V. Kuenen-Boumeester,1 T.H. Van der Kwast,2 C.C. Claassen,1 M.P. Look,3 G.S. Liem,3 J.G.M. Klijn3 and S.C. Henzen-Logmans1

1Department of Pathology and 3Division of Endocrine Oncology, Dr Daniel den Hoed Cancer Centre, P.O. Box 5201, 3008 AE Rotterdam, The Netherlands; and 2Department of Pathology, Erasmus University, Rotterdam, The Netherlands.

V. Kuenen-Boumeester - Correspondence


To analyse the clinical significance of the presence of androgen receptors (AR) in breast carcinomas, clinical and histological parameters of 153 primary breast carcinomas (median follow-up 46 months) were examined. Oestrogen (ER) and progesterone receptor (PR) levels were determined in cytosol preparations using enzyme immunoassay assays and in cryostat sections by immunohistochemistry. AR and Ki-67 levels were only determined immunohistochemically. Data were analysed by uni- and multivariate models. 94/153 (61%) breast carcinomas were ER+ PR+ AR+, while 14 cases were only positive for AR. All grade III tumours (n = 17) were steroid receptor negative and 14 (76%) of these cases demonstrated high Ki-67 values suggestive of more aggressive behaviour. Strikingly, 14 ductal carcinomas negative for ER and PR were positive for AR. In univariate analysis, AR as well as ER, tumour size, lymph node status, grade and Ki-67 proved to be significant prognostic factors for disease- free survival (DFS). Multivariate analysis, however, showed lymph node status, tumour size and ER status to be the only independent prognostic factors for DFS within this model. We conclude that simple histological and cell biological parameters, including AR, can be used to select high- and low- risk patients at the time of primary surgery and can provide valuable information on treatment options.

Key words: breast cancer, androgen receptor, immunohistochemistry, prognosis, multivariate analysis.


THE IMPORTANCE of the histological subclassifkation of breast cancer as one of the indicators for tumour aggressiveness is well recognised. Histological subtypes, such as tubular, cribriform and colloid cancers, are known to have a good prognosis, while not otherwise specified (NOS) ductal carci-nomas have a relatively poor prognosis; the lobular, medullary and mixed carcinomas are considered to represent an intermediate prognostic group [1-5]. Other tumour characteristics, such as lymph node status and histological grading, are considered to be of major importance [1, 2, 5-9]. The prognostic impact of additional parameters, such as proliferative activity [10-12] and the expression of oestrogen (ER) and progesterone receptors (PR), is also well established [13, 14]. A relationship between subtypes of breast cancer and proliferative activity has been demonstrated in both histological [12, 15] and cytological material [15].

The clinical significance and functional role of androgen receptor (AR) expression in breast cancer is less well defined. In a previous study [16] on human breast cancer, we found a close relationship between AR, ER and PR expression. AR was detectable in 76% of the 100 cases of breast cancer investigated. Nine per cent of the tumours expressed AR as the only sex steroid receptor. As various studies [17, 18] have reported that the combined use of androgen and anti-oestrogen therapy has therapeutic advantages over anti-oes-trogen treatment alone, AR determination may give additional information regarding the response to different endocrine treatment modalities.

In this study, we report on the prognostic impact of the aforementioned tumour characteristics, with special emphasis on the relationship between AR and these factors.

Using these parameters, we performed uni- and multivariate analyses. In addition, we correlated ER and PR detected immunohistochemically with ER and PR levels measured by a biochemical method based on the enzyme immunoassay (EIA).

Patients and methods


This study was performed on a group of 153 patients (mean age 55 years, range 29-88 years) with primary breast cancer who underwent either breast conserving surgery (n = 71 ) or modified mastectomy (n = 82) with axillary lymph node dissection, from 1988 to 1991 in the Dr Daniel den Hoed Cancer Centre (median follow-up 46 months, range 12-73 months). Patients’ age and menopausal status were recorded at the time of primary surgery. The tumours were graded according to Bloom and Richardson, with minor modifications as described by Elston in Page and Anderson [19]. Histological typing was performed following the WHO classification [20], modified according to Page [19]. Only patients without signs of distant metastasis at the time of surgery, with known immunohistochemically determined ER, PR, and AR status and known follow-up were included in this study. All patients were routinely examined every 3-6 months during the first 5 years. In the follow-up period, 46 patients (30%) showed evidence of recurrent disease, and 25 of these women died. 3 patients died without recurrent disease.


Biochemistry. In 133/153 cases, material was also available for cytosolic ER and PR assays. As described previously [14], tumour tissue (0.4-0.8 g) was pulverised and homogenised as recommended by the EORTC for processing breast tumour tissue for cytosolic ER and PR determinations. The homogenate was centrifuged for 30 min at 100000 g at 4°C, and the supernatant fraction (cytosolic extract) was used for ER and PR determination by enzyme immunoassays (ER-EIA and PGR-EIA kits, Abbott, Chicago, Illinois) (cut-off values 10 fmol/mg protein).


The immunohistochemical methods used have been previously described [15, 16]. In short, representative tissue samples were snap-frozen in liquid nitrogen and stored at -70°C until use. Proliferative activity was assessed with MAb Ki-67 (DAKO, Glostrup, Denmark). Cryostat sections, 4 pm thick, were air-dried and fixed in acetone for 10 min, after which the indirect immunoperoxidase technique was used for visualising Ki-67. For immuno- staining of ER, PR and AR, the cryostat sections were fixed with formalin (4%) diluted in phosphate-buffered saline (PBS). Incubation was performed with the ERICA or PRICA kits for ER and PR, respectively (Abbott). For immunostain- ing of AR, the MAb F39.3 was used, specific to a unique epitope in the IV-terminal domain of the human AR molecule [21]. Specific binding of ER and PR was visualised using the peroxidase-antiperoxidase (PAP) technique; for AR the s-ABC (streptavidin-biotin-enzyme complex) technique was applied. All immunostained sections were counterstained with Mayer’s haematoxylin for 1 min. Control sections consisted of known positive and negative specimens identified by ligandbinding assay. In addition, for negative control sections, the primary antibody was replaced with PBS or non-immune ascites fluid.


The percentage of ER-, PR- and AR-posi- tive tumour cells was calculated by counting the number of positive cells in a total of 300 cells in three different areas of the tumour [16]. A staining percentage of less than 10 was regarded as negative.

The Ki-67 score was assessed by counting 300 cells in the areas with the highest proliferative activity, as described previously [15]. Arbitrarily, tumours with a Ki-67 score equal to or over 20% were defined as tumours of high proliferative activity. In 5 cases, no Ki-67 could be assessed, due to inadequate staining results (n = 4) and loss of material (n = 1).


Spearman rank correlations (rs) were used to study associations between continuous variables. The associations between continuous and grouped variables were tested using the Wilcoxon rank sum or Kruskal-Wallis tests and the associated trend test, when appropriate. When patients died of unknown causes, they counted as failures for disease-free survival (DFS) at time of death (n = 3). The Cox proportional hazards model was applied for both univariate and multivari-ate analyses, using the associated likelihood ratio test to test for differences. Cox regression analyses are summarised in Table 4 by the relative relapse rates.

Relapse-free survival probabilities were calculated by the actuarial method of Kaplan and Meier. The log rank test for trend was used for curves with three ordered groups. For all tests, we considered a two-sided P-value of less than 5% as significant. Because of the relatively short follow-up period of the patients and, as a result the low number of events in overall survival, we chose to focus only on DFS.


Patient and tumour characteristics

Patient and tumour characteristics are summarised in Table 1. The histological subtypes included 114 ductal, 14 lobular, five colloid, two tubular, one cribriform, 14 mixed, two medullary and one metaplastic carcinomas. Of the pre-menopausal women with positive lymph nodes (n = 36), 27 received adjuvant chemotherapy, 1 hormonal therapy and 8 received no adjuvant therapy. Postmenopausal women with positive lymph nodes (n = 38) received adjuvant hormonal therapy in 10 cases, adjuvant chemotherapy in 13 cases, while 15 received neither chemo- nor hormonal therapy.

Table 1. Patient and tumour characteristics.


Correlation between ER and PR data assessed immunohisto-chemically and cytosolic ER and PR levels measured with EIA

Immunohistochemically detectable nuclear ER and PR were present in variable percentages of the tumour cells, but not in stromal cells.

The Spearman rank correlation between data obtained immunohistochemically and data obtained by EIA wasrs= 0.66 (P<0.0001) for ER, andrs= 0.74 (P<0.0001) for PR. When dichotomised, a discordance was observed in 11 cases for ER and in 20 cases for PR.

Association between expression of three steroid receptors and Ki-67 score

The relationship between the immunohistochemical expression of the three examined steroid receptors is shown in Table 2. In 94 cases (61%), expression of all three receptors was observed. Interestingly, in 14 cases (9%) only AR expression was found.

Table 2. Relationship between the immunohistochemical expression of the three steroid receptors Positivity for AR.


An inverse relationship between AR, ER and PR expression and the Ki-67 score was demonstrated; a high Ki-67 score was significantly (P<0.01) associated with low receptor values (data not shown).

Correlation between steroid receptor expression, Ki-67 score and histological typing and grading

A variable steroid receptor expression and Ki-67 score were found in the group of NOS ductal carcinomas (Table 3). Of the 17 receptor-negative tumours, 14 were NOS ductal carcinomas. Strikingly, 13 of these 14 cases expressed high (3=20) Ki-67 values. In addition, the 14 cases negative for ER and PR and positive for AR were all NOS ductal carcinomas.

Table 3. Correlation between steroid receptor expression, Ki-67 and histological typing and grading.


The colloid, cribriform and tubular carcinomas combined ER, PR and AR expression with low (< 20) Ki-67 values (Table 3). Figure 1 shows this variation in staining for a colloid carcinoma.


Figure 1. Comparison of Ki-67 and androgen receptor (AR) immunostaining of frozen sections of the same colloid breast carcinoma (magnification x40). (a) Only few tumour cells (darker nuclei) show positive staining of Ki-67. (b) Positive nuclear staining (darker nuclei) for AR is present in the majority of tumour cells.

The mixed (n = 14) and lobular (n = 14) type carcinomas formed an intermediate group (Table 3); only one mixed type carcinoma lacked all three receptors and 2/14 lobular (14%) and 4/14 mixed (29%) carcinomas expressed 3=20% Ki-67 positive cells. In contrast, the medullary (n = 2) carcinomas combined high Ki-67 scores with the absence of steroid receptor expression, and the metaplastic carcinoma combined receptor negativity with a low Ki-67 score (Table 3).

The correlation between immunohistochemically detected steroid receptor expression and grade is presented in Table 3. Grade I and II carcinomas showed a higher percentage of receptor-positive cases in comparison to grade III carcinomas. Moreover, all receptor-negative cases (n = 17) were grade III cancers, in 14 cases characterised by a high percentage of Ki- 67 positive cells. The 54 grade III tumours consisted predominantly of NOS ductal carcinomas (n = 49).

Correlations between steroid receptors, Ki-67 and clinical parameters

No significant association between immunohistochemically detected steroid receptor expression and lymph node status was observed. A significant correlation was seen for ER and age (rs = 0.28, P<0.01) and an inverse relationship for tumour size and PR (Kruskal-Wallis associated trend test, P= 0.03) was seen. A borderline significant inverse relationship was found for Ki-67 score and agers= -0.16, P= 0.05) and a significant rank correlation (rs = 0.21, P= 0.01) for Ki- 67 score and lymph node status.

DFS according to various parameters

At 60 months, the DFS of the 153 patients was 64.7% with 16 patients at risk. The results of univariate analysis (Table 4) indicated an increased risk of relapse for patients with larger tumours, with high grade tumours (Figure 2a) and with more than three lymph nodes affected (Figure 2b). Similar findings were observed when patients were stratified according to ER positivity (P = 0.008), AR expression (P= 0.043) and Ki-67 score (P = 0.052). Not surprisingly, ELA-ER showed a similar correlation with DFS as did immunohistochemically detected ER (Figure 3).


Figure 2. (a) Disease-free survival curve stratified according to grade (P - 0.01). (b) Disease-free survival curve stratified according to lymph node status (P = 0.02).

Table 4. Cox univariate and multivariate analysis in 147* patients with primary breast cancer.


Figure 4 gives an indication of the DFS according to the various immunohistochemically determined receptor combi-nations. For DFS, the patients with ER+, PR+, AR+ tumours had the best prognosis, and the group with a combination of negative AR, ER and PR had the worst prognosis (P= 0.026), but the number of patients was low (n = 17).

The results of the Cox regression analysis are shown in Table 4. Relative hazard rates (HR), the 95% confidence intervals (95% Cl) and P values are given for both the uni-variate and multivariate analyses. The results of both the initial and final multivariate models are shown. The addition of adjuvant therapy, which was not statistically significant in the univariate analysis, had no influence on the estimations of the initial model. It appeared that tumour size and nodal status were the major prognostic factors, with ER status showing an additional prognostic value. Given these factors, the others showed no statistically significant associations in the final model. Estimations of the relative HR of tumour size (HR 1.70, 95% Cl 1.15-2.51) and immunohistochemically detected ER (HR 0.36, 95% Cl 0.19-0.69) were not influenced by adding adjuvant therapy as an indicator variable to the final model as compared to Table 4.


In recent years, the EIA has become an alternative method to the conventional dextran-coated charcoal (DCC) assay. Like others [22, 23] we found good correlation (P<0.0001) between values obtained by EIA and those using immunohis- tochemistry. In addition, the prognostic impact of ER/PR detected by either method was similar (Figure 3). The observed discrepancies between EIA and immunohistochem-ical determination may be partly related to variations in the proportion of tumour tissue in the specimens examined, and partly to the presence of receptor-positive benign epithelium in EIA samples, resulting in false positive ER and PR values.


Figure 3. Disease-free survival curve for oestrogen receptors (ER) according to the method of detection.

Loss of antigenicity due to material processing, especially in the case of the known labile PR, may give an additional explanation for the discordances. No biochemical data for AR were available for this material, but Ruizeveld and associates [24] observed that immunoreactivity with this antibody was generally consistent with earlier biochemical and autoradio- graphical data. As primary tumours are of increasingly smaller size at primary surgery, due to breast cancer screening pro-grammes, immunohistochemical measurements of steroid expression may, in the future, become an acceptable alterna-tive to the binding assay, particularly as monoclonal steroid receptor antibodies for use on paraffin sections have become available.

The inverse relationship of Ki-67 positivity with ER, PR and AR expression was similar to the results obtained in other studies for ER and PR [10, 25, 26]. However, Wintzer and associates [27] found no correlation between Ki-67 and ER status, and an inverse relationship between Ki-67 and PR; Gasparini and associates [28] made the opposite observation. These controversies may partly be explained by differences in the assays used and partly by differences in cut-off levels. Isola [29] correlated AR with proliferation as determined by the S-phase fraction, but found only a weak inverse association. The relationship between pure histological parameters, such as typing, grading and the biological behaviour of breast carcinomas, has already been emphasised [1-5]. As the aggressiveness of a tumour may also be reflected by the presence or absence of a number of other tumour characteristics, it seemed logical to relate the histological parameters to these markers. In agreement with an earlier study [15], we found that the Ki-67 defined proliferation index was related to particular types of breast cancer (see Table 3), and that most cases with high Ki-67 positivity were found in the group of NOS ductal carcinomas. No significant correlation was demonstrated between either Ki-67 expression or steroid receptor expression and grade. Yet all steroid receptor negative tumours (n = 17), that is AR-, ER-, PR-, appeared to be grade III tumours. Moreover, 14 cases of this group appeared to be NOS ductal carcinomas and 13 demonstrated high (s= 20) Ki-67 activity. These findings emphasise the view that lack of all three steroid receptors, combined with high proliferative activity and grading, are associated with a more aggressive tumour behaviour. This is substantiated by the DFS data (Figure 4). In contrast, and in line with published results [1-5, 25], we found that the colloid, tubular and cribriform tumours combined steroid receptor positivity, including AR, with low Ki-67 values and low grading (Table 3). In addition, consistent with the view that the group of mixed and lobular carcinomas represent a group of tumours with intermediate prognosis, we observed that only one of 28 mixed and lobular type carcinomas lacked all three steroid receptors (Table 3) and only six of 28 contained s=20% Ki-67 positive tumour cells. The same relationship was found with respect to grading: only two of 28 mixed and lobular type carcinomas were grade III tumours.


Figure 4. Disease-free survival curves for patients with tumours of different oestrogen/progesterone/androgen receptor (ER/PR/AR) phenotypes (P= 0.026). Numbers in parentheses show failures/total amount of patients in each group.

Teulings and associates [30] stated that the positive response to treatment with high-dose progestin (megestrol acetate) in a group of patients with metastatic breast cancers was determined by the AR level. Moreover, Hackenberg and colleagues [31] demonstrated in vitro that the progestin med-roxyprogesterone acetate inhibits proliferation of a ER- and PR-negative cancer cell line via AR. Therefore, AR positivity may have additional therapeutic consequences, particularly in the group of carcinomas with high proliferative activity, and with the presence of only the AR sex steroid receptor, as was observed for 14 NOS ductal carcinomas. Unfortunately, in our series, the number of patients receiving hormonal treat-ment for advanced disease was too small to study this relationship.

The prognostic impact of tumour size, lymph node status, ER, AR, Ki-67 and grade in univariate analysis emphasises the importance of these parameters. In the multivariate analy-sis, we confirmed [6, 26] that lymph node status, tumour size and ER expression were the only independent prognostic factors and that neither AR nor Ki-67 offered additional discriminating ability.

confirmed [6, 26] that lymph node status, tumour size and ER expression were the only independent prognostic factors and that neither AR nor Ki-67 offered additional discriminating ability. In contrast to the large studies of Dixon and associates [2] and Pereira and associates [5], our study included a relatively small group of 34 patients with non-ductal carcinomas. Consequently, a reliable statistical analysis of these subsets of breast cancers was not possible. Nevertheless, we feel that our results emphasise the importance of histological typing and grading of individual tumours. Moreover, they indicate that these histological parameters in combination with a few common cell biological parameters, including AR, may help in the initial selection of high- and low-risk patients. Further refinement in choice of treatment for each individual patient may then be realised by using these parameters.


We thank Mrs M.E. Meijer-Van Gelder for her assistance in collecting the clinical data of the patients included in this study, and W.L.J. Van Putten, statistician of the Dr Daniel den Hoed Cancer Centre, for his contribution to the final preparation of the manuscript. Supported by the Dutch Cancer Society (grants NKB 91-15 and 92-04).


Simpson JF, Page DL. Prognostic value of histopathology in the breast. Semin Oncol 1992, 19, 254-262. 

Dixon JM, Page DL, Anderson TJ, et al. Long-time survivors after breast cancer. BrJ Surg 1985, 72, 445-448. 

Fisher ER, Redmond C, Fisher B, Bass G. Pathologic findings from the national surgical adjuvant breast and bowel projects (NSABP). Cancer 1990, 65, 2121-2128. 

Ellis IO, Galea M, Broughton N, Locker A, Blarney RW, Elston CW. Pathological prognostic factors in breast cancer. II. Histological type. Relationship with survival in a large study with longterm follow-up. Histopathology 1992, 20, 479-489. 

Pereira H, Pinder SE, Sibbering DM, et al. Pathological prognostic factors in breast cancer. IV: Should you be a typer or a grader? A comparative study of two histological prognostic features in operable breast carcinoma. Histopathology 1995, 27, 219-226. 

Fisher B, Bauer M, Wickerham DL, Redmond CK, Fisher ER. Relation of number of positive axillary nodes to the prognosis of patients with primary breast cancer. Cancer 1983, 52, 1551-1557. 

Fisher ER, Redmond C, Fisher B. Histologic grading of breast cancer. Pathol Annu 1980, 15, 239-251. 

Clark GM, Sledge GW, Kent Osborne C, McGuire WL. Survival from first recurrence: relative importance of prognostic factors in 1015 breast cancer patients. J Clin Oncol 1987, 4, 55-61. 

Elston CW, Ellis IO. Pathological prognostic factors in breast cancer. I. The value of histological grade in breast cancer: experience from a large study with long-term follow-up. Histopathology 1991, 19, 403-410. 

McGurrin JF, Doria MI, Dawson PJ, Karrison Th, Stein HO, Franklin WA. Assessment of tumor cell kinetics by immunohisto- chemistry in carcinoma of breast. Cancer 1987, 59, 1744-1750. 

Tubiana M, Pejovic MH, Chavandra N, Contessa G, Malaise EP. The longterm prognostic significance of the thymidine labeling index in breast cancer. IntJ Cancer 1984, 33, 441-445. 

Meyer JS, Bauer WC, Rao BR. Subpopulations of breast carcinoma defined by S-phase fraction, morphology and estrogen receptor content. Lab Invest 1978, 39, 225-235. 

Thorpe SM. Estrogen and progesterone receptor determinations in breast cancer. Acta Oncol 1988, 27, 1-18. 

Foekens JA, Portengen H, Van Putten WLJ, et al. Prognostic value of estrogen and progesterone receptors measured by enzyme immunoassays in human breast tumor cytosols. Cancer Res 1989, 49, 5823-5828. 

Kuenen-Boemeester V, Van Der Kwast ThH, Van Laarhoven HAJ, Henzen-Logmans SC. Ki-67 staining in histological sub- types of breast carcinoma and fine needle aspiration smears. J Clin Pathol 1991, 44, 208-210. 

Kuenen-Boumeester V, Van Der Kwast ThH, Van Putten WLJ, Claassen C, Van Ooijen B, Henzen-Logmans SC. Immunohistochemical determination of androgen receptors in relation to oestrogen and progesterone receptors in female breast cancer. Int J Cancer 1992, 52, 581-584. 

Tormey DC, Lippman ME, Edwards BK, Cassidy JG. Evaluation of tamoxifen doses with and without fluoxymesterone in advanced breast cancer. Ann Intern Med 1983, 98, 139-144. 

Ingle JN, Twito DI, Schaid DJ, et al. Combination hormonal therapy with tamoxifen plus fluoxymesterone versus tamoxifen alone in postmenopausal women with metastatic breast cancer. Cancer 1991,67, 886-891. 

Page DL, Anderson TJ. Diagnostic Histopathology of the Breast. New York, Churchill Livingstone, 1987, Chapter 17, 303-307. 

World Health Organisation. International Histological Classification of Tumours, No. 2, Histological Typing or Breast Tumours, 2nd edn. Geneva, World Health Organisation, 1981. 

Zegers ND, Claassen E, Neelen C, et al. Epitope prediction and confirmation for the human androgen receptor: generation of monoclonal antibodies for multi-assay performance following the synthetic peptide strategy. Biochim Biophys Acta 1991, 1073, 23-32. 

Henzen-Logmans SC, Fieret EJH, Bems EMJJ, Van Der Burg MEL, Klijn JGM, Foekens JA. Ki-67 staining in benign, borderline, malignant primary and metastatic ovarian tumors: correlation with steroid receptors, epidermal-growth-factor receptor and cathepsin D. IntJ Cancer 1994, 57, 468-472. 

Heubner A, Beck T, Grill HJ, Pollow K. Comparison of immuno- cytochemical estrogen receptor assay, estrogen receptor enzyme immunoassay, and radioligand-labeled estrogen receptor assay in human breast cancer and uterine tissue. Cancer Res 1986, 46 (Suppl), 4291s-4295s. 

Ruizeveld de Winter JA, Trapman J, Vermey M, Mulder E, Zegers ND, Van der Kwast ThH. Androgen receptor expression in human tissues: an immunohistochemical study. J Histochem Cytochem 1991, 39, 927-936. 

Helin HJ, Helle MJ, Kallioniemi O, Isola JJ. Immunohistochemical determination of estrogen and progesterone receptors in human breast carcinoma. Cancer 1989, 63, 1761-1767. 

Di Stefano D, Mingazzini PI, Scucchi L, Donnetti M, Marinozzi V. A comparative study of histopathology, hormone receptors, peanut lectine binding, Ki-67 immunostaining and nucleolar organizer region associated proteins in human breast cancer. Cancer 1991, 67, 463-467. 

Wintzer HO, Zipfel I, Schulte-Mônting J, Hellerich U, Von Kleist S. Ki-67 immunostaining in human breast tumors and its relationship to prognosis. Cancer 1991, 67, 421-428. 

Gasparini G, Dal Fior S, Pozza F, Bevilacqua P. Correlation of growth fraction by Ki-67 immunohistochemistry with histologic factors and hormone receptors in operable breast carcinoma. Breast Cancer Res Treat 1989, 14, 329-336. 

Isola JJ. Immunohistochemical demonstration of androgen receptor in breast cancer and its relationship to other prognostic factors. J Pathol 1993, 170, 31-35. 

Teulings FAG, Van Gilse HA, Henkelman MS, Portengen H, Alexieva-Figusch J. Estrogen, androgen, glucocorticoid and progesterone receptors in progestin-induced regression of human breast cancer. Cancer Res 1980, 40, 2557-2561. 

Hackenberg R, Hawighorst T, Filmer A, Huschmand Nia A, Schulz K. Medoxyprogesterone acetate inhibits the proliferation of estrogen- and progesterone-receptor negative MFM-223 human mammary cancer cells via the androgen receptor. Breast Cancer Res Treat 1993, 25, 217-224.