06 (2007) 325–333www.elsevier.com/locate/ygyno

Gynecologic Oncology 1

Relationship of estrogen and progesterone receptors to clinical outcomein metastatic endometrial carcinoma: A Gynecologic

Oncology Group Study

Meenakshi Singh a, Richard J. Zaino b, Virginia J. Filiaci c, Kimberly K. Leslie d,⁎

a Department of Pathology, The University of Colorado Health Sciences Center, Denver, CO 80262, United Statesb Department of Pathology, Hershey Medical Center, Hershey, PA 17033, United States

c Gynecologic Oncology Group Statistical and Data Center, Roswell Park Cancer Institute, Buffalo, NY 14263, United Statesd Departments of Obstetrics and Gynecology and Biochemistry and Molecular Biology, the University of New Mexico, Albuquerque, NM 87131, United States

Received 18 January 2007Available online 25 May 2007

Abstract

Introduction. The goal of this study was to explore the relationship between the expression of hormone receptors in metastatic endometrialtumors and clinical response to daily tamoxifen citrate and intermittent weekly medroxyprogesterone acetate.

Study design. Patients with measurable recurrent or advanced endometrial cancer were enrolled on a clinical trial, Gynecologic OncologyGroup Study 119. A pretreatment tumor biopsy was obtained and subjected to immunohistochemical analyses. Estrogen receptor-α (ER-α) andprogesterone receptor (PR) were assessed on frozen tissues, and PR isoforms A and B were detected on fixed tissues. The receptors were scoredusing a semi-quantitative HSCORE, with a cut off greater than 75 considered positive.

Results. Of the 60 eligible patients, 45 had evaluable tissues for all receptors. For ER, 40% of the cases were positive; for PR, 45% were positive.The sub-cellular distribution of PRA was exclusively nuclear, and 16% of the tumors demonstrated positive staining. PRB was nuclear andcytoplasmic, with 22% of the tumors staining for nuclear PRB and 36% of the tumors staining for cytoplasmic PRB. ER and PR from frozen tissuesand PRA and cytoplasmic PRB from fixed tissues significantly decreased with increasing tumor grade. The co-expression of ER-αwith PR from thefrozen tissues (r = 0.68, p b 0.001) and PRA (r = 0.58, p b 0.001) from the fixed tissues was statistically significant. The ERHSCOREwas related toboth response and overall survival; there was no statistically significant correlation of PR with clinical response in this small number of patients.

Conclusion. ER-α measured in metastatic endometrial carcinoma tissue prior to hormonal therapy was statistically significantly related toclinical response to daily tamoxifen and intermittent medroxyprogesterone acetate.© 2007 Elsevier Inc. All rights reserved.

Keywords: Uterus; Estrogen; Medroxyprogesterone acetate; Receptors; Tamoxifen; Endometrial; Cancer

Introduction

The role of progesterone in the glandular epithelium of theendometrium is primarily antagonistic to estrogen-mediated cellproliferation [1]; this is in contrast to the breast, where proges-

⁎ Corresponding author. University of New Mexico, The Division ofMaternal-Fetal Medicine, The Department of OB/GYN2211 Lomas Blvd., NEACC-4, Albuquerque, NM 87131, United States.

E-mail address: kleslie@salud.unm.edu (K.K. Leslie).

0090-8258/$ - see front matter © 2007 Elsevier Inc. All rights reserved.doi:10.1016/j.ygyno.2007.03.042

terone mediates both proliferative and anti-proliferative effects[2,3]. Therefore, the study of progestin action in the endome-trium has particular importance because the epithelium relies onprogesterone to induce cell differentiation and to counter uncon-trolled growth. While progestins have been used with greatsuccess to reverse endometrial hyperplasia [4,5] they are notconsistently effective in the treatment of primary endometrialcancer. Progestins as single agents have been used traditionallyin the treatment of recurrent or metastatic endometrial adeno-carcinoma. However, the overall response rates range from only8% to 55% [6–11]. Progestin exposure results in a rapid and

Table 1Number and proportion of patients with positive or negative expression

Receptor HSCORE 75or less: negative

HSCORE greaterthan 75: positive

Total

ER 29 (60%) 19 (40%) 48PR 27 (55%) 22 (45%) 49PRA 42 (84%) 8 (16%) 50PRB-nuclear 39 (78%) 11 (22%) 50PRB-cytoplasm 32 (64%) 18 (36%) 50

ER and PR were measured from frozen tissue. The PR isoforms A and B weremeasured from formalin fixed tissues. Receptor data were assessed semi-quantitatively using the HSCORE. A cut-off of 75 was used to dichotomize theHSCORE data into two groups: an HSCORE that was 75 or less was considerednegative and an HSCORE over 75 was considered positive.

326 M. Singh et al. / Gynecologic Oncology 106 (2007) 325–333

predictable ligand-dependent loss of progesterone receptors(PR) in both normal and neoplastic endometrial glands [12].Two recent publications of Gynecologic Oncology Group(GOG) studies, 119 and 153, reported attempts to circumventPR down-regulation with a strategy combining tamoxifen, as anestrogen surrogate to induce PR, and intermittent progestintreatment [13,14]. However, neither of these papers provided thesteroid hormone receptor expression data needed to correlateclinical response with these variables. This report now providesinformation on the link between hormone receptor status andclinical response in one of these trials, GOG 119.

Two isoforms of PR, PRA and PRB, are expressed in humans:PRA encodes a 90 kDa protein, and PRB encodes a 120 kDaprotein. Both forms arise from alternative promoters on the samegene and can form homo (A/A, B/B) or hetero (A/B) dimericunits. The isoforms are identical except that PRB has a longer N-terminus consisting of 164 amino acids not present in PRA. Theunique PRBN-terminus encodes a third activating domain, AF-3,that confers different functional characteristics to the isoforms.PRB is a stronger transcriptional activator of many genes com-pared to PRA [15,16]; PRA also has been reported to repress PRB[17]. A number of studies suggest that a relative change in PRisoform expression may be as important as the total level of PR inthe genesis of endometrial cancer [15,16,18].

Tamoxifen is a complex nonsteroidal compound, structur-ally similar to diethylstilbestrol, which binds to the estrogenreceptor (ER). The ability of tamoxifen-bound ER to activategene transcription appears to be gene and tissue-specific anddetermines whether the drug acts as an estrogen agonist(tamoxifen-bound ER is capable of activating gene transcrip-tion) or as an antagonist (competes for ER binding with otherestrogens, but forms complexes that are incapable of inducingtranscription). The estrogen-like effects of tamoxifen on steroidhormone receptor expression in the endometrium wereexploited in this phase II clinical trial of tamoxifen plusintermittent weeks of MPA, GOG 119, where tamoxifen wasused in theory as an estrogen surrogate to induce PR andmaintain its up-regulation [13]. Patients were treated withtamoxifen citrate, 20 mg, p.o. twice daily. On alternating (even-numbered) weeks, they also received medroxyprogesteroneacetate (MPA), 100 mg, p.o., twice daily. The clinical results ofthis study have been reported previously [13].

The original aim of GOG 119 was to determine clinicalresponse to this hormonal regimen; in addition, a pretreatmentbiopsy was obtained specifically to determine whether hormonereceptors correlated with response. Therefore, in the presenttranslational study, tissues from GOG 119 were analyzed todetermine, first, whether the expression of receptors on a pre-treatment biopsy predicts for clinical outcome, and second, whatthe relationships are between ER expression and the PR isoformsA and B in advanced endometrial cancer. Interestingly, therelationship between hormone receptor expression and clinicalresponse has yet to be demonstrated prospectively in a GOG trial;thus our hypothesis was that ER and/or PR expression would berequired to achieve a clinical response. Also, as ER is generallyrequired to induce PR, we tested the hypothesis that tumors withER would also express PR.

Materials and methods

Patient population

Between 1991 and 1996, the GOG conducted a prospective phase II trial oftamoxifen combined with intermittent MPA. This trial included patients withhistologically-confirmed advanced, persistent or recurrent endometrial carcino-ma considered incurable by local therapy or refractory to local therapy. Inaddition, all patients were to have disease that wasmeasurable in 2 dimensions bypalpation or imaging. Any lesion imaged by CTscan or ultrasound was to have aminimum diameter of 3 cm. The measurable tumor to be followed from entrymust not have received radiation therapy within 3 months prior to entry. Priortherapy with cytotoxic drugs or hormonal therapy for endometrial cancer was notallowed. Normal hepatic, renal and hematologic functions, the absence ofsignificant infection and GOG performance status 0–2 were required. Patientswith past or concomitant malignancy, other than non-melanoma skin cancer,were ineligible. Awritten informed consent conforming to federal, state and localregulations was obtained from all participants prior to study entry.

Tissue requirements

Pre-treatment tumor tissue from the metastatic or recurrent lesion was requiredfor all patients for analysis of estrogen and progesterone receptors. A portion of thefresh tumor, greater than 3×5×5mm,was to be immediately placed inOCTor otherembedding media into a foil cup or on cork or cardboard, rapidly frozen andmaintained at a temperature of nomore than−20 °C.The labeled vial of frozen tissuewas to be shipped on dry ice by overnight carrier to the initial testing laboratory.

ER and PR immunohistochemistry on frozen tissue

First, ER and PR were assessed directly on the frozen specimens. Frozensections were cut at 5 μM in a cryostat, and thaw mounted onto poly-L-lysinecoated glass slides. The slideswere placed in Zamboni's fixative at 4 ° C for 15minthen rinsed twice in PBS for 5 min. Immunohistochemistry was performedmanually for both of the antibodies.

The primary antibody to ERwas theD222 clone, which recognizes ER-α, withlittle or no affinity for ER-β. For ER, the reagents provided by the manufacturer ofthe ER-ICA kit (Abbott, Chicago, IL) were used as recommended for the blockingstep and the control or the ER primary antibody, with avidin and biotin used foramplification of the reaction, using a rat peroxidase kit (Vector, Burlingame, CA).

For PR, the primary antibody used on the frozen tissue was mPRI. Mouse IgGwas used in place of the primary antibody as a negative control. Endogenousperoxidase was blocked with 1% hydrogen peroxide. and biotin was used foramplification of the reaction using the Mouse IgG Elite Vectastain kit (Vector,Burlingame, CA).

For immunolocalization of both ER and PR, 3,3-diaminbenizidine was usedas the chromogen, and the slides were coverslipped with permount mountingsolution. Non-gravid rabbit uterus was used as the positive control tissue forboth ER and PR, since there was a known, predictable variation in the intensityof staining in the glands, stroma and myometrium. Positive staining of thecontrol tissue was limited to an intranuclear location for both ER and PR.

Fig. 1. (A–C) Immunostaining for hormone receptors showing subcellulardistributions. ER (panel A) and PRA (panel B) are nuclear, while PRB (panel C)are cytoplasmic and nuclear.

Table 2Receptor HSCORE Spearman correlation statistics

ER PR PRA PRB-nuclear PRB-cytoplasm

ER 1.00PR 0.68 ⁎ 1.00PRA 0.58 ⁎ 0.63 ⁎ 1.00PRB-nuclear 0.09 −0.08 −0.15 1.00PRB-cytoplasm 0.12 0.14 0.29 ⁎⁎ 0.24 1.00

⁎ pb0.0001; ⁎⁎ p=0.04.

327M. Singh et al. / Gynecologic Oncology 106 (2007) 325–333

Isoform Immunohistochemistry on fixed and paraffin embedded tissue

When discriminatory antibodies for PRA and PRB became available andprotocols derived for the analysis of formalin fixed tissues [19], the investigatorstook portions of the original frozen tumor blocks and processed them for theseadditional studies. The frozen tissues were thawed slowly to room temperature,fixed in 4% neutral buffered formalin, processed through graded alcohols toacetone, and embedded in paraffin. Tissue slides were deparaffinized through 3changes of xylene to alcohol to water. Antigen retrieval was performed by micro-waving slides in 1 mM EDTA in a pressure cooker (Nordicware, MN) for 30 min.Following a cooling period at room temperature for 20 min, tissue was incubated

with 3% hydrogen peroxide in methanol for 30 min to quench endogenousperoxidase. Slides were rinsed in Tris buffer pH 7.6 for 10 min. The instrument(Ventana NexES) was programmed to incubate the tissue slides with primarymonoclonal antibodies against PRA+B (to common regions of PRA and PRB, butpreferentially identifying PRA) manufactured by Ventana (clone 1A6). This will bereferred to as PRA in this report. A PRB-specific antibody, Ab-6 (NeoMarkers/LabVision), has also been tested with success on the Ventana NexES. The primaryantibodywas incubated for 32min at room temperature, and the slideswerewashed.Secondary antibody, streptavidin, and chromogen incubations were programmedinto the instrument following themanufacturer's instructions. Slideswere rinsed andcounterstained with hematoxylin prior to evaluation.

Cells and controls

To test the specificity of the PRA and PRB antibodies, cells were plated onchamber slides. The cell lines employed were well-differentiated Ishikawa endo-metrial cancer cells (that express PRA and PRB) and poorly differentiatedHec50cocells (that do not express or PR). Hec50co cells were infected with adenovirusvectors encoding PRA or PRB to obtain high PR isoform expression. Antibodyspecificity for PRA and PRBwas tested using IHC and co-immunoprecipitation, aspreviously reported [19].

Semi-quantitation of IHC results

The results of immunostaining for all receptors were semi-quantified using theHSCORE, calculated based on estimates of percentages of positive stainedepithelial cells in each of 5 intensity categories (0, 1+, 2+, 3+, and 4+). TheHSCORE represents the sum of each of the percentagesmultiplied by theweightedintensity of staining, as follows: HSCORE = (i+1)π, where i = 1, 2, 3, 4, and πvaries from 0 to 100%. A threshold of N75 as a positive result has previously beendetermined by McCarty et al. to provide the maximum sensitivity and specificitywhen compared with the biochemical assay for ER [20].

In addition to the nuclear staining clearly visible, cytoplasmic staining was alsoscored because of the recent report that some PR (particularly PRB) is located in thecytoplasm in the absence of ligand [19]. The specificity of PRB cytoplasmic stainingwas ascertained by the use of cell models expressing PRB compared to PRA, where50% of the PRB introduced into cells remained cytoplasmic in the absence ofprogesterone but translocated into the nucleus when progesterone was added.

Clinical response and overall survival

A partial response was defined as a 50% or greater reduction in the product ofthe largest diameter and its perpendicular of eachmeasurable lesion on two separateexaminations at least 2 weeks apart. Additionally, there must have been nodeterioration in performance status and no new lesions. If all gross evidence ofdisease disappeared for at least 1 month, this was classified as a complete response.All other patients were classified as non-responders. Overall survival is defined asthe length of life measured from the date of entry on the study. Patients who werelast known to be alive were censored at the date of last contact.

Statistical analysis

HSCORE values for ER, PR, PRA and nuclear and cytoplasm PRB wereassessed both as continuous covariates and as dichotomous covariates using acutoff of 75 to categorize as negative versus positive receptor status. The Spearmancorrelation statistic was employed to measure the association of HSCORE data

Table 3AJonckheere–Terpstra test results of receptor and tumor grade

Receptor p-value a J–T teststatistic

Standardized teststatistic b

ER (n=48) 0.051 292.5 −1.63PR (n=49) 0.024 286.0 −1.99PRA (n=50) 0.027 310.0 −1.96PRB-nuclear (n=50) 0.391 394.0 −0.29PRB-cytoplasm (n=50) 0.045 323.5 −1.65a Monte Carlo estimate for exact J–T test one-sided p-value using 10,000

samples.b Provides information on the direction of the alternative hypothesis: a small

p-value for a negative value of the standardized test statistic supports thealternative of decreasing HSCORE with increasing grade.

Table 4Frequency of response within dichotomized receptor HSCORE category

Responders Non-responders Total number treatedwith Receptor data

ER HSCORE≤75 7 (26%) 20 (74%) 27ER HSCOREN75 9 (47%) 10 (53%) 19PR HSCORE≤75 8 (32%) 17 (68%) 25PR HSCOREN75 9 (41%) 13 (59%) 22PRA HSCORE≤75 13 (32%) 27 (68%) 40PRA HSCOREN75 4 (50%) 4 (50%) 8PRB-nuclearHSCORE≤75

13 (35%) 24 (65%) 37

PRB-nuclearHSCOREN75

4 (36%) 7 (64%) 11

PRB-cytoplasmHSCORE≤75

10 (32%) 21 (68%) 31

PRB-cytoplasmHSCOREN75

7 (41%) 10 (59%) 17

This table summarizes the response data for each receptor with an HSCOREgreater that 75 considered positive.

328 M. Singh et al. / Gynecologic Oncology 106 (2007) 325–333

between pairs of receptors. The Jonckheere–Terpstra (JT) test [23] was used to testif HSCOREvalues variedwith increasing tumor grade. A negative standardized JTtest statistic indicates receptor HSCORES decreased with increasing tumor grade.Monte Carlo estimates of the exact p-values are reported. The JT test is a non-parametric test for ordered alternatives. When ordered alternatives are of interest,the JT test is preferred to a more general test such as the Kruskal–Wallis test thattests for class differences without any regard to ordering. Monte Carlo estimates(using 10,000 samples) of the exact test p-values are reported since given thesample size, an exact test was preferred over an asymptotic test, but the exact testwas computationally intensive [24]. The relationship between receptors, tumorgrade and clinical response was analyzed using logistic regression. The effect ofeach receptor HSCORE on the odds of response was estimated as an odds ratio(OR) and reported with a 95% Wald confidence interval (CI). A receiver operatorcurve (ROC) was constructed to assess the utility of the ER HSCORE as a test topredict clinical outcome. The value of c was used to estimate the area under theROC [25]. The Kaplan–Meier method was used to estimate the probability ofsurvival within ER positive and ER negative subgroups. Cox proportional hazardsmodels were used to estimate the relative effect of receptor HSCOREs and tumorgrade on the hazard of death. Hazard ratio (HR) estimates are reported with 95%Wald confidence intervals. Martingale residuals calculated from a model withoutthe ER HSCORE covariate were plotted against ER HSCORE. Residuals greaterthan zero suggest worse than expected survival and residuals less than zero suggestbetter than expected survival. A smoothing spline function was fit to evaluate thefunctional form of the dependence of the residuals on ER HSCORE [26]. Allstatistical tests were two-sided, unless otherwise stated, with a p-value less than0.05 considered as statistically significant. These analyses are exploratory; noadjustment was made to any p-value to account for multiple comparisons.

Results

Patient evaluability

Sixty-one patients were entered onto the study. One patientwas ineligible following central pathology review which con-

Table 3BFrequency of tumor grade category

Grade 1 Grade 2 Grade 3 or not specified

ER HSCORE ≤75 7 8 14ER HSCORE N75 5 7 7PR HSCORE ≤75 6 7 14PR HSCORE N75 6 9 7PRA HSCORE ≤75 9 13 20PRA HSCORE N75 3 4 1PRB-nuclear HSCORE ≤75 8 14 17PRB-nuclear HSCORE N75 4 3 4PRB-cytoplasm HSCORE ≤75 4 14 14PRB-cytoplasm HSCORE N75 8 3 7

cluded the primary tumor was not endometrial carcinoma.Tissue was never submitted for five patients and was inevalu-able for all analyses in two patients. Additionally, tissue fromfour patients was inevaluable in the initial analysis of ER-α andPR, and tissue from three patients was inevaluable for isoformanalysis. One additional patient was inevaluable for ER-α. Twopatients were never treated and, thus, not evaluated for re-sponse. Forty-five patients had enough available tissues toperform each of the immunoassays, and of these, 43 hadclinical response data for correlation.

Percent of cells staining for receptors

Table 1 summarizes the receptor status of the pre-treatmenttumor tissues. Using an HSCORE of greater than 75, 40% of thefrozen tumors demonstrated positive staining for ER-α, while45% had positive staining for PR using the antibody mPRI.Receptor isoforms were evaluated on formalin fixed tissues: PRAwas found to be exclusively nuclear in distribution, while PRBwas both nuclear and cytoplasmic. Some tumors demonstratednuclear and cytoplasmic PRB, while other tumors demonstratedonly cytoplasmic PRB. Sixteen percent of the tumors werepositive for PRA, 22% for nuclear PRB, and 36% for cytoplasmic

Table 5Associations between receptor HSCORE and response or overall survival

Response Overall survival

Receptor OR a 95% CI C b HR c 95% CI

ER 1.007 1.001 to 1.014 0.708 0.996 0.993 to 0.999PR 1.002 0.997 to 1.008 0.573 0.997 0.994 to 1.000PRA 1.003 0.997 to 1.009 0.630 0.998 0.994 to 1.002PRB-nuclear 0.999 0.993 to 1.005 0.463 0.999 0.996 to 1.002PRB-cytoplasm 1.001 0.995 to 1.006 0.507 0.999 0.996 to 1.001a Change in relative odds of response per unit increase in receptor HSCORE.b Estimates the area under the ROC curve; values closer to 1 would indicate

better classification of responders; values closer to 0.5 indicate a uselessdiscriminator of responders versus non-responders.c Change in relative hazard of death per unit increase in receptor HSCORE.

Fig. 3. Martingale residuals for ER HSCORE.

Table 6Association between estrogen receptor HSCORE and response or overallsurvival adjusted for tumor grade

Response Overall survival

Factor OR 95% CI C HR 95% CI

ER (continuous) 1.008 1.001 to 1.015 0.692 0.996 0.993 to 1.000ER (dichotomous) 2.589 0.737 to 9.093 0.626 0.474 0.244 to 0.919

This table provides a summary of the tumor grade-adjusted associations betweenER HSCORE and response and overall survival. Adjustment was accomplishedby including two indicator variables into the models: one for grade 1 and one forgrade 2 tumors.

329M. Singh et al. / Gynecologic Oncology 106 (2007) 325–333

PRB using antibodies 1A6 for PRA and Ab-6 for PRB. Figs. 1Aand B illustrate the typical nuclear staining for ER-α and PRA,while 1C demonstrates PRB in the nucleus and the cytoplasm.

Table 2 displays the Spearman correlation statistic for each pairof receptors. The expression of ER from frozen tissues, PR fromfrozen tissues, and PRA from formalin fixed tissues are mode-rately correlated, with statistically significant (p b 0.0001) valuesbetween 0.58 and 0.68; there is a weak correlation between PRAand cytoplasmic PRB on fixed tissues (r = 0.29, p = 0.04).

Relationship between tumor grade and receptor expression

Tables 3A and 3B illustrate that as tumor grade increased,expression of ER, PR, PRA and cytoplasmic PRB decreased,indicating an inverse relationship between receptor expressionand grade.

Clinical response rates and association with receptors

Overall, the clinical study observed a 10% complete responserate and a 22% (13/58) partial response rate for a total of 33%(19/58). These results have been previously reported and dis-cussed [13]. Table 4 provides the frequency of response cate-gorized by receptor HSCORE. For ER-positive tumors, theclinical response rate was 47% compared to 26% for ER-nega-

Fig. 2. ER receiver operator curve (ROC) for classifying response.

tive tumors. For frozen tissues with positive PR (HSCORE N 75)0, 41% responded. From fixed tissues, 50% of patients withPRA responded. Response among patients with nuclear PRB(HSCORE N 75) was 36%; whereas, among patients with cyto-plasm PR, 41% responded.

Except for ER, no correlation between receptor expression andclinical response was found. The positive correlations betweenER HSCORE and response and between ER HSCORE andsurvival are summarized in Table 5. Adjusted for tumor grade,there is a significant effect of ER HSCORE on the log odds ofresponse when analyzed as a continuous covariate (OR: 1.008,95% CI 1.001 to 1.015) (Table 6). A change in ER HSCORE of100 translates into a predicted 222% increase in the odds ofresponse. As theHSCOREvaries from 0 to 500, an increase in thescore of 100 would roughly translate into a 20% up-regulation ofreceptor staining. Fig. 2 is a receiver operator curve for the ER

Fig. 4. Kaplan–Meier Survival plot grouped by ER status. An HSCORE cut offof 75 was used to classify tissues as ER-positive or ER-negative.

330 M. Singh et al. / Gynecologic Oncology 106 (2007) 325–333

HSCORE as it relates to response. The estimated area under thecurve is 0.708. Fig. 3 exhibits the relationship between ERHSCORE and the hazard of death using martingale residuals. AKaplan–Meier plot of OS categorizing ER HSCORE as aboveor below 75 is shown in Fig. 4. The estimated median survivalamong the patients with ER HSCORES equal to or below 75 is8 months compared with 19 months for patients withHSCORES above 75. The estimated reduction in the deathhazard ratio associated with ER HSCORES greater than 75relative to HSCORES 75 or less adjusted for tumor grade is53% (HR: 0.47, 95% CI 0.24 to 0.92). When evaluating theeffect of ER HSCORE as a continuous covariate on thehazard of death, there is a predicted 33% [calculated as follows:(1− (0.996100))×100%] reduction in the death hazard associat-ed with an increase of 100 in ER HSCORE (HR 0.996, 95% CI0.99 to 1.00).

Discussion

Single agent progestins have been used in the treatment ofrecurrent and metastatic endometrial carcinoma. Upon treat-ment with a progestin, there is activation of PR as a transcriptionfactor, and the induction of genes associated with endometrialdifferentiation follows [21–23]. However, the ligand-dependentactivation of PR is closely linked to the ligand-dependent down-regulation of PR, a process that occurs upon PR phosphoryla-tion that is both a prerequisite for transcriptional activation anda preliminary step for PR ubiquitination and destruction in theproteasome [12,24,25]. Therefore, long-term continuous expo-sure to progestins sets the stage for loss of PR effects.

The strategy chosen in this study, GOG 119, was to treatcontinuously with tamoxifen, a partial estrogen agonist, which ina model system stimulated PR as potently as estradiol but withless growth stimulation. Since continuous treatment with pro-gestin alone down-regulates PR, it was hypothesized that inter-mittent treatment withMPAwould permit tamoxifen to induce PRand enhance the effect of progestin therapy. This phase II clinicaltrial demonstrated the tolerability and activity of combining theestrogen-like drug tamoxifen with intermittent progestin. Patientswere treated with tamoxifen citrate, 20 mg, p.o. twice daily. Onalternating (even-numbered) weeks, they also received MPA,100 mg, p.o., twice daily [13]. GOG 119 also allows the eva-luation of pretreatment tumor receptor expression and permits ananalysis of whether ER or PR expression correlates with eachother, with tumor grade, and with clinical outcome. These resultsare reported in this study.

The presence of hormone receptors in endometrial cancerand the link to clinical response has been studied extensively.Initial studies utilized cytosolic ligand binding assays in whichprotein extracts from ground tissue, including contaminatingsurrounding stroma, were analyzed. These early techniques hadlimitations, as previously reviewed [26,27]. Despite the draw-backs of the early technology, significant correlations betweenreceptors and clinical outcome, survival, and response toprogestins were documented for patients with early and advancedendometrial cancer in many studies [28–36]; however, othersfailed to show that receptors predicted for improved survival [37].

The development of monoclonal antibodies to ER and PR andtheir isoforms has permitted immunohistologic localization ofhormone receptors in endometrial cancers and demonstrated theirheterogeneous distribution within primary tumors and when pri-mary tumors are compared with their metastases [38]. While thepresence of antigenicity does not guarantee the functionality ofthe receptor or its downstream pathways, a number of studieshave shown that ER, PR, or both predict for an improved clinicaloutcome compared to receptor negative tumors [39–43]. A recentevaluation of ER and PR in 164 endometrial carcinomas found atrend towards improved survival for patients with ER and PRpositive tumors, but the level did not reach statistical significance[44]. Another report failed to find a relationship between the ERand PR status of tumors and recurrence in 62 high risk endo-metrial cancers [45]. These studies have limitations in that thetreatments were not uniform; therefore, the strength of the cor-relation of receptor status with clinical outcome was confoundedby different therapies. Nevertheless, despite the early enthusiasmfor directing care based upon the presence of ER and/or PR, thereceptor status of endometrial tumors is not routinely evaluated inclinical practice in the United States.

In the present study, ER and PR were assessed initially infrozen tissue using a method available in the early years of thestudy. Later, it became possible to discern two isoforms of PR, Aand B, by immunostaining on formalin-fixed tissues [19,46].Both methods are reported in this study and correlated withclinical response. Also, cytoplasmic as well as nuclear PRA andB staining was evaluated because of the recent report that PRB isoften cytoplasmic as well as nuclear, particularly in the absenceof ligand [19]. Previous immunohistochemical studies oftenignored cytoplasmic staining because the potential non-genomicfunctions of such receptors and their ability to rapidly translocateto the nucleus with the addition of ligand were not understood.

Despite the fact that this study used two immunohistochem-ical methodologies to detect receptors, concordance between PR(frozen tissues, assessed by the antibody mPRI) and PRA (for-malin-fixed tissues, assessed by the antibody 1A6) was reas-suringly high and statistically significant. However, PR fromfrozen tissues did not correlate well with PRB, identified usingthe antibodyAb-6with an epitope in the unique N-terminus of B.This indicates that, like many of the currently available PRantibodies raised against shared regions of PRA and PRB, mPRIdoes not have a high affinity for PRB. Therefore, to reliablyidentify this isoform, an antibody to the unique N-terminus ofPRB must be employed, as previously reported [19,47].

The single best predictor of response in the present study wasER, which was significantly associated with clinical responseand survival. Also, ER expression was highly correlated with theexpression of PRA, but not PRB. Nevertheless, 26% of thetumors without ER and 32% of the tumors without PRA or PRB,as determined on a pre-treatment biopsy, responded to therapy.

The HSCORE cutoff of 75 recommended by McCarty et al.[20] to categorize those with positive versus negative receptorexpression appears reasonable within this study. The fitted splinefunction in Fig. 3 crosses from positive to negative residuals veryclose to 75. However, the relationship appears linear inHSCORE.Additionally, the difference in survival between ER receptor

331M. Singh et al. / Gynecologic Oncology 106 (2007) 325–333

expression groups, after adjusting for tumor grade, is apparent.Furthermore, the maximum difference between sensitivity and 1-specificity along the ROC for predicting response is associatedwith ER HSCORE equal to 40 (sensitivity = 0.75, specifici-ty = 0.63). The difference at ER HSCORE equal to 70 is similar(sensitivity = 0.69, specificity = 0.67).

An interesting question to consider is why some tumorswithout receptors on the pre-treatment biopsy responded. Thereare two possibilities: first, tamoxifen induced the steroid receptorsas expected based upon the study design, or second, tamoxifenand/or MPA worked through non-receptor mediated pathways.The core rationale for this clinical trial was that tamoxifen, as anestrogen surrogate, would induce ER and PR [48,49] Tamoxifenis well known to act as an estrogen agonist in endometrial cancers[50]. Therefore, as shown in previouswork in the rat uterus [51], itis likely that ER and PR expression was induced in some tumorsinitially found to be devoid of receptors on the pre-treatmentbiopsy. This could explain the findings that 26% of ER negativetumors and 32% of PR negative tumors (as determined prior totreatment) responded to therapy. While this trial required a pre-treatment biopsy for entrance, no post-treatment biopsy wasobtained— such a biopsy would have been for research purposesonly and is difficult to justify to human subject review boards.Therefore, we do not know whether tamoxifen treatment didindeed up-regulate PR and/or ER. The alternative explanation isthat tamoxifen or MPA had receptor independent anti-tumoreffects. Particularly in the case of tamoxifen, such effects areproposed to occur through inhibition of protein kinase C, which isthe rationale for the use of high dose tamoxifen in the treatment ofmalignant glioma [52].

A second question surrounds the significance of the cyto-plasmic PRB found in many tumors. The presence of PR in thecytoplasm indicates a potential non-genomic function for thisreceptor. PR is known to interact with signal transduction path-ways that initiate or enhance responses in conjunction withmembrane-bound growth factor receptors and G-protein recep-tors. For example, PR associates with p42MAPKand is involvedin phosphatidylinositol 3-kinase signaling in Xenopus oocytes[53]. In addition, PRs contain a proline-rich motif within theshared A domain of PRA and PRB which interacts directly withSH3 domains of c-Src and its family member Hck [54]. BothPRA and PRB demonstrate this interaction in vitro, but sincePRA is largely restricted to the nucleus in endometrial cancercells, it is predicted that cytoplasmic signaling events dependupon PRB. It has been suggested that a dynamic situation existswhereby receptors diffuse into the cytoplasm and are constantlyand actively transported back to the nucleus [55]. Themechanism(s) underlying PR cytoplasmic to nuclear shuttlingare currently under investigation and hopefully will shed newlight on the functions of the PRA compared to the PRB isoform.Given the difference in localization of PRA compared to PRBnoted on clinical specimens in this study, these investigations areparticularly relevant.

In conclusion, recent calls for the incorporation of translationalstudies in clinical trials have been heard [56]; this study inves-tigated the relationship between ER and PR expression, tumorgrade, and clinical response in patients with advanced endome-

trial cancer. In this study, ER expression was significantly andpositively associatedwith PR and PRAexpression, aswas PRA tocytoplasmic PRB. No relationship was appreciated between ERand PRB expression, indicating an uncoupling of the expected co-expression of ER with both PR isoforms in these advancedtumors. In addition, receptor expression was down-regulated astumor grade increased, confirming our previous observations andthose of others [15,16,57].

The principal strength of the study is the availability of pre-treatment tissue receptor data linked to clinical response to astandardized hormonal regimen; however, there were limitationsto the study, including its relatively small size and the lack ofpost-treatment biopsies to follow the effects of therapy onreceptors. Nevertheless, this study demonstrated that ERreceptor status, as reported by the semi-quantitative HSCORE,was related to clinical response and survival in patients treatedwith tamoxifen and intermittent MPA, but PR expression did notsignificantly correlate with clinical outcome. However, it ispossible that the small number of tumors in this study thatexpressed PRA, only 16%, may have prevented us from detec-ting PRA as a prognostic marker even though it was statisticallylinked to ER expression.

Importantly, the median survival of patients with an ERHSCORE above 75 on this hormonal regimen was 19 months,which is comparable to the 15- to 18-month median survival timereported with doxorubicin, cisplatin and paclitaxel with granulo-cyte colony-stimulating factor support and paclitaxel and cisplatinwith granulocyte colony-stimulating factor support [58]. From theGOG 119 cohort, 40% of the patients expressed ER at levelsabove an HSCORE of 75, indicating that many patients couldpotentially benefit from hormonal therapy. These data demon-strate the potential value of routine characterization of endome-trial tumors with respect to receptor expression. Therefore, theauthors recommend ER and PR testing for all endometrial tumorsbecause the results may impact the choice of therapy. For tumorswith high ER expression, it is possible that hormonal therapymayachieve response rates equal to that of chemotherapy with fewerexpected side effects, thereby improving quality of life. Furtherinvestigation is warranted to determine whether this result isspecific to the use of tamoxifen combined with intermittent MPA,or is applicable to hormonal therapy in general. In addition, it willbe important to confirm these findings in larger studies.

Acknowledgments

Further support was provided by the Dean and Alice IrvinFamily Foundation and Mrs. Shirley Leslie, who made contribu-tions to the research activities of the GOG Core Laboratory forReceptors. Administrative support was provided by Ms. AnneReardon, the Gynecologic Oncology Group, and Ms. LorettaCampbell, the University of New Mexico.

This study was supported by National Cancer Institute grantsto the Gynecologic Oncology Group Administrative Office (CA27469) and the Gynecologic Oncology Group Statistical andData Center (CA 37517). This work was also supported by NIHR01CA 99908-1 (KL), by NIH CA 27469 to the GynecologicCore Laboratory for Receptors, by the Cory/Beach Family Fund

332 M. Singh et al. / Gynecologic Oncology 106 (2007) 325–333

(KL), and by a University of New Mexico Cancer Research andTreatment Center Translational Research Grant (KL).

The following Gynecologic Oncology Group member institu-tions participated in this study: Duke University Medical Center,AbingtonMemorial Hospital,Walter ReedArmyMedical Center,University of Minnesota Medical School, University of Mis-sissippi Medical Center, The Milton S. Hershey School ofMedicine of the Pennsylvania State University, University ofCincinnati College ofMedicine, University of IowaHospitals andClinics, University of Texas Southwestern Medical Center atDallas,Wake Forest University School ofMedicine, University ofCalifornia, Irvine Medical Center, Tufts New England MedicalCenter, The Cleveland Clinic Foundation, State University ofNewYork at Stony Brook, Columbus Cancer Council, Fox ChaseCancer Center, Medical University of South Carolina, Universityof Oklahoma, University of Virginia, University of Chicago,Tacoma General Hospital.

References

[1] Clarke CL, Sutherland RL. Progestin regulation of cellular proliferation.Endocr Rev 1990;11:266–301.

[2] AndersonTJ,BattersbyS,KingRJ,McPhersonK,Going JJ.Oral contraceptiveuse influences resting breast proliferation. Hum Pathol 1989;20:1139–44.

[3] Longacre TA, Bartow SA. A correlative morphologic study of human breastand endometrium in themenstrual cycle. Am J Surg Pathol 1986;10:382–93.

[4] Kurman RJ, Kaminski PF, Norris HJ. The behavior of endometrialhyperplasia. A long-term study of “untreated” hyperplasia in 170 patients.Cancer 1985;56:403–12.

[5] Wood ER, Truesdale AT, McDonald OB, Yuan D, Hassell A, Dickerson SH,et al. A unique structure for epidermal growth factor receptor bound toGW572016 (Lapatinib): relationships among protein conformation, inhibitoroff-rate, and receptor activity in tumor cells. Cancer Res 2004;64:6652–9.

[6] Kelley RM, Baker WH. Progestational agents in the treatment ofcarcinoma of the endometrium. N Engl J Med 1961;264:216–22.

[7] PiverMS, Barlow JJ, Lurain JR, Blumenson LE.Medroxyprogesterone acetate(Depo-Provera) vs. hydroxyprogesterone caproate (Delalutin) in women withmetastatic endometrial adenocarcinoma. Cancer 1980;45:268–72.

[8] Reifenstein Jr EC. The treatment of advanced endometrial cancer withhydroxyprogesterone caproate. Gynecol Oncol 1974;2:377–414.

[9] Bonte J, Decoster JM, Ide P, Billiet G. Hormonoprophylaxis andhormonotherapy in the treatment of endometrial adenocarcinoma by meansof medroxyprogesterone acetate. Gynecol Oncol 1978;6:60–75.

[10] Thigpen JT, BradyMF,AlvarezRD,AdelsonMD,HomesleyHD,ManettaA,et al. Oral medroxyprogesterone acetate in the treatment of advanced orrecurrent endometrial carcinoma: a dose–response study by the GynecologicOncology Group. J Clin Oncol 1999;17:1736–44.

[11] Quinn MA, Cauchi M, Fortune D. Endometrial carcinoma: steroid receptorsand response to medroxyprogesterone acetate. Gynecol Oncol 1985;21:314–9.

[12] Lange CA, Shen T, Horwitz KB. Phosphorylation of human progesteronereceptors at serine-294 by mitogen-activated protein kinase signals their degra-dation by the 26S proteasome. Proc Natl Acad Sci U S A 2000;97:1032–7.

[13] Whitney CW, Brunetto VL, Zaino RJ, Lentz SS, Sorosky J, ArmstrongDK, et al. Phase II study of medroxyprogesterone acetate plus tamoxifen inadvanced endometrial carcinoma: a Gynecologic Oncology Group study.Gynecol Oncol 2004;92:4–9.

[14] Fiorica JV, Brunetto VL, Hanjani P, Lentz SS, Mannel R, Andersen W.Phase II trial of alternating courses of megestrol acetate and tamoxifen inadvanced endometrial carcinoma: a Gynecologic Oncology Group study.Gynecol Oncol 2004;92:10–4.

[15] Kumar NS, Richer J, Owen G, Litman E, Horwitz KB, Leslie KK.Selective down-regulation of progesterone receptor isoform B in poorlydifferentiated human endometrial cancer cells: implications for unopposedestrogen action. Cancer Res 1998;58:1860–5.

[16] Leslie KK, Kumar NS, Richer J, Owen G, Takimoto G, Horwitz KB, et al.Differential expression of the A and B isoforms of progesterone receptor inhuman endometrial cancer cells. Only progesterone receptor B is inducedby estrogen and associated with strong transcriptional activation. Ann NYAcad Sci 1997;828:17–26.

[17] Vegeto E, ShahbazMM,WenDX, GoldmanME, O'Malley BW,McDonnellDP. Human progesterone receptor A form is a cell- and promoter-specificrepressor of human progesterone receptor B function [see comments]. MolEndocrinol 1993;7:1244–55.

[18] Smid-Koopman E, Blok LJ, Kuhne LC, Burger CW, Helmerhorst TJ,Brinkmann AO, et al. Distinct functional differences of human progester-one receptors A and B on gene expression and growth regulation in twoendometrial carcinoma cell lines. J Soc Gynecol Investig 2003;10:49–57.

[19] LeslieKK, SteinMP,KumarNS,DaiD, Stephens J,Wandinger-NessA, et al.Progesterone receptor isoform identification and subcellular localization inendometrial cancer. Gynecol Oncol 2005;96:32–41.

[20] McCarty Jr KS, Miller LS, Cox EB, Konrath J, McCarty Sr KS. Estrogenreceptor analyses. Correlation of biochemical and immunohistochemicalmethods using monoclonal antireceptor antibodies. Arch Pathol Lab Med1985;109:716–21.

[21] Dai D, Litman ES, Schonteich E, Leslie KK. Progesterone regulation ofactivating protein-1 transcriptional activity: a possible mechanism of proges-terone inhibition of endometrial cancer cell growth. J Steroid Biochem MolBiol 2003;87:123–31.

[22] Dai D, Wolf DM, Litman ES, White MJ, Leslie KK. Progesterone inhibitshuman endometrial cancer cell growth and invasiveness: down-regulationof cellular adhesion molecules through progesterone B receptors. CancerRes 2002;62:881–6.

[23] Davies S, Dai D, Feldman I, Pickett G, Leslie KK. Identification of a novelmechanism of NF-kappaB inactivation by progesterone through proges-terone receptors in Hec50co poorly differentiated endometrial cancer cells:induction of A20 and ABIN-2. Gynecol Oncol 2004;94:463–70.

[24] Qiu M, Lange CA. MAP kinases couple multiple functions of humanprogesterone receptors: degradation, transcriptional synergy, and nuclearassociation. J Steroid Biochem Mol Biol 2003;85:147–57.

[25] Qiu M, Olsen A, Faivre E, Horwitz KB, Lange CA. Mitogen-activatedprotein kinase regulates nuclear association of human progesteronereceptors. Mol Endocrinol 2003;17:628–42.

[26] Mortel R, Zaino R, Satyaswaroop PG. Heterogeneity and progesterone–receptor distribution in endometrial adenocarcinoma. Cancer 1984;53:113–6.

[27] Soper JT, McCarty Jr KS, Creasman WT, McCarty Sr KS. Histologiccontrol of biochemical steroid receptor analysis in endometrial carcinomas.Am J Obstet Gynecol 1985;153:520–3.

[28] Ehrlich CE, Young PC, Cleary RE. Cytoplasmic progesterone and estradiolreceptors in normal, hyperplastic, and carcinomatous endometria:therapeutic implications. Am J Obstet Gynecol 1981;141:539–46.

[29] Ehrlich CE, Young PC, Stehman FB, Sutton GP, Alford WM. Steroidreceptors and clinical outcome in patients with adenocarcinoma of theendometrium. Am J Obstet Gynecol 1988;158:796–807.

[30] McCarty Jr KS, Barton TK, Fetter BF, Creasman WT, McCarty Sr KS.Correlation of estrogen and progesterone receptors with histologic differenti-ation in endometrial adenocarcinoma. Am J Pathol 1979;96:171–83.

[31] Benraad TJ, Friberg LG, Koenders AJ, Kullander S. Do estrogen and proges-terone receptors (E2R and PR) in metastasizing endometrial cancers predictthe response to gestagen therapy?ActaObstetGynecol Scand 1980;59:155–9.

[32] Martin PM, Rolland PH, Gammerre M, Serment H, Toga M. Estradiol andprogesterone receptors in normal and neoplastic endometrium: correlationsbetween receptors, histopathological examinations and clinical responsesunder progestin therapy. Int J Cancer 1979;23:321–9.

[33] Creasman WT, Soper JT, McCarty Jr KS, McCarty Sr KS, Hinshaw W,Clarke-Pearson DL. Influence of cytoplasmic steroid receptor content onprognosis of early stage endometrial carcinoma. Am J Obstet Gynecol1985;151:922–32.

[34] Creasman WT. Prognostic significance of hormone receptors in endome-trial cancer. Cancer 1993;71:1467–70.

[35] Ingram SS, Rosenman J, Heath R, Morgan TM, Moore D, Varia M. Thepredictive value of progesterone receptor levels in endometrial cancer. Int JRadiat Oncol Biol Phys 1989;17:21–7.

333M. Singh et al. / Gynecologic Oncology 106 (2007) 325–333

[36] Kauppila AJ, Isotalo HE, Kivinen ST, Vihko RK. Prediction of clinicaloutcome with estrogen and progestin receptor concentrations and theirrelationships to clinical and histopathological variables in endometrialcancer. Cancer Res 1986;46:5380–4.

[37] Friberg LG, Noren H. Prognostic value of steroid hormone receptors for 5-year survival in stage II endometrial cancer. Cancer 1993;71:3570–4.

[38] Soper JT, Segreti EM, Novotny DB, Mutch D, Creasman WT, McCarty JrKS. Estrogen and progesterone receptor content of endometrial carcino-mas: comparison of total tissue versus cancer component analysis. GynecolOncol 1990;36:363–8.

[39] Pertschuk LP, Masood S, Simone J, Feldman JG, Fruchter RG, Axiotis CA,et al. Estrogen receptor immunocytochemistry in endometrial carcinoma: aprognostic marker for survival. Gynecol Oncol 1996;63:28–33.

[40] Iwai K, Fukuda K, Hachisuga T, Mori M, Uchiyama M, Iwasaka T, et al.Prognostic significance of progesterone receptor immunohistochemistryfor lymph node metastases in endometrial carcinoma. Gynecol Oncol1999;72:351–9.

[41] Fukuda K, Mori M, Uchiyama M, Iwai K, Iwasaka T, Sugimori H.Prognostic significance of progesterone receptor immunohistochemistry inendometrial carcinoma. Gynecol Oncol 1998;69:220–5.

[42] Chambers JT, Carcangiu ML, Voynick IM, Schwartz PE. Immunohisto-chemical evaluation of estrogen and progesterone receptor content in 183patients with endometrial carcinoma. Part II: correlation betweenbiochemical and immunohistochemical methods and survival. Am J ClinPathol 1990;94:255–60.

[43] Chambers JT, MacLusky N, Eisenfield A, Kohorn EI, Lawrence R,Schwartz PE. Estrogen and progestin receptor levels as prognosticators forsurvival in endometrial cancer. Gynecol Oncol 1988;31:65–81.

[44] Sivridis E, Giatromanolaki A, Koukourakis M, Anastasiadis P. Endome-trial carcinoma: association of steroid hormone receptor expression withlow angiogenesis and bcl-2 expression. Virchows Arch 2001;438:470–7.

[45] Fanning J, Brown S, Phibbs G, Kramer T, Zaher A. Immunohistochemicalevaluation is not prognostic for recurrence in fully staged high-riskendometrial cancer. Int J Gynecol Cancer 2002;12:286–9.

[46] Mote PA, Johnston JF, Manninen T, Tuohimaa P, Clarke CL. Detection ofprogesterone receptor forms A and B by immunohistochemical analysis.J Clin Pathol 2001;54:624–30.

[47] Mote PA, Johnston JF, Manninen T, Tuohimaa P, Clarke CL. Detection ofprogesterone receptor forms A and B by immunohistochemical analysis.J Clin Pathol 2001;54:624–30.

[48] Kreitmann B, Bugat R, Bayard F. Estrogen and progestin regulation of theprogesterone receptor concentration in human endometrium. J ClinEndocrinol Metab 1979;49:926–9.

[49] Janne O, Kontula K, Luukkainen T, Vihko R. Oestrogen-induced proges-terone receptor in human uterus. J Steroid Biochem 1975;6:501–9.

[50] Satyaswaroop PG, Zaino RJ, Mortel R. Estrogen-like effects of tamoxifenon human endometrial carcinoma transplanted into nude mice. Cancer Res1984;44:4006–10.

[51] Janssens JP, Billiet G, Bonte J, De Loecker W. Independent regulation ofgrowth and steroid receptors in uterus and mammary tumors of rats.Anticancer Res 1983;3:385–91.

[52] Chen TC, Su S, Fry D, Liebes L. Combination therapy with irinotecanand protein kinase C inhibitors in malignant glioma. Cancer 2003;97:2363–73.

[53] Bagowski CP, Myers JW, Ferrell Jr JE. The classical progesterone receptorassociates with p42 MAPK and is involved in phosphatidylinositol 3-kinase signaling in Xenopus oocytes. J Biol Chem 2001;276:37708–14.

[54] Boonyaratanakornkit V, Scott MP, Ribon V, Sherman L, Anderson SM,Maller JL, et al. Progesterone receptor contains a proline-rich motif thatdirectly interacts with SH3 domains and activates c-Src family tyrosinekinases. Mol Cell 2001;8:269–80.

[55] Guiochon-Mantel A, Lescop P, Christin-Maitre S, Loosfelt H, Perrot-Applanat M, Milgrom E. Nucleocytoplasmic shuttling of the progesteronereceptor. EMBO J 1991;10:3851–9.

[56] Herzog TJ. What is the clinical value of adding tamoxifen to progestins inthe treatment [correction for treatment] of advanced or recurrentendometrial cancer? Gynecol Oncol 2004;92:1–3.

[57] Fujimoto J, Ichigo S, Hori M, Nishigaki M, Tamaya T. Expression ofprogesterone receptor form A and B mRNAs in gynecologic malignanttumors. Tumour Biol 1995;16:254–60.

[58] Dimopoulos MA, Papadimitriou CA, Georgoulias V, Moulopoulos LA,Aravantinos G, Gika D, et al. Paclitaxel and cisplatin in advanced orrecurrent carcinoma of the endometrium: long-term results of a phase IImulticenter study. Gynecol Oncol 2000;78:52–7.