightly superior for women with ER negative tumors, RR of death: 0.89 (95% CI: 0.79–1.00, p = 0.049). Results were unchanged in patients who did not receive adjuvant systemic therapy (n = 6,272). Thus, positive ER status does not confer a negative impact on survival in young women as has been previously reported. The inferior prognosis for ER negative patients during the first 5 years after diagnosis changes into a slightly superior residual prognosis compared to ER positive patients independent of use of adjuvant systemic therapy. This may have an impact on future designing of guidelines for adjuvant endocrine therapy beyond 5 years. © 2007 Wiley-Liss, Inc.

The prognosis of patients with primary breast cancer depends on several factors identified through histopathological examination of the tumor and the regional lymph nodes. Estrogen receptor (ER) status is regarded as both a prognostic factor and a predictive factor for endocrine responsiveness. Although it is well known that the prevalence of ER positive tumors increases with age, it has been a matter of debate, whether the prognostic effect of having an ER positive breast cancer varies with age. Generally, ER positive tumors are considered to be less aggressive than ER negative tumors, but it has been argued that the opposite relation might exist for young patients.1 Furthermore, studies including smaller number of patients have indicated that the favorable prognostic effect for ER positive tumors is limited to the first year after diagnosis.2

We addressed these questions by analyzing a large cohort of women with primary breast cancer from the nationwide population-based database of Danish Breast Cancer Cooperative Group, DBCG, which contains detailed information on clinical and histopathological presentation, postoperative therapy and follow-up status. The main endpoint of these analyses was overall survival (OS). We further evaluated how OS was influenced by ER-receptor status over time since diagnosis as well as over age at diagnosis.

Material and methods

In 1977, the DBCG started nationwide prospective studies on the treatment of breast cancer.3 All primary clinical data and histopathological data as well as information concerning adjuvant therapy and follow-up status have been prospectively registered by the DBCG for breast cancer patients in Denmark. Data were linked between DBCG and the Danish Cancer Registry, which is considered almost complete in reporting of cancer diagnosis among residents in Denmark.4 In the present study population, a 96% concordance was found between the 2 registers.5 These data have been linked to data on survival, death and emigration registered by the Danish Civil Registration System (CRS), which implies a unique identification number given to all residents in Denmark.6

The present study was based on registration of all women <75 years (N = 34,817) diagnosed with primary breast cancer from December 1989 to January 2004. Evaluation of ER status by immunohistochemical technique gradually became standard procedure through the first study-years. The analysis was performed in the pathology laboratories serving the breast surgery units. The quality and thus uniformity of the analysis was guaranteed by participation in external quality assurance schemes.7 A tumor was defined as ER positive if more than 10% of the cells stained positively regardless of the intensity of the reaction. This definition was chosen in order to simplify the evaluation of the immunohistochemical reaction and thus minimizing intra- and interobserver variation. The definition is in accordance with standards in most laboratories in Europe.7

Nationwide guidelines for the treatment of primary breast cancer in Denmark are given by the DBCG according to treatment protocols. The present study includes the protocols DBCG 89 and DBCG 99. Patients were treated according to DBCG 89 from 1989 to 1999 and according to DBCG 99 from 1999 to 2004. Guidelines for risk group allocation and treatment have been described in detail elsewhere,3, 8, 9, 10 and are summarized in Table I.

Table I. Guidelines for Adjuvant Systemic Treatment of women with Primary Breast Cancer in Denmark According to Protocols DBCG 89 and DBCG 99 Including Patients <75 Years
ProtocolIncluded patientsTreatment-allocation
  1. CMF: Cyclophosphamide plus methotrexate plus fluorouracil; CEF: Cyclophosphamide plus epirubicin plus fluorouracil; TAM: Tamoxifen.

DBCG 89-ALow-risk or ER- neg. 70–74 yearsNone
DBCG 89-BHigh-risk premenopausal ER-pos., node-pos.1. CMF or
2. Ovarian ablation
DBCG 89-CHigh-risk postmenopausal, ER-pos.1. TAM 1 years or
2. TAM 2 years or
3. TAM half year + megace half year or
4. TAM 5 years
DBCG 89-DHigh-risk1. CMF or
I Premenopausal Node-neg. or ER-neg.2. CEF or
II Postmenopausal ER neg <70 years3. CMF + oral pamidronate or
4. CEF + oral pamidronate
DBCG 99-ALow-risk or ER- neg. 70–74 yearsNone
DBCG 99-BHigh-risk premenopausal ER-pos.1. CMF or
2. Ovarian Ablation or
3. CEF + TAM 5 years
DBCG 99-CHigh-risk postmenopausal ER-pos.TAM 5 years
DBCG 99-DHigh-risk ER-neg.Premenopausal: CEF
Postmenopausal: CMF

The primary surgical treatment included mastectomy or lumpectomy with axillary clearance. Patients were classified as having either low-risk or high-risk disease according to histopathological criteria. Low-risk patients were observed without further adjuvant treatment apart from radiation therapy to the residual breast after lumpectomy.

High-risk patients were allocated to adjuvant systemic therapy (chemotherapy and/or endocrine therapy) and/or radiotherapy. Patients with bilateral breast cancer, inflammatory cancer, distant metastases, or patients not treated according to the surgical guidelines or with other contraindications to standard protocol therapy were excluded from the prospective studies.

In protocol DBCG 89, high-risk patients included patients with positive nodal status or with tumors >5 cm or premenopausal with ductal carcinomas grade II or III. ER status did not serve as an independent criterion for allocation to high-risk group.

In protocol DBCG 99 inclusion to high-risk group was defined as patients being either node-positive or having tumors >2 cm or ductal grade II–III or under 35 years of age. In this protocol negative ER status was included as an independent risk factor and only patients 70–74 years did not receive adjuvant systemic treatment based on this criterion.

Permission to perform the study was granted from The National Scientific Ethics Committee and The Danish Data Protection Board.

Statistical analyses

Associations between ER status, age at time of diagnosis and tumor characteristics, respectively, were evaluated using χ2 tests. Overall survival was defined as time from date of surgery until death of all causes. The date of last follow-up was January 1, 2005 and all patients known to be alive at that date were censored. Because of a linkage to the Danish Civil Citizen Registration System, no patients were lost to follow-up. Curves illustrating overall survival (OS) were computed by ER-status and follow-up periods 0–5 and 5–10 years using the Kaplan-Meier method. Differences in OS between subgroups were analyzed using the log-rank test.

The association between ER status and survival was analyzed by means of Cox proportional hazard regression models adjusted for the covariates: axillary nodal status, tumor size, histological type and grade, age at diagnosis, and protocol allocation and version grouped as shown in Table II. In all analyses the stratum with the most observations was chosen as reference.

Table II. Characteristics of 26,944 Danish women with Primary Breast Cancer <75 Years Operated 1989–2004
 Number (%)ER positive (%)ER negative (%)p-value*
  • *

    p-values from χ2 tests testing whether ER-status is equally distributed across each of the other covariates.

Total26,944 (100)19,917 (73.9)7,027 (26.1) 
Age (years)   <0.0001
 −34530 (2.0)271 (51.1)259 (48.9) 
 35–39991 (3.7)598 (60.3)393 (39.7) 
 40–442,045 (7.6)1,382 (67.6)663 (32.4) 
 45–493,511 (13.0)2,462 (70.1)1,049 (29.9) 
 50–544,311 (16.0)3,083 (71.5)1,228 (28.5) 
 55–594,349 (16.1)3,287 (75.6)1,062 (24.4) 
 60–644,185 (15.5)3,219 (76.9)966 (23.1) 
 65–693,871 (14.4)3,036 (78.4)835 (21.6) 
 70–743,151 (11.7)2,579 (81.8)572 (18.2) 
Tumour size (mm)   <0.0001
 1–104,162 (15.5)3,374 (81.1)788 (18.9) 
 11–2011,291 (41.9)8,784 (77.8)2,507 (22.2) 
 21–5010,227 (38.0)6,965 (68.1)3,262 (31.9) 
 51–1,264 (4.7)794 (62.8)470 (37.2) 
Positive lymph nodes   <0.0001
 014,201 (52.7)10,667 (75.1)3,534 (24.9) 
 1–37,708 (28.6)5,823 (75.5)1,885 (24.5) 
 4–93,173 (11.8)2,220 (70.0)953 (30.0) 
 10–1,862 (6.9)1,207 (64.8)655 (35.2) 
Histological type/grade   <0.0001
 Nonductal5,089 (18.9)3,992 (78.4)1,097 (21.6) 
 Ductal grade I7,163 (26.6)6,434 (89.8)729 (10.2) 
 Ductal grade II9,562 (35.5)7,384 (77.2)2,178 (22.8) 
 Ductal grade III5,130 (19.0)2,107 (41.1)3,023 (58.9) 
Protocol Version   <0.0001
 DBCG 8914,922 (55.4)10,468 (70.2)4,454 (29.8) 
 DBCG 9912,022 (44.6)9,449 (78.6)2,573 (21.4) 
Protocol allocation   <0.0001
 Yes24,428 (90.7)18,186 (74.4)6,242 (25.6) 
 Not treated according to surgical guidelines620 (2.3)426 (68.7)194 (31.3) 
 Not allocated due to other reasons1,896 (7.0)1,305 (68.8)591 (31.2) 
Adjuvant treatment   <0.0001
 None8,956 (33.2)7,026 (78.5)1,930 (21.5) 
 Chemotherapy6,747 (25.0)2,868 (42.5)3,879 (57.5) 
 Endocrine8,305 (30.8)8,010 (96.4)295 (3.6) 
 Not-protocolized treatment2,936 (10.9)2,013 (68.6)923 (31.4) 

The proportional hazard assumption for each covariate was verified graphically by means of plots of log minus log of the survival density function vs. log time. A plot of the scaled Schoenfeld residual was made to illustrate the fluctuation of the hazard over time, followed by a Pearson's test for correlation. In fact the actual points were not plotted, but a smoothed spline with 95% confidence bands was fitted to the points.

The residuals were calculated for each event, and for the ith individual the residual is the difference between the observed value of the covariates, Xi and its conditional expectation given those individuals still under observation when the ith individual fails. If the proportional hazard holds the expectation of the ith residual should be ∼0, and a plot of the residuals versus time will be centered round 0.

It is possible to show that when the Schoenfeld residual is scaled with the variance matrix of the covariates to the minus 1st, the expectation of the scaled Schoenfeld residual plus the estimated constant coefficient β (the parameter estimate) equals approximately the time-dependent coefficient.11 If the assumption of proportional hazard was violated, the parameter was included in the analysis as a stratification variable. However for the parameter ER status the Schoenfeld residual plot was used to investigate the hypothesis that ER status had different effect according to follow-up time. This was further investigated using a likelihood ratio test.

To investigate the effect of ER status in each age group on OS an analysis was performed, where interaction between ER status and age group was added to the Cox model. Using the same model as above a subanalysis was performed in a group of low-risk patients who had not received any adjuvant treatment. All analyses were performed with the use of SAS.

Results

In all, 34,817 women with primary breast cancer <75 years operated during the period from December 1989 to January 2004 were registered. ER status determined with the immunohistochemical (IHC) method was available in 28,615 patients (82.2%). Patient characteristics and distribution of ER status according to age at diagnosis, tumor size, nodal status and histological type/grade are summarized in Table II. The proportion of ER positive patients increases with age from 51.4% in the youngest patients aged <35 years to 81.6% in the oldest age group 70–74 years. When separating the whole study period into 2 consecutive periods (years 1989–1996 and years 1997–2004, respectively) there is a decrease within all age groups in the percentage of ER-negative patients from 31% (3,108 ER-negative of 9,869 cases) in the first period to 23% (3,919 ER-negative of 17,075 cases) in the second period. This indicates that the general increase in incidence of breast cancer over time in this 15 years period primarily occurs among the group of ER positive patients by 27%, while the absolute number of ER negative patients remains unchanged over time.

Cases without information on histological type, grading, size or nodal status were excluded [n = 1,671 (5.8%)], leaving 26,944 patients for further analysis. At time of follow-up (January 1, 2005), 20,718 patients were alive (76.9%). The median time of follow-up was 4.8 years (range 0–15) corresponding to 148,690 person-years of follow-up. The distribution of patients according to adjuvant therapy is shown in Table III.

Table III. Distribution of Danish women with Primary Breast Cancer According to Adjuvant Treatment According to age Group for 26,944 Danish women Operated 1989–2004
Age (years)Adjuvant treatmentTotal
NoneChemotherapy ± endocrineEndocrine onlyNot allocated to protocol1
  • 1

    Treated not according to protocol, N = 2516. Unknown for other reasons, N = 420.

−3452 (9.8)391 (73.8)13 (2.5)74 (14.0)530
35–39141 (14.2)654 (66.0)62 (6.3)134 (13.5)991
40–44434 (21.2)1,179 (57.7)172 (8.4)260 (12.7)2,045
45–49906 (25.8)1,706 (48.6)437 (12.5)462 (13.2)3,511
50–541396 (32.4)1,253 (29.1)1,203 (27.9)459 (10.7)4,311
55–591,512 (34.8)629 (14.5)1,845 (42.4)363 (8.4)4,349
60–641,579 (37.7)532 (12.7)1,710 (40.9)364 (8.7)4,185
65–691,506 (38.9)399 (10.3)1,583 (40.9)383 (9.9)3,871
70–741,430 (45.4)4 (0.1)1,280 (40.6)437 (13.9)3,151
Total8,9566,7478,3052,93626,944

Overall survival according to ER status is shown in Figure 1. To evaluate the independent prognostic effect of ER status we performed a multivariate analysis including axillary lymph node status, age at diagnosis, tumor size, year of treatment and protocol allocation. Histological grading was included in the model as a stratification covariate. The overall relative risk of death for women with ER negative tumors was 1.65 (1.56–1.74) for the whole study period with women with ER positive tumors as reference (p < 0.0001).

Figure 1.

Overall survival according to ER-status 0–10 years after diagnosis in 26,944 Danish women with primary breast cancer operated 1989–2004.

In Figure 2 the Kaplan-Meyer plot is divided into follow up periods 0–5 years and 5–10 years, respectively. During the first 5-year period, ER negative patients had a significantly inferior prognosis (p < 0.0001). However, looking at all patients alive after 5 years since diagnosis (Fig. 2b) the picture changes: The prognosis of ER positive patients becomes slightly but significantly worse during the 5–10 years observation period compared to ER negative patients (p = 0.019).

Figure 2.

(a + b) Overall survival according to ER status 0–5 years and 5–10 years after diagnosis in 26,944 Danish women with primary breast cancer operated 1989–2004. Figure b includes 12,992 women alive 5 years after diagnosis.

The conditional probabilities of survival—which is given the patient survives the first 5 years after diagnosis—is shown in Table IV. The percentage called conditional survival (last column) is the percentage of patients alive after 10 years when looking specifically at the group of patients that were still alive after the first 5 years, which corresponds to the Kaplan-Meyer plot shown in Figure 2b.

Table IV. Probabilities of Survival at 5- and 10-Year after Diagnosis According to ER-Receptor Status among 26,944 Danish women with Primary Breast Cancer Operated 1989–2004
 Years after surgeryPatients at riskSurvivalConditional survival if alive after 5 years
ER positive019,917100%80%
59,73985%
102,48868%
ER negative07,027100%83%
53,25369%
101,10357%

The relative risk ratios over time for patients with ER negative tumors is visualized by a scaled Schoenfeld residual plot in Figure 3. The figure illustrates a nonproportional hazard and a worse survival for the ER negative patients during the first 5 years after surgery. From 5 years after surgery and onwards the curve is horizontal and the parameter estimate is almost equal to 0, i.e., in this time interval the hazard is proportional and the survival for ER negative patients seems slightly better than survival for ER positive patients. Subsequently the parameter for ER status was subdivided into the 2 periods of time in order to take into account the change in proportional hazard, as indicated by the smoothed residual plots. The other parameters are kept constant over time. The likelihood ratio test showed a significant effect of the splitting, (p < 0.0001).

Figure 3.

Smoothed scaled Schoenfeld residual plot for risk of death according to ER status among 26,944 Danish women with primary breast cancer treated 1989–2004. ER positive status = Reference. Dotted lines: 95% confidence bands.

Interaction terms between ER status and age were added to the multivariate Cox model. The interaction terms are close to significant, p = 0.0515 (Wald-test). A simple reparametrization of the model gives the prognostic effect of ER status across age groups as given in Table V. The table shows the relative hazard ratio for patients with ER negative tumors compared to ER positive tumors in each age group. There is a significant effect of ER status in each age group. Omitting all the interaction terms, the adjusted relative risk of death for patients with ER negative disease, keeping ER positive status as reference, was 2.08 (95% CI: 1.95–2.22) in the first 5 years after diagnosis. In comparison, the risk was 0.89 (95% CI: 0.79–1.00) for the period 5–10 years after diagnosis (Table VI).

Table V. Adjusted Relative risk for Death According to Receptorstatus Divided into age groups for 26,944 Danish women with Primary Breast Cancer Operated 1989–2004
AgeER status
PositiveNegative
0–5 years5–10 years
−34 years11.66 (1.19–2.30)0.71 (0.41–1.25)
35–39 years11.67 (1.26–2.22)0.75 (0.47–1.21)
40–44 years12.34 (1.87–2.93)0.97 (0.68–1.39)
45–49 years12.24 (1.88–2.67)0.82 (0.60–1.11)
50–54 years12.41 (2.08–2.81)0.63 (0.46–0.88)
55–59 years11.99 (1.71–2.31)0.77 (0.57–1.03)
60–64 years12.12 (1.84–2.45)1.15 (0.88–1.49)
65–69 years11.96 (1.70–2.26)1.05 (0.81–1.35)
70–74 years11.98 (1.70–2.31)0.99 (0.74–1.32)
Table VI. Adjusted Relative risk of Death for 26,944 Danish women with Primary Breast Cancer with ER Negative Disease 0–5 and 5–10 Years After Diagnosis
Variablep-valueHazard ratio95% CI
  1. p-values from Wald test. ER positive status reference = 1.

ER status<0.0001  
 Positive 0–5 years 1 
 Negative 0–5 years 2.08(1.95–2.22)
 Positive 5–10 years 1 
 Negative 5–10 years 0.89(0.79–1.00)
Age<0.0001  
 −34 years 1.34(1.13–1.59)
 35–39 years 1.07(0.92–1.24)
 40–44 years 0.86(0.76–0.98)
 45–49 years 0.83(0.74–0.92)
 50–54 years 1 
 55–59 years 1.23(1.12–1.36)
 60–64 years 1.50(1.36–1.65)
 65–69 years 1.74(1.58–1.91)
 70–74 years 2.25(2.05–2.48)
Tumour size<0.0001  
 1–10 mm 0.69(0.62–0.77)
 11–20 mm 1 
 21–50 mm 1.36(1.28–1.45)
 51– mm 1.90(1.73–2.09)
Positive lymphnodes<0.0001  
 0 1 
 1–3 1.55(1.45–1.66)
 4–9 3.06(2.84–3.29)
 10– 4.84(4.46–5.26)
Version<0.0001  
 89 protocol 1 
 99 protocol 0.70(0.65–0.75)
Protocol<0.0001  
 Yes 1 
 Not treated according to surgical guidelines 1.20(1.03–1.40)
 Not allocated because of other reasons 1.94(1.81–2.09)

Bias may be introduced when calculating the prognostic effect of ER status since the choice of adjuvant treatment may be based on receptor status. To investigate the natural course of the disease, we performed an analysis restricted to low-risk patients not receiving any adjuvant systemic treatment, n = 6,272. This subgroup analysis left the results unchanged, indicating that the time-limited positive prognostic effect of being ER positive is unrelated to the applied treatment. As shown in Table VII, the risk of death for ER-negative patients during the period 0–5 years after diagnosis was 2.13 (95% CI: 1.69–2.67) and during the 5–10 years period it was 0.80 (95% CI: 0.59–1.08). Correspondingly the Schoenfeld residual plot shows an overall similar configuration over time (Fig. 4).

Figure 4.

Smoothed scaled Schoenfeld residual plot for risk of death according to ER status among 6,272 Danish women with primary breast cancer diagnosed 1989–2004, not receiving adjuvant systemic treatment. ER positive status = Reference. Dotted lines: 95% confidence bands.

Table VII. Adjusted Relative risk of Death for Danish women with ER Negative Breast Cancer 0–5 and 5–10 Years after Diagnosis for 6,272 Danish women not Receiving Adjuvant Systemic Treatment
Variablep-valueHz ratio95% CI
  1. p-values from Wald test. ER positive status reference = 1.

ER status<0.0001  
 Positive 0–5 years 1 
 Negative 0–5 years 2.13(1.69–2.67)
 Positive 5–10 years 1 
 Negative 5–10 years 0.80(0.59–1.08)
Age<0.0001  
 −39 years 1.67(1.05 – 2.65)
 40–44 years 0.96(0.62 – 1.49)
 45–49 years 0.88(0.61 – 1.26)
 50–54 years 1 
 55–59 years 1.28(0.92–1.77)
 60–64 years 1.95(1.45–2.63)
 65–69 years 3.08(2.32–4.08)
 70–74 years 3.60(2.71–4.79)
Tumour size<0.0001  
 1–10 mm 0.67(0.55–0.81)
 11–20 mm 1 
 21–50 mm 1.43(1.21–1.70)
Hist. grade0.8181  
 Grade I 1 
 Nonductal 1.02(0.88–1.18)
Version0.1728  
 89 protocol 1 
 99 protocol 0.83(0.64–1.08)

In both Schoenfeld plots (Figs. 3 and 4) all the follow up time is depictured although we have limited the follow-up in the Cox regression models to 10 years as it seems obvious that it becomes difficult to draw safe conclusions based on the number of patients observed more than 10 years and thus the confidence intervals becomes too wide.

All premenopausal high-risk patients received adjuvant cytotoxic chemotherapy. Looking specifically into this subgroup, the prognostic effect of ER negative status was in line with the result of the whole study group (data not shown).

Discussion

In accordance with previous studies,12 we found that the proportion of ER positive breast cancers increases with age. When we analyze the prognostic effect of ER status, we observe a superior survival related to positive ER status for all women although less pronounced for women <40 years. Thus, we cannot confirm previous findings by Aebi et al. that ER-positivity confers a negative impact on survival for patients <40 years.1 Our study is based on ER status measured by IHC analysis, whereas the receptor analysis in the study by Aebi et al. used the biochemical (BC) method. Although there is a generally good accordance between the IHC and the BC analyses, it seems more consistent to include only patients where the same quality-assessed IHC method has been used in all patients. Moreover, it has been reported that endogenous hormone levels in premenopausal women may interfere with the BC receptor analysis whereas this should not affect the IHC method.13, 14 Thus, we cannot support the conclusion displayed at “Adjuvant Online” that patients <35 years at diagnosis carry an additional increased risk of death by a factor 1.5 if the tumor is ER positive (https://www.adjuvantonline.com/breastnew.jsp).

The patients' overall risk of dying is the sum of contributory factors: the disease, possible side effects of treatment of breast cancer and other causes of death. We chose OS as endpoint in order to avoid bias that may be introduced using breast cancer specific mortality or relapse free survival.

We confirm that patients with ER negative tumors have a relatively worse prognosis compared to patients with ER positive status, but the negative impact on outcome of being receptor negative is limited to the first 5 years after diagnosis as has also been reported by other investigators based on smaller materials.3 Beyond 5 years the positive effect of having an ER positive tumor changes into a slight disadvantage in relation to OS. In line with the latest Oxford overview we confirm that the prognosis of women with ER positive and ER negative tumors converge over time.15 Furthermore, the timeframe of this phenomenon is clarified, showing that the positive prognostic effect of positive ER status declines after 5–6 years. In 2005, the St. Gallen Consensus Panel 2005 redefined ER status as only a predictive factor and not a simple prognostic factor, i.e. ER status was excluded as a risk factor when defining the risk categories.16 Our study supports that ER status is not a simple prognostic factor since the prognostic effect changes from positive to negative over time. It is important to notice, that our observation remains unchanged when analyzing patients not receiving adjuvant systemic treatment. The outcome of this patient group represents the natural course of the disease following only local therapy. This observation makes it unlikely that adjuvant treatment is the reason for the time dependant change of the prognostic effect of ER status.

Duration of adjuvant endocrine treatment is a matter of considerable concern since more than half of all recurrences in women with ER positive tumors appear following the completion of the hitherto standard of 5 years tamoxifen.15 It is still not clearly defined that patients should be offered extended endocrine treatment i.e. more than 5 years.

In conclusion, we cannot confirm that ER positive status has a negative impact on survival in young women with primary breast cancer. The present study supports the view that ER status should not be regarded as a simple prognostic factor since prognostic impact varies profoundly over time after primary treatment despite use of adjuvant treatment.

Ancillary