Fax: +205-934-7094 or +205-975-7128
Hormone receptors and proliferation in breast carcinomas of equivalent histologic grades in pre- and postmenopausal women
Version of Record online: 7 DEC 2001
Copyright © 2001 Wiley-Liss, Inc.
International Journal of Cancer
Volume 98, Issue 1, pages 118–127, 1 March 2002
How to Cite
Talley, L. I., Grizzle, W. E., Waterbor, J. W., Brown, D., Weiss, H. and Frost, A. R. (2002), Hormone receptors and proliferation in breast carcinomas of equivalent histologic grades in pre- and postmenopausal women. Int. J. Cancer, 98: 118–127. doi: 10.1002/ijc.10171
- Issue online: 29 JAN 2002
- Version of Record online: 7 DEC 2001
- Manuscript Accepted: 12 OCT 2001
- Manuscript Revised: 20 AUG 2001
- Manuscript Received: 21 MAY 2001
- University of Alabama at Birmingham Center for Aging
- Avon Products Foundation
- National Cancer Institute. Grant Number: 5-R25-CA47888-13
- breast cancer;
- estrogen receptors;
- progesterone receptors;
Breast cancers in younger, premenopausal women are more likely to exhibit an adverse prognostic profile (including negative steroid hormone receptors and a high rate of cellular proliferation) and poor outcome than breast cancers in postmenopausal women. It has been hypothesized that this adverse prognostic profile is a result of the higher histologic grade of breast cancers in pre- compared with post-menopausal women. To assess the association of expression of steroid hormone receptors and indicators of proliferation while controlling for histologic grade, we identified 100 infiltrating ductal carcinomas from premenopausal women 45 years of age or younger and 100 from postmenopausal women 65 years of age or older. The carcinomas were selected so that the histologic grades (low versus high) were distributed equally between the 2 groups. Estrogen receptors (ER), progesterone receptors (PR), p27Kip1 and Ki-67 (to measure rate of proliferation) were assessed by immunohistochemistry and compared between groups. Clinical information and survival data were also analyzed. ER content was lower and proliferation was higher in carcinomas in premenopausal women (p = 0.048 and p = 0.005, respectively). By univariate analysis, p27Kip1 and PR were not different between the groups; however, in multivariate analysis, p27Kip1 was higher in postmenopausal women, but only in a subgroup with highly proliferative carcinomas. Overall survival was similar in the pre- and postmenopausal women. Furthermore, low p27Kip1 and African-American ethnicity predicted a poorer overall survival in the premenopausal, but not in the postmenopausal, women in our study. After controlling for histologic grade, a lower expression of ER and a higher proliferative index were detected in breast carcinomas in premenopausal women. Therefore, some prognostic indicators, such as ER and proliferative rate, may be more closely associated with menopausal status than histologic grade. Our data also suggest that some prognostic factors are not equally effective as predictors of survival in pre- and postmenopausal women. © 2001 Wiley-Liss, Inc.
Prognosis in women with breast cancer has been reported to differ according to age and menopausal status. Studies examining this issue have varied in design, age categorization and outcome measures. A few large population-based studies and more limited analyses1–5 have indicated that women under the ages of 30–40 years have a worse prognosis. Some retrospective studies have attempted to explain the less favorable outcome in younger women and have reported that breast cancers in younger women exhibit more adverse characteristics, including a high clinical stage at diagnosis, high histologic grade, lack of steroid hormone receptors and increased proliferation.6–9 Multivariate analyses to assess associations of age and menopausal status with other clinical and pathologic characteristics have not consistently retained either variable as an independent adverse prognosticator,8 although several multivariate analyses have indicated that young age or premenopausal status are independent predictors of disease recurrence.10–13 In addition, a few studies have found that those women at the extremes of age, i.e., premenopausal women younger than 33 years of age11 or postmenopausal women older than 80–85 years of age,3, 14 had a worse prognosis than other age groups.
Breast carcinomas in young women have been reported to express fewer estrogen receptors (ER),10, 15–18 to have a higher index of proliferation8, 19–22 and to be more likely to exhibit an abnormal nuclear accumulation of mutant p53 protein8, 23–26 than those in older women. Also, comparisons of the pathology of breast carcinomas have demonstrated that there is a greater proportion of high-grade carcinomas in young women (younger than 40 years)27, 28 and that special histologic types of invasive carcinoma (e.g., mucinous and papillary carcinomas) with more favorable prognoses than the more common infiltrating ductal carcinoma are more prevalent in older than younger populations.29, 30 However, histologic grade and type are closely related to the presence of other prognostic markers, including ER, rate of cell proliferation and p53 status.24, 31, 32 Thus, it has been speculated that the less favorable prognostic profile of breast carcinomas in younger women is a result of their higher histologic grade rather than the hormonal environment or other biologic differences.28 Previous studies comparing breast carcinomas in older and younger women have not been specifically designed to control for the association of histologic grade with the expression of prognostic markers.
To test the hypothesis that expression of ER and PR and indicators of cellular proliferation are more closely associated with menopausal status than histologic grade, we assessed the expression of ER, PR and p27Kip1 (a cyclin-dependent kinase inhibitor that prevents the transition from G1 to S-phase of the cell cycle), the rate of cellular proliferation and survival in premenopausal women 45 years of age or younger (≤45) and postmenopausal women 65 years of age and older (≥65) after controlling for histologic grade and type of carcinoma.
MATERIAL AND METHODS
Study design and population
Our study utilizes a case series design in which cases were selected retrospectively from the University of Alabama at Birmingham Hospital Tumor Registry. Case identification and accrual were performed after approval by the University of Alabama at Birmingham Institutional Review Board. Cases were selected from 2 randomly sorted lists of patients with breast cancer treated and/or diagnosed at the University of Alabama at Birmingham from 1988 to 1995. One group consisted of 442 women who were ≤45 at the time of diagnosis, and the second consisted of 501 women who were ≥65 at the time of diagnosis. Invasive ductal carcinomas from the first 100 women in each group who met the inclusion criteria (specified below) were selected. This case definition reduced the likelihood of misclassification of pre- and postmenopausal women, as those women who were perimenopausal (ages 46–64 years) were not included. Because the inclusion criteria were so exclusive, all available breast carcinomas (from over 900 patients) were required for initial evaluation and screening in order to identify the 200 patients that met the study criteria and were ultimately included.
Only invasive ductal carcinomas that had adequate formalin-fixed, paraffin-embedded tissue available from the archives of the Department of Pathology of the University of Alabama at Birmingham were selected. Special histologic subtypes were excluded. The medical records were reviewed and carcinomas from women who had been on hormonal replacement therapy at any time or who had had a hysterectomy before the age of 46 years were excluded. Furthermore, cases were selected so that the proportions of carcinomas of high and low histologic grades were distributed equally in the pre- and postmenopausal groups. The Nottingham modification of the Bloom and Richardson histologic grading system33 was used to categorize carcinomas as high grade, corresponding to a total score of 8 or 9, or low grade, corresponding to a total score of less than 8.
Collection of clinical data
Clinical information was obtained from review of medical records and the University of Alabama at Birmingham Tumor Registry. The information obtained included the following: date of birth, date of diagnosis, age at diagnosis, ethnic group, tumor size at diagnosis, number of involved lymph nodes and stage at diagnosis (using the American Joint Committee on Cancer staging criteria),34 gravida, parity, age when first pregnant, family history of breast cancer (i.e., breast cancer in the mother, paternal or maternal aunt, grandmother or sister), hysterectomy status, use of hormone replacement therapy and last date of contact with vital status or date and cause of death.
Our methods of performing immunohistochemistry have been reported in the literature.35–37 Specifically, sections (5 μm thick) were cut from each block and mounted on Superfrost/Plus slides (Fisher Scientific, Pittsburgh, PA). All sections were cut 1 day prior to immunostaining, to avoid antigen decay, and were attached to the slide by heating in a 60°C oven for 1 hr. The sections were deparaffinized in xylene (3 changes of xylene, 2 min each), hydrated in graded alcohols (absolute, 95% and 70% ethanol for 2 min each) and placed in Tris-buffered saline (0.05 M Tris base, 0.15 M NaCl, Triton X-100 4 drops/l, pH 7.6) prior to antigen recovery and immunostaining.
Sections were subjected to low temperature antigen retrieval with enzymatic pretreatment, which consists of predigestion in 0.1% trypsin (Type II-S from porcine pancreas, Sigma, St. Louis, MO) in PBS for 15 min in a 37°C oven followed by incubation in 10 mM citrate buffer, pH 6, for 2 hr at 80°C, as previously described.37–39 The slides were cooled and rinsed with deionized water. All sections were incubated with an aqueous solution of 3% hydrogen peroxide for 5 min and rinsed with Tris-buffered saline to reduce endogenous peroxidase activity, and then incubated with 1% goat serum for 1 hr at room temperature to reduce nonspecific background staining. Subsequently, sections were incubated with the appropriate monoclonal antibodies, diluted in PBS (pH 7.6) containing 1% BSA, 1 mM ethylenediamine tetraacetic acid and 1.5 mM sodium azide for 1 hr at room temperature. The antibodies used were the antiestrogen receptor mouse monoclonal antibody, clone ER88 (Biogenex, San Ramon, CA), at 1:30 dilution (0.33 mg/ml total protein), the antiprogesterone receptor mouse monoclonal antibody, clone PR88 (Biogenex) at 1:30 dilution (0.33 mg/ml total protein), the anti-Ki-67 mouse monoclonal antibody, clone MIB-1 (Biogenex) at 1:30 dilution (0.37 mg/ml total protein) and the anti-p27Kip1 mouse monoclonal antibody, clone 1B4 (Novocastra, Newcastle upon Tyne, UK) at 1:30 dilution (8.0 μg/ml IgG).
Primary antibody detection was accomplished using the Biogenex Super Sensitive Biotin-Streptavidin Horseradish Peroxidase Detection Kit (Biogenex). Sections were incubated with biotinylated goat anti-mouse antibody for 10 min and streptavidin peroxidase for 5 min with Tris-buffered saline washes between steps. The diaminobenzidine tetrachloride Super Sensitive Substrate Kit (Biogenex) was used to visualize the antibody-antigen complex. Sections were incubated with the diaminobenzidine tetrachloride for 7 min. Appropriate negative controls, consisting of histologic sections of each case processed without the addition of primary antibody, were prepared for each antigen retrieval method along with a positive/negative, multitissue control section.
After immunostaining, slides were stained for 1 min in hematoxylin, rinsed in water, dehydrated in graded alcohols (70%, 95% and absolute ethanol for 2 min each) and washed in xylene before coverslipping.
All slides were examined and scored by 2 of the authors (A.R.F. and W.E.G.) concurrently. The intensity of nuclear immunostaining of individual cells was scored on a scale of 0 (no staining) to 4+ (strongest intensity), and the percentage of cells with nuclei staining at each intensity was estimated. The proportion of cells at each intensity was multiplied by the corresponding intensity value, and these products were added to obtain an immunostaining score (immunoscore) ranging from 0 to 4.35, 36
ER, PR and p27Kip1 content was assessed as high or low using the median immunostaining score (based on both extent and intensity of staining) of all 200 cases for each antigen as a cutoff value. Scores above the median were categorized as high and those below the median as low. Proliferation, as reflected by immunohistochemical staining for Ki-67, was similarly categorized as high or low. ER and PR content were classified additionally as negative or positive in some analyses with positive cases defined as those with 10% or more of cells exhibiting nuclear staining. In a confirmatory analysis, Ki-67 staining also was assessed as high or low based only on the percentage of positively staining cells using the median percent positive cells (28%) as a cutoff point. Spearman's rank correlation coefficients were computed to assess interobserver agreement regarding evaluation of immunoscores. The average of the observers' immunoscores or percentage of positively staining cells for each antigen in each carcinoma was used in subsequent analyses.
Associations of ER, PR, p27Kip1 and Ki-67 with age and the other variables were evaluated with the Mantel-Haenszel chi-square test. All significance tests were 2-sided, with an α level of 0.05.
Multivariate logistic regression models with backward elimination procedures were performed to identify factors significantly associated with ER, PR, p27Kip1 and Ki-67 while retaining menopausal status in the model. The odds ratio (OR) and 95% confidence interval (CI) were computed as the point estimate and corresponding precision of the measure of effect. Due to the numbers of women for whom data were missing for gravida, parity and age when first pregnant, these variables were not included in the multivariate models. All other clinical factors assessed were included in all multivariate models with menopausal status and biomarker expression. Terms representing possible interaction between covariates were included in the model. Any interaction term found to be statistically significant was retained in all further models in addition to the main effects terms. To evaluate whether the apparent association between menopausal status with ER content and proliferation was in fact an effect of menopausal status or rather was an effect of age, additional multivariate logistic regression models that included age were performed. Separate models were performed for pre- and postmenopausal women, and age categories chosen were based on tertiles of the age distribution (less than 37, 37–41, 42 and older in the premenopausal women and less than 68, 68–73, and 74 and older in the postmenopausal women) within each menopausal group. Indicator variables were created to compare expression of ER, PR, p27Kip1 and Ki-67 using the lowest age tertile as the referent group.
Kaplan-Meier methods were employed to estimate the survival function for overall survival for pre- and postmenopausal women.40 Overall survival was defined as time from the date of diagnosis to the date of death. Patients who were alive at the last date of contact, died from unknown causes or died from a disease other than breast cancer were censored at the date of last contact. Differences in overall survival functions between younger, premenopausal and older, postmenopausal women were assessed using the log-rank test.
Multivariate Cox proportional hazards models were performed to assess the effect of ER, PR, Ki-67, p27Kip1 and other covariates on survival.41 All models were tested to ensure that assumptions regarding proportionality were met. Next, interaction terms representing possible interaction between covariates on overall survival were studied. Any interaction term that was found to be statistically significant was retained in all further models as well as the corresponding main effects terms. A stepwise procedure was used to determine order of entry for all covariates of interest. The hazard ratio and 95% CI were reported for the significant covariates in the model.
Characteristics of the study population
A comparison of the characteristics of the younger, premenopausal and older, postmenopausal cases is presented in Table I. These cases are not representative of breast carcinomas in pre- or postmenopausal women in general because these cases were selected so that the distribution of histologic grades was similar in each group. If these carcinomas had not been selected in this manner, a higher percentage of high-grade carcinomas would be expected in the premenopausal group and a lower percentage expected in the postmenopausal group.27, 28 In spite of this selection process, the younger, premenopausal group of carcinomas had a higher stage (greater than stage I; p = 0.003), a greater incidence of positive lymph nodes (p = 0.041) and larger tumors (p < 0.001). There was a higher percentage of Caucasian women (84%) than African American women (16%) who were premenopausal and ≤45 (p = 0.012).
|Premenopausal||Postmenopausal||Crude OR||95% CI||p-value|
|Tumor size (cm)|
|Age first pregnant (yr)|
The frequencies of radiation therapy (p = 0.882) and lumpectomy or mastectomy (p = 0.594) were similar in pre- and postmenopausal women (data not shown). However, premenopausal women were more likely to receive chemotherapy (70%) than postmenopausal women (16%) (p < 0.001), and postmenopausal women were much more likely to receive hormonal therapy (31%) than premenopausal women (1.2%; p < 0.001). The median duration of follow-up for the 200 women was 5.9 years.
Univariate comparison of ER, PR, p27Kip1 and proliferation in carcinomas in pre- and postmenopausal women
When ER, PR and p27Kip1 content and Ki-67 were categorized as high (immunoscore above the median) or low (immunoscore below the median), ER content was significantly higher and Ki-67 was significantly lower in breast carcinomas in postmenopausal women (p = 0.048 and p = 0.005, respectively; Table II). PR and p27Kip1 content was not significantly different in the 2 groups in this analysis. When the ER content was categorized as positive or negative, there was no difference between carcinomas in pre- and postmenopausal women. Forty-seven percent of younger, premenopausal women had ER-positive carcinomas compared with 53% of older, postmenopausal women (p = 0.315). When PR content was reclassified as positive or negative, there was still no difference between the groups with 54% of carcinomas in younger women and 46% in older women being PR positive (p = 0.393). Because Ki-67 expression is often assessed as the percentage of cells with nuclear staining, rather than as an immunoscore, we also compared Ki-67 expression in the 2 groups of carcinomas by designating those carcinomas with less than 28% (the median percentage) of the cells staining as low and those with 28% or more cells staining as high. Using this categorization, 57% versus 43% of carcinomas in younger and older women, respectively, had a high rate of proliferation (or Ki-67 expression). This difference was significant (p = 0.048).
|Biomarker1||Menopausal status (%)||OR||95% CI||p-value|
Multivariate analysis of ER, PR, p27Kip1 and Ki-67 in carcinomas in pre- and postmenopausal women
Multivariate logistic regression analyses were conducted using a backward elimination procedure to determine factors associated with ER, PR, p27Kip1 and Ki-67 while forcing menopausal status to remain in the model. The significance level to stay in the model was set at 0.05. It was found that there was a modification of the association of menopausal status and ER immunoscore by lymph node status (Table III). Furthermore, there was a modification of the association of p27Kip1 immunoscores and menopausal status by Ki-67 immunoscores (Tables III, IV). Thus, interaction terms between menopausal status and lymph node status and between menopausal status and Ki67 were included in the logistic regression models.
|ER multivariate model|
|Positive lymph nodes2|
|Negative lymph nodes2|
|Positive lymph nodes||3.87||(1.29–11.65)||—|
|Negative lymph nodes||1.00||—||—|
|Positive lymph nodes||0.65||(0.22–1.98)||—|
|Negative lymph nodes||1.00||—||—|
|PR multivariate model|
|p27Kip1 multivariate model|
|Ki-67 multivariate model|
In the multivariate analysis, a high ER content was significantly more likely in breast carcinomas in postmenopausal women, but only among those with uninvolved lymph nodes. In this group, postmenopausal women were 5.03 times more likely than premenopausal women to have a high ER score (OR = 5.03; 95% CI: 1.83–13.86). In women with involved lymph nodes, there was no significant difference in ER content in carcinomas from pre- and postmenopausal women. However, among premenopausal women, those with positive lymph nodes were almost 4 times more likely than those with negative lymph nodes to have a high ER score (OR = 3.87; 95% CI: 1.29–11.65) in their primary carcinomas (Table III). Low histologic grade, high PR and high p27Kip1 also were independently associated with a high ER.
Menopausal status was not associated with PR expression (OR = 0.63; 95% CI: 0.31–1.28; Table III). However, low histologic grade, high ER, high p27Kip1 and low Ki-67 were independently and significantly associated with high PR.
Although in univariate analysis there was no significant difference in p27Kip1 content in pre- and postmenopausal women, by multivariate analysis, there was a significant association between postmenopausal status and a high p27Kip1, but only in those cases with a high Ki-67 (Table IV). When considering only those cases with a high Ki-67, postmenopausal women were more likely than premenopausal women to have a high p27Kip1 (OR = 3.56; 95% CI: 2.89–4.39). In general, a high Ki-67 would be expected to be associated with a low p27Kip1, and this was the case in premenopausal women. In premenopausal women, those with high Ki-67 scores were significantly less likely to have high p27Kip1 than those with low Ki-67 scores (OR = 0.30; 95% CI: 0.15–0.61). For postmenopausal women, there was no significant difference in p27Kip1 content between those with high and low Ki-67 (OR = 1.92; 95% CI: 0.74–4.96). Alternatively, when analyzed on the basis of factors associated with a high Ki-67, Ki-67 was not associated with a high or low p27Kip1 (Table IV). Additionally, high ER and high PR were independently and significantly associated with a high p27Kip1.
Postmenopausal status was independently associated with a low Ki-67 (OR = 0.34; 95% CI: 0.17–0.66; Table IV). High histologic grade and low PR were also significantly associated with a high Ki-67.
When assessing the relationship of age (ordered in tertiles as described in the Material and Methods section) within each menopausal status to high ER, PR, p27Kip1 or Ki-67, there was no significant association with age and any of these factors. In addition, there were no significant trends (decreasing or increasing) in expression of ER, PR, p27Kip1 and Ki-67 according to age in either pre- or postmenopausal women (data not shown). This suggests that menopausal status, rather than age, is the primary variable significantly associated with ER, p27Kip1 and Ki-67 expression in our study.
Overall survival, as assessed by the Kaplan-Meier method, was not different for pre- and postmenopausal women (p = 0.731; Fig. 1). As expected, there was a significantly poorer overall survival in women with high-grade carcinomas than those with low-grade carcinomas (p < 0.001); however, the poorer survival in those with high-grade carcinomas was found only in premenopausal women (Fig. 2). By multivariate Cox regression analysis (Table V), axillary lymph node involvement was significantly associated with poorer survival in both pre- and postmenopausal women; however, in univariate analysis, the difference in survival rates for those with involved and uninvolved lymph nodes was significant in postmenopausal women but was only of borderline significance in premenopausal women (p = 0.056; Fig. 3). African-American ethnicity and a low p27Kip1 immunoscore were also associated with poorer survival, but only in premenopausal women (Table VI, Fig. 3). In postmenopausal women, the association between a low p27Kip1 score and poorer survival approached statistical significance (p = 0.054, Kaplan Meier analysis not shown). By multivariate Cox regression analysis, histologic grade was not an independent prognostic indicator in pre- or postmenopausal women. None of the other clinical or pathologic factors assessed in our study affected overall survival.
|Variable||Hazard ratio||95% CI||p-value|
|Involved lymph nodes||2.59||(1.26–5.31)||0.010|
|Involved lymph nodes||2.20||(1.08–4.52)||0.031|
Prior analyses of the effect of age and menopausal status on the expression of prognostic or predictive markers in breast carcinomas have been limited in the factors assessed; however, studies have demonstrated that breast carcinomas in younger or premenopausal women are more likely to exhibit biologic and prognostic features that are known to be associated with a high histologic grade, such as an absence of steroid hormone receptors,31, 42, a high rate of proliferation43 and abnormalities in p53 protein.8, 24–26 It has been hypothesized that this adverse pattern of prognostic indicators and the poorer clinical outcome of younger, premenopausal patients were secondary to the greater proportion of higher grade carcinomas found in these women. Our study was designed to assess whether the expression of some prognostic indicators, specifically ER, PR, p27Kip1 and the rate of cellular proliferation, are more closely associated with menopausal status than with histologic grade. To our knowledge, this is the first comparison of carcinomas in premenopausal and postmenopausal women in which the carcinomas were selected so that the distribution of histologic grades was similar in the 2 groups. After controlling for the effects of histologic grade, some differences in prognostic indicators, such as ER content and proliferation, persisted, suggesting that these factors may be more closely associated with menopausal status than histologic grade.
A higher rate of proliferation in invasive breast carcinomas in younger women has been documented in multiple studies using thymidine labeling indices,19 S-phase fraction8, 12, 20, 21 or the number of Ki-67-positive cells;22, 44; however, not all investigators have found a correlation between proliferation rate and age or menopausal status.17, 43, 45 Similarly, most studies have reported that ER-positive invasive breast carcinomas are more common in older or postmenopausal women than in younger or premenopausal women.8, 10, 11, 17, 18, 42, 46–49 Some studies have not detected a relationship between ER and menopausal status after controlling for the effect of age,15, 50–53 suggesting that age may be more important than menopausal status in determining ER expression. The proportion of PR-positive breast carcinomas has been variably reported to be higher in older age groups,8, 11–13, 46 lower in postmenopausal women54 or not varying with age or menopausal status.16, 17 Several studies have found that p27Kip1 expression does not differ significantly by age or menopausal status.55–58 None of these studies, however, were designed to specifically consider the independence of these findings with respect to histologic grade.
Most prior studies that have assessed ER content in younger or premenopausal and older or postmenopausal women have utilized biochemical assays with various cutoff points (usually at a relatively low level of expression) to designate positive and negative ER status.11, 46, 54 To our knowledge, only 1 prior study has utilized immunohistochemistry to compare breast carcinomas in pre- and postmenopausal women.59 Our study demonstrated a greater number of ER-positive carcinomas, defined as more than 10% of cells with nuclear staining, in postmenopausal women. When we utilized a cutoff value of 10% of nuclei staining to classify carcinomas as ER-positive or -negative, we found no differences in ER content in pre- and postmenopausal women. Only by classifying ER expression as high or low based on the median immunoscore was a difference in ER expression identified. This suggests that the proportion of cases expressing any level of ER was similar between the 2 groups, perhaps as a result of controlling for histologic grade. However, postmenopausal women were more likely to express higher levels of ER than premenopausal women. We similarly analyzed PR content by both methods of quantification and found no differences between the groups by either method.
The variability in ER expression in breast carcinomas in premenopausal versus postmenopausal women may reflect different levels of circulating estrogens in pre- and postmenopausal women. Breast carcinomas in premenopausal women are reported to express ER in both phases of the menstrual cycle.60 The percentage of ER-positive tumors detected during the first or second halves of the cycle has been reported to be similar61 or higher in the first half of the cycle than in the second half.54 No consistent relationship between plasma estradiol and the ER content of breast carcinomas has been reported.62–65 There is a blood-tissue gradient of estradiol that varies considerably between individuals, and, in general, plasma estradiol levels do not reflect tissue estradiol levels.66, 67 In postmenopausal women, the levels of estradiol in breast tissue are much greater than plasma levels.67, 68 The concentration of estradiol in breast tumors from postmenopausal women remains as high as the plasma level in premenopausal women with breast cancer, and the concentrations of estradiol in carcinomas in pre- and postmenopausal women are similar.67, 69 Furthermore, tumor levels of estrogen are higher than that of the normal breast.67 The high tumor estrogen concentrations are derived from in situ estradiol production from plasma estrogen precursors.70
A correlation between tissue estradiol content and ER expression in breast carcinomas has been documented. Higher estradiol concentrations have been found in ER-positive tumors than ER-negative tumors.67, 69, 71 This suggests a positive relationship between estrogen and ER; however, such a relationship has not been reported consistently. Estrogen has been shown to both positively and negatively regulate ER.72, 73 This regulation has varied with tissue or cell type examined and the culture conditions used. Therefore, evidence suggests that circulating levels of estrogen are not responsible for the tissue level of ER in carcinomas and do not explain the differences in ER expression observed in our study.
There are other hormonal differences in pre- and postmenopausal women that may explain the observed differences in ER expression in breast carcinomas. Progesterone, in the presence of a functional PR, can inhibit transcription of the ER gene.74 The higher progesterone levels in premenopausal women54 could result in lower ER expression, but only in the presence of PR, which represents less than 50% of the breast carcinomas in premenopausal women. Some peptide hormones, cytokines and growth factors have been shown to alter ER expression60 and may be present at different levels in pre- and postmenopausal women. In addition, the breast tissue of younger women is composed of a higher proportion of fibrous breast stroma and epithelial cells and a lower proportion of adipose tissue than the breast tissue of older women. Stromal cells may modulate the growth of normal and neoplastic breast epithelial cells and secrete growth factors in response to various levels of endogenous hormones,75 and adipose tissue produces estrogen.70 These differences in the proportions of fibrous and adipose tissue and epithelial cells may have differential effects on ER expression.
ER expression has been correlated inversely with the rate of proliferation of breast carcinomas.76–78 By multivariate analysis, we did not find a significant association between a high Ki-67 and ER; however, the immunoscores for ER and Ki-67 were inversely correlated (Spearman's rank correlation coefficient = −0.415, p < 0.001). Although all evidence does not support this, estrogen stimulation of ER could result in proliferation and downregulation of ER, contributing to the observed differences in the expression level of ER and rate of proliferation in pre- and postmenopausal women. Alternatively, regulatory proteins of the cell cycle, such as cyclin E,79 or other growth factors, such as epidermal growth factor80 or fibroblast growth factor81 may be stimulating proliferation independently of estrogen or ER.
In the multivariate analysis for factors associated with high ER, the association between a high ER content and postmenopausal status was found only in women with uninvolved lymph nodes. There is a general consensus that ER content is a weak prognosticator, with most studies conferring a more favorable prognosis to ER-positive tumors.82–84 Our results support this contention, but only in postmenopausal women. In contrast, other investigators have reported that, in postmenopausal women, a high ER content may confer a more aggressive tumor behavior prior to receiving adjuvant therapy.85, 86 It has also been reported that in women younger than 35 years of age who were treated with cyclophosphamide, methotrexate and fluorouracil, those with ER-positive tumors experienced a significantly worse disease-free survival than those with ER-negative tumors.87 This was not true for older women. This report is in accordance with our finding that in premenopausal women, those with positive lymph nodes were more likely to have a primary carcinoma with a high ER than those with negative lymph nodes. Although these results suggest a difference in the prognostic effect of ER in pre- and postmenopausal women, ER was not an independent predictor of survival in either group.
p27Kip1 is a member of the Cip/Kip family of cyclin-dependent kinase inhibitors, which function to inhibit the transition from G1 to the S-phase of the cell cycle.88 p27Kip1 is highly expressed in nondividing cells and is also directly affected by estrogen and androgens. Treatment of growth-arrested MCF-7 cells with estradiol resulted in downregulation of p27Kip1.89 Therefore, it appears that the induction of proliferation by estrogen occurs in part through a decrease in the expression and inhibitory activity of p27Kip1. We detected no significant difference in p27Kip1 expression in pre- and postmenopausal women by univariate analysis; however, in multivariate analysis, postmenopausal women with a high Ki-67 were significantly more likely than premenopausal women with a high Ki-67 to have a high p27Kip1. Because p27Kip1 is an inhibitor of proliferation, an inverse association between p27Kip1 and Ki-67 would be expected, but, as is evident, this is not always the case. Others have reported a lack of correlation between p27Kip1 and proliferation.55, 56, 90, 91 These findings suggest that other proliferative factors, such as cyclin D1,58, 92 cyclin E93 or c-myc,94 may work to overcome the inhibitory effects of p27Kip1 to allow cell proliferation, especially in postmenopausal women. We, as have others,31, 55, 57, 58, 90, 91 have observed a significant positive correlation between ER and PR and p27Kip1.
We found no association between ER, PR, p27Kip1 and Ki-67 and age, rather than menopausal status, in our study, suggesting that tumor behavior may be influenced more by the hormonal environment or menopausal status than the age of the patient.
Even after controlling for histologic grade and type of carcinoma in our study, breast carcinomas in the premenopausal group were more likely to present at a higher stage of disease than those in postmenopausal women. These results are similar to those in previous studies in which cases were not selected to control for histologic grade or type.3, 8, 12, 49
Although the premenopausal women in our study presented with a higher stage of disease, overall survival in the pre- and postmenopausal women was similar. When comparing overall survival in women with high- and low-grade carcinomas, those with high-grade carcinomas had a worse prognosis, but, unexpectedly, this difference was largely a result of the particularly poor outcome of premenopausal women with high-grade carcinomas. In multivariate Cox survival analysis, however, histologic grade was not found to be an independent prognostic factor for either pre- or postmenopausal women. That the tumors in the premenopausal women were more likely to present at a high stage of disease and with involved lymph nodes than the tumors in the postmenopausal women, despite a similar overall survival (in high- and low-grade tumors combined) between the 2 groups, may well be a result of the sample size of 200 and the median follow-up period of 5.9 years. A larger sample size and/or a longer duration of follow-up might allow the detection of a difference in survival between the groups. Additionally, the premenopausal women in our study were more likely to receive adjuvant chemotherapy. This may have contributed to the equivalent survival between the groups. Another possible explanation is that postmenopausal women are less able to tolerate their breast carcinomas than premenopausal women, even though their cancers are less advanced. This may be a result of age-related differences, such as decreased effectiveness of the immune system,95 alterations in hormonal, growth factor or growth factor inhibitor levels,96–98 a decline in overall health, or an increase in comorbid conditions.99
Tumor stage, including lymph node status, has long been considered the strongest predictor for survival in breast cancer.100, 101 In our study, lymph node status was a significant independent prognostic factor in both pre- and postmenopausal women, although in univariate analysis, its prognostic value in premenopausal women was only of borderline significance. In premenopausal women, ethnic group was also an independent prognostic indicator of survival, with African-American women having a poorer prognosis. This is consistent with much of the literature on breast cancer in African-American women. African-American women with breast cancer are more likely to present at a relatively young age (<50), have a poorer survival within each tumor stage and are more likely to present with high-grade cancers.102–108 Overall mortality rates are higher in African-American than Caucasian women younger than 45 years; however, mortality rates are higher in Caucasian than African-American women older than 65 years.102, 109
A low expression of p27Kip1 was found to predict a poorer overall survival in our group of premenopausal women and approached significance as a prognostic indicator in postmenopausal women. With a larger sample size, a low p27Kip1 may have also proved to be prognostic in the postmenopausal group. A low or absent nuclear expression of p27Kip1 detected by immunohistochemistry has been shown to be a clinical marker of disease progression in several types of tumors, including breast cancer. Low p27Kip1 was associated with a worse relapse-free and overall survival in several univariate and multivariate analyses,57 most of which included proliferation rate and/or histologic grade in the multivariate model.55, 56, 93, 110, 111 Gillett et al.112 found a significant association between high p27Kip1 expression and relapse-free and overall survival in univariate analysis; but by multivariate analysis, p27Kip1 was not an independent predictor of survival when either histologic grade or proliferative activity was included in the model. In an analysis of 512 consecutive cases of breast carcinoma with a median follow-up of 9 years, Barbareschi et al.58 found that p27Kip1 expression did not predict outcome. However, in the node-negative subgroup of these cases (n = 249), a high p27Kip1 expression indicated a poor prognosis. In contradiction to this, Reed et al.91 did not find p27Kip1 to be a significant prognostic indicator in 77 node-negative breast carcinomas with a median follow-up of 13 years. In a study confined to grade 1 infiltrating ductal carcinomas (148 cases) with a median follow-up of 11.3 years, Leong et al.90 determined that p27Kip1 was insufficiently sensitive to predict those patients with a poor outcome, a relatively small number of 19%.
Our study supports the importance of histologic grade as an indicator of prognosis and of tumor biology in premenopausal and postmenopausal women. However, even after controlling for histologic grade, there remain differences in ER content, p27Kip1 expression and proliferation in breast cancers in pre- and postmenopausal women. Furthermore, we have also demonstrated the potential importance of considering menopausal status when assessing the significance of prognostic factors. It is likely that other molecular markers may differ in expression and prognostic utility in breast carcinomas in pre- and postmenopausal women.
L.I. Talley was partially supported by the Cancer Prevention and Control Training Program, funded by the National Cancer Institute, R25 grant number 5-R25-CA47888-13.
- 12Age and the risk of breast cancer recurrence. Cancer Control 1996;3: 421–7., , , et al.
- 13Menopausal status and the impact of early recurrence on breast cancer survival. Cancer Control 1997;4: 335–41., , , et al.
- 29Breast carcinoma in elderly women: pathology, prognosis, and survival. Pathol Annu 1984; 19 Pt 1: 195–219., , , et al.
- 1FlemingID, CooperJS, HensonDE, et al., eds. AJCC cancer staging handbook, 5th ed. Philadelphia: Lippincott-Raven, 1998.
- 35Immunohistochemical evaluation of biomarkers in prostatic and colorectal neoplasia. In: HanausekM, WalaszekZ, eds. John Walker's methods in molecular medicine—tumor marker protocols. Totowa, NJ: Humana Press, 1998. 143–60., , , et al.
- 36Factors affecting immunohistochemical evaluation of biomarker expression in neoplasia. In: HanausekM, WalaszekZ, eds. John Walker's methods in molecular medicine—tumor marker protocols. Totowa, NJ: Humana Press, 1998. 161–79., , , et al.
- 41Regression models and life tables. J R Stat Soc 1972;34: 187–220..
- 60Estrogens, estrogen receptor and breast cancer. Amsterdam: OS Press, 2000. 154–60..
- 83Relative effect of steroid hormone receptors on the prognosis of patients with operable breast cancer. A univariate and multivariate analysis of 3089 Japanese patients with breast cancer from the Study Group for the Japanese Breast Cancer Society on Hormone Receptors and Prognosis in Breast Cancer. Cancer 1992;69: 153–64., , , et al.
- 84Relative worth of estrogen or progesterone receptor and pathologic characteristics of differentiation as indicators of prognosis in node negative breast cancer patients: findings from National Surgical Adjuvant Breast and Bowel Project Protocol B-06. J Clin Oncol 1988;6: 1076–87., , , et al.
- 92High level expression of p27(kip1) and cyclin D1 in some human breast cancer cells: inverse correlation between the expression of p27(kip1) and degree of malignancy in human breast and colorectal cancers. Proc Natl Acad Sci USA 1997;94: 6380–5., , , et al.
- 101Prognostic and predictive factors in breast cancer. In: ManniA, ed. Contemporary endocrinology: endocrinology of breast cancer. Totowa, NJ: Humana Press, 1999. 205–20., , ,
- 102SEER cancer statistics review 1973–1990. NIH Publication No. 93-2789. Bethesda, MD: U.S. Department of Health and Human Services, 1993., , , et al.