Significance of epidermal growth factor receptor expression in primary human endometrial cancer

Int. J. Cancer: 56,26-30 (1994) 0 1994 Wiley-Liss, Inc.

Publication of the InternationalUnion Against Cancer Publication de I'Union Internationale Contre le Cancer

SIGNIFICANCE OF EPIDERMAL, GROWTH FACTOR RECEPTOR EXPRESSION IN PRIMARY HUMAN ENDOMETRIAL, CANCER Giovanni SCAMBIA', Pierluigi BENEDETTIPANICI',Gabriella FERRANDINA', Francesco BAITAGLIA', Mariagrazia DISTEFANO', Giorgia D'ANDREAI,Rosa DE VINCENZO~, Francesco MANESCHI',Franco Oreste RANELLETTI~ and Salvatore M A N C ~ S O ' ~ ~ 'Department of Gynecology/ Obstetrics and 2Departmentof Histology, Catholic University,L.go A. Gemelli, 8, 00168, Rome, Italy. Radioreceptwialassessmentof EGFR expressionwas prospectively performed on 60 primary human endometrial tumors. Of these, 26 were EGFR-positive while 13 expressed high EGFR levels. High EGFR levels correlated well with poor histopathological grading. No correlation with histopathological type, stage, myometrial invasion, lymph-node involvement or steroid hormone receptor status was observed. Disease-free survival rate was significantly shorter in the cases with high than in the cases with low EGFR levels. These results suggest a potential role of EGFR expression assessment in prognostic characterization of endometrial cancer patients.

to June 1992. None of the patients received pre-operative hormonal or radiation treatment. Patients were treated by abdominal hysterectomy plus bilateral salpingo-oophorectomy. In 22 cases systematic pelvic and para-aortic lymphadenectomy was performed. Of the 66 eligible patients, 6 were lost to follow-up. The median age of the 60 assessable patients was 62 years (range = 30-84). Seven (12%) patients were pre-menopausal and 53 (88%) were post-menopausal. All patients were staged according to F.I.G.O. classification and their tumors graded as well (Gl), moderately (G2) or poorly (G3) differentiated. Characteristics of the patients are listed in Table I.

Endometrial cancer is generally considered to have a more favorable prognosis than other gynecological malignancies. However, approximately 30-35% of women will have recurrences and die of disease (Petterson, 1988). At present surgical stage, depth of myometrial invasion and grade are commonly used to predict clinical outcome, but none of these parameters can be considered as accurate. Moreover, histopathological grading is subject to inter-observer variability and is difficult to reproduce. Studies have been carried out on the biological factors that could be used, apart from common histopathological findings, in order to identify patients at risk of treatment failure, in whom more individualized therapies should be attempted. In particular, steroid hormone receptor content and cytofluorimetric data on tumor ploidy and DNA replicating activity have been associated with prognosis in endometrial cancer (Utaaker et al., 1987; Britton et al., 1989). However, their precise role in improving the present prognostic characterization needs to be more clearly investigated. Recently, much attention has been focused on the role of peptidic growth factors, in particular epidermal growth factor (EGF) and its receptors (EGFR), in regulating the growth of endometrial cells. EGFRs have been found in normal endometrium in which they undergo estrogen and progesterone regulation (Mukku and Starcel, 1985; Taketani and Mizuno, 1991). In only a few studies, to date, has EGFR been detected and characterized by various techniques in primary endometrial tumors (Berchuck et al., 1989; Battaglia et al., 1989; Reynolds et al., 1990; Birmelin et al., 1992; Bigsby et al., 1992). It has been proposed that abnormal EGFR expression may be one of the mechanisms by which unopposed estrogens are involved in malignant endometrial transformation (Reynolds et al., 1990). Evidence has been reported concerning the unfavorable prognostic role of high EGFR expression in several human neoplasms such as breast (Sainsbury et al., 1987), ovarian (Scambia et al., 1992), esophageal (Ozawa et al., 1989) and laryngeal (Maurizi et al., 1992) cancer. In the current study, the expression of EGFR in a group of primary endometrial carcinomas was analyzed. Correlation with clinico-pathological parameters, receptor status and clinical outcome was also investigated.

Processing of tumor tissue Tissue specimens, collected during primary surgery, were frozen on dry ice shortly after surgical removal and stored at -80°C until processing. A representative section of the specimens was retained for histopathological examination. The membrane fraction and cytosol were prepared as described elsewhere (Scambia et al., 1992). Briefly, tumor specimens were minced and homogenized in 5 vol ice-cold buffer consisting of 25 mM TRIS, 1.5 mM EDTA, 5 mM NaN3, 20% glycerol (TENG) and 0.1% monothioglycerol, by applying intermittent bursts of an Ultra-Turrax homogenizer. The crude homogenate was centrifuged at 7,000g for 20 min at O"C, then the supernatants were centrifuged at 105,OOOg for 75 min at 0°C to obtain a membrane pellet and a cytosolic fraction.

o 1994 Wiley-Liss,Inc.

MATERIAL AND METHODS

Our study was prospectively conducted on a group of 66 primary endometrial cancer patients enrolled from June 1986

l2'I-EGF binding measurement The membrane pellet was resuspended in TENG plus 10 mM MgC12. Aliquots of the suspension (100 pl, containing 300 to 500 pg protein) were incubated with I=I-EGF (3.2 nM) in the presence or absence of unlabelled EGF (1 mM) for 16 hr at room temperature in a final volume of 400 pI. Binding was stopped by addition of TENG plus 0.1% BSA. Pellets were obtained by centrifugation at 2,000 g for 20 min at 0°C and counted in a gamma-counter for 1min. Results were expressed as fmoles per mg of membrane protein (fmol/mg prot). Assessment of tissue weight permitting Scatchard analysis was carried out with concentrations of I=I-EGF from 0.025 to 4.25 nmol/l. Protein concentration was measured by the Bradford (1976) method. EGFR values of 3.00 fmol/mg prot and 5.00 fmol/mg prot were arbitrarily chosen as cut-off values to define EGFR status. ER and PR measurement ER and PR were assayed by the dextran-coated charcoal (DCC) method according to the EORTC protocol (EORTC, 1980; Battaglia et al., 1988) using increasing concentrations of 3H-estradiol (spec. act. 116 Ci/mmol from 0.05 nM to 1 nM) or 3H-ORGANON 2058 (spec. act. 47 Ci/mmol from 0.5 nM to 8 nM) (Amersham, Aylesbury, UK) as radiolabelled ligands. ER and PR values of 10 fmol/mg protein and 20 fmol/mg protein were chosen to define ER and PR status. 3T0 whom correspondence and reprint requests should be addressed. Fax:06-3051343. Received: April 16,1993 and in revised form July 30,1993.

27

EGFR EXPRESSION IN PRIMARY HUMAN ENDOMETRIAL TUMORS

TABLE I - DISTRIBUTION OF EGFR LEVELS ACCORDING TO CLINICO-PATHOLOGICALCHARACTERISTICS EGFR

>3.00 >5.00 (fmol/mg rot) Number &mber

Number Median Range (fmol/mg prot)

All cases Histo pe Engmetr ioid Adenosquamow Stage I I1 I11 Grade 1 2 3 Myometrial invasion Ne ative 50% Lymph-node involvement No Yes

(%)

(%)

60

1.98

0-18.40 26(43) 13(22)

59 1

1.93

0-18.40 25 (42) 12(20) 1 1

-

I

l 0

I

0

41 4 15 13 34 13 3 32 25

1751

1.93 0-18.40 16 39 3.45 1.53-7.40 1.76 0-15.10 37 47 1.16 2.41 1.83

-.0

8 19 411251 27

rl 6 CI

0

3

3 20 1 [14]

00

0

0

lI0

0

0

5 33 4 [57)

B

0

Y

1.75 0-18.40 3.00 0.2-6.60

3

2 r = 0.063; p N.S.

0

o

'

O',

0 04

so

h

15 7

2

1

Log (x) of ER

0-10.46 0-15.10 16 47 8 23 0-18.40 46 el1 46 41 31

1.53 0.28-1.75 3.07 0-15.10 17 1.76 0-18.40 9

- . - - - - - I.

-

0 0

O0

@0

0

Specific 1251-EGF bound (pM) 10 8 8

:I :I

/

4 2

/

\* r.O.96

b '

\

O P 4 e 8 1 0

0

BOUND (pM)

0

1

2

3 12SI-EGF concentrations (nM)

4

1

FIGURE1 - Scatchard analysis of IzI-EGF bindin to endometrial tumor membranes. Specifically bound lZSI-E&Fwas measured as detailed in the text. 301

i

EGFR (fmoVmg of protein) FIGURE2 - Scattergram of EGFR values in 60 primary human endometrial cancers.

-1 -1

0 0

- - . - . - - . . .0 . - - -- . 0

1

2

' -

3

L q (x) ofPR FIGURE 3-Relationship between ER and PR (a), ER and EGFR (b), PR and EGFR (c) levels in primary human endometrial tumors. In order to investigate the correlations between ER, PR, and EGFR, the variance of error was normalized by transforming receptor data in log,, of data before performing the regression analysis.

Irnrnunohistochernical analysis Immunohistochemical analysis was performed by the avidinbiotin-peroxidase complex (ABC) method (Vector, Burlingame, CA) with the anti-EGFR mouse monoclonal antibody (Oncogene Science, Uniondale, NY).Tumor tissue was fixed in 10% buffered formalin (pH = 7.4) and embedded in paraffin wax. Cryostat sections of 5 km were de-waxed, rehydrated, and washed in PBS, then incubated for 5 min in 3.0% (w/v) Hz02 solution at room temperature to block any endogenous peroxidase. After washing in PBS, sections were incubated with normal serum to minimize non-specific binding. A 1:lOO dilution in PBS of the specific MAb or normal rat IgG (Sigma, St. Louis, MO) was used. The sections were then incubated with biotinylated horse anti-rat IgG (1:200 in PBS), and with the ABC complex reagent for 30 min at room temperature.

28

SCAMBIA ET AL.

A

D

C

FIGURE4 - Immunohistochemical analysis of EGFR expression in primary human endometrial tumors. (a) Control sections of endometrial adenocarcinoma with normal rat serum as primary antibody; (b to d ) endometrial adenocarcinoma: most of the tumor cells show staining reaction. Original magnifications are a and b x 100,c and d x 400. Scale bar: 1cm = 100 c~ (u,b) and 1cm = 25 p (c,d).

Finally, the sections were washed in PBS, stained by incubation with 0.5 mg/ml 3-3'-diaminobenzidine (Sigma) in 0.01% H202 for 5 min, and counterstained with hematoxylin. Statistical analysis The Mann-Whitney rank sum non-parametric test was used to analyze the relationship between EGFR expression and clinico-pathological characteristics. All medians and life tables were computed using the product-limit estimate of Kaplan and Meier (1958) and the curves were examined by the log-rank test (Mantel, 1966). Disease-free survival (DFS) was calculated from the day of surgery until the day of clinical recurrence or the date last seen. The median follow-up was 30.5 months. Analysis was as of September, 1992. RESULTS

EGFR binding Figure 1 shows a representative example of saturation analysis of lZI-EGF binding in endometrial cancer. Scatchard analysis indicated the presence of one specific high-affinity

binding site (I(,= 0.75 nmol/l) with an x-intercept of 5.93 fmol/mg protein. Among the endometrial cancer specimens, & values ranged from 0.5 to 1.8 nmol/l (median = 1.2 nmol /I). The specificity of EGF binding was evaluated by testing the ability of various peptides, such as FSH, GH, TSH-releasing hormone and insulin, used at the same concentration of unlabeled EGF, to prevent binding of lZI-EGF. All the compounds tested failed to effectively displace lZI-EGF, whereas unlabeled EGF displaced 70% of the lZI-EGF (data not shown). Figure 2 shows the distribution of EGFR levels in the overall population of 60 primary endometrial tumors. EGFR ranged from 0 to 18.40 fmol/mg protein (median: 1.98 fmol/mg protein, mean: 3.55 fmol/mg protein, SD: 4.04 fmol/mg protein). Twenty-six (43%) of the specimens were EGFRpositive (> 3.00 fmol/mg prot) while 13 (22%) of the cases expressed high EGFR levels ( > 5.00 fmol/mg prot). Comparison of EGFR with clinicopathological characteristics Table I shows the distribution of EGFR levels according to clinico-pathological characteristics of the patients. Well-

29

SCAMBIA ET AL.

18-month DFS was 90% in EGFR-positive cases compared to 70% in EGFR-negative cases.

r

DISCUSSION

L

I--* QI W

n

c (

EGFR EGFR

-

8

ENTERED ECURRENCE 47 Log-ranktest 4 + 13 p z0.04

-

16

2h

i2

MONTHS

RCURE 5 - Disease-free survival rate according to EGFR status defined at cut-off of 5.00 fmol/mg prot in 60 primary endometrial cancer patients. differentiated tumors are likely to express lower EGFR levels than moderately/poorly differentiated tumors (median: 1.16 fmol/mg protein; range: 0-10.46 vs. median: 2.21 fmol/mg protein; range: 0-18.40) (p = 0.057). No difference was found in EGFR levels in relation to myometrial invasion and stage. Seven (31.8%) out of the 22 cases in which systematic pelvic and para-aortic lymphadenectomy was performed showed metastatic lymph-node involvement. No difference in EGFR levels in primary endometrial tumors in relation to nodenegative and node-positive patients was found. Compatison of EGFR levels with ER and PR Figure 3 shows the linear regression between ER, PR, and EGFR levels. Overall, 52.5% of the cases were ER-positive and 53% were PR-positive. A strong correlation between E R and PR log levels was found (p = 0.OOOl). No correlation between EGFR and ER (p = 0.67) and EGFR and PR (p = 0.42) log values was observed. No difference was found between distribution of EGFRpositive cases at any of the cut-off points tested, according to ER status and PR status, respectively (data not shown). Immunohistochemical analysis Figure 4 shows a representative example of primary endometrial cancer immunohistochemically stained by MAb against EGFR. Immunoreactivity was evident in most tumor cells while the stroma was unreactive. Staining was localized at the cell membrane but in most cases staining of both the membrane and the cytoplasm of tumor cells was found. Survival analjsb During the follow-up period, recurrence of disease was observed in 8 cases (13%). Disease-free survival analysis of 60 primary endometrial cancer patients in relation to EGFR status defined at cut-off of 5.00 fmol/mg prot is shown (Fig. 5). A statistically significant relationship was shown between EGFR positivity and a shorter DFS (p = 0.04). In fact, the

The radioreceptorial assessment of EGFR expression was prospectively conducted in a series of primary human endometrial tumors. In our series, the proportion of EGFR-positive cases is rather similar to that observed by Birmelin et al. (1992). On the other hand, Berchuck et al. (1989), using an immunohistochemical technique, reported a higher percentage of endometrial cancer expressing EGFR. Our immunohistochemical analysis demonstrated that EGFR immunostaining was confined to endometrial tumor cells, thus minimizing the possibility that false-positive results could derive from the presence of stromal elements. We did not find any correlation between EGFR and tumor stage or the degree of myometrial invasion according to Birmelin et al. (1992). In contrast, we found a direct correlation between the presence of higher EGFR levels and poor histopathological grade of differentiation, suggesting that EGFR expression may characterize biologically more aggressive tumors. Birmelin et al. (1992) reported that, in 22 cases in which nodal status was assessed, lymph-node-positive tumors seemed to be more frequently EGFR-positive. In our series we failed to observe any relationship between lymph-node involvement and EGFR status, but the number of cases in both studies is too small for any significant conclusion to be drawn. There is evidence that estradiol and progesterone may respectively up- and down-regulate EGFR levels in normal endometrium, and it has been suggested that the EGF/EGFR system may have a physiological role in endometrial trophism (Mukku and Stancel, 1985; Taketani and Mizuno, 1991). Modulation of EGFR by unopposed estrogens has been suggested to be an important step in the onset of endometrial hyperplasia and cancer (Reynolds et al., 1990). According to Berchuck et al. (1989) we observed no correlation between EGFR and ER and PR levels, suggesting that the normal steroidal control mechanisms of EGFR expression are lost when malignant transformation has occurred. There are 2 other reports dealing with EGFR and steroid receptors in endometrial cancer. One (Llorens et al., 1989) showing an inverse relationship between EGFR and ER and no correlation with PR, is based on only 22 cases. On the contrary, Birmelin et al. (1992) found no correlation with ER and only a weak inverse relationship between EGFR and PR. The current prognostic characterization of endometrial cancer patients is still inadequate. Therefore, the identification of possible biologic factors correlating with tumor aggressiveness would be of the utmost importance in the selection of individualized primary and adjuvant therapy for patients at higher risk of recurrence. Consistent data have been reported about the unfavorable prognostic role of EGFR expression in breast (Sainsbury et al., 1987), ovarian (Scambia et al., 1992), esophageal (Ozawa et al., 1989), and laryngeal (Maurizi et al., 1992) cancer. In this study, EGFR-positive tumors behave more aggressively than EGFRnegative ones. The difference in relapse rate and the prognostic information obtained from DFS analysis are unlikely to be related to any difference in tumor stage or to myometrial invasion since these parameters are not associated with EGFR positivity. Birmelin et al. (1992) found a significant difference in the overall survival in relation to EGFR status. This difference was more evident when a higher cut-off for EGFR positivity was used, as reported in our own study. Our data, therefore, suggest that assessment of EGFR status could be useful in identification of patients with a more unfavorable prognosis. A multivariate analysis of a large number of cases with a longer follow-up period should be

30

SCAMBIA ET AL.

carried out. h a k e et al. (1991) reported the presence of EGF in about 50%, and of transforming growth factor-alpha in nearly all, of t h e endometrial cancers examined. Moreover, preliminary data suggested a relationship between high TGFalpha levels, tumor invasiveness and survival ( h a k e et al., 1991). Since over-production of the ligand by endometrial cancer cells in a n autocrine manner may lead t o an underestimation of EGFR levels, a complete evaluation of t h e prognostic role of EGFR in endometrial cancer should include the assessment of EGF/TGF-alpha levels in tumor extracts. Finally, it is worth noting that MAb anti-EGFR (Masui et al., 1988) and synthetic EGF-like peptides (Eppstein et al.,

1989) inhibit the growth of cancer cells. Moreover, Hancock et al. (1991) reported that the MAb against a n oncogene product similar to EGFR, i e . erbB2/neu (Downward et al., 1984), is able to synergize the anti-tumor activity of cis-diamminedichloroplatinum in vitro and in vivo. Thus, it is conceivable that drugs and/or biologic response modifiers (Scambia et al., 1991) able to interfere with the EGF/EGFR system may b e useful in endometrial cancer therapy. ACKNOWLEDGEMENTS

G.F. and R.D.V. are the recipients of a fellowship from the Italian Association for Cancer Research (A.I.R.C.).

REFERENCES

BAITAGLIA, F., SCAMBIA, G., BENEDEITIPANIC],P., BAIOCCHI, G., LLORENS, M.A., BERMEJO,M.J., SALCEDO, M.C., CHARRO, A.L. and PERRONE, L., IACOBELLI, S. and MANCUSO, S., Epidermal growth PUENTEM., Epidermal growth factor receptors in human breast and factor receptor ex ression in gynecological malignancies. Gynecol. endometrial carcinomas. J. Steroid Biochem., 34,505-509 (1989). obstet. Invest., 27,4!44 (1989). MANTEL,N., Evaluation of survival data and two new rank order BAITAGLIA, F., SCAMBIA, G., Ross], S., BENEDEITI PANICI,P., IACO- statistics arising in its consideration. Cancer Chemother. Rep., 50, BELLI,S., CRUCIITI,F. and MANCUSO, S., Epidermal growth factor 163-170 (1966). receptor in human breast cancer: correlation with steroid hormone MASUI,H., KAWAMOTO, T., SATO,J.D., WOLG,B., SATO,G. and receptors and axillary lym h node involvement. Europ. J. Cancer MENDELSOHN. J.. Growth inhibition of human tumor cells in athvmic Oncol., 24,1685-1690 (1988r mice by antiepidermal growth factor receptor monoclonal antibobies. BERCHUCK, A., SOISSON, A.P., OLT,G.J., SOPER,J.T., CLARKE,D.L., CancerRes., 44,1002-1007 (1988). BAST,R.C. and MCCARTHY, K.S., Epidermal growth factor receptor M., SCAMBIA, G., BENEDEITI PANCI,P., FERRANDINA, G., expression in normal and malignant endometrium. Amer. J. Obstet. MAURIZI, ALMADORI, G., PALUDEITI,G., DE VINCENZO, R., DISTEFANO, M., Gynecol., 161,1247-1252 (1989). BRINCHI, D., CADONI, G. and MANCUSO, S., EGF receptor expression BIGSBY, R.M., AIXIN,L., BOMALASKY, J., STEHMAN, F.B., LOOK,K.Y. in primary laryngeal cancer: correlation with clinico-patholo ical and SUITON,G.P., Immunohistochemicalstudy of HER-a/neu, epider- features and prognostic significance. Int. J. Cancer, 52,862-866 (1892). mal growth factor receptor, and steroid rece tor expression in normal MUKKU, V.R. and STANCEL, G.M., Regulation of e idermal and malignant endometrium. Obstet. Gyneco!, 79,95-100 (1992). factor receptor by estr0gen.J. biol. Chem., 260,9820-9824 ( 1 9 8 r w h BIRMELIN, G., ZIMMER, V., SAUERBREI, W., PFLEIDERER, A. and N. and ABE,O., Prognostic BAUKNECHT, T., Relationship between epidermal owth factor recep- OZAWA,S., UEDA,M., ANDO,N., SHIMIZU, tor (EGF-R) and various prognostic factors in E m a n endometrial significance of Epidermal Growth factor Rece tor in esophageal squamous cell carcinomas. Cancer, 63,2169-2173 fi989). carcinoma. Int. J. gvnecol. Cancer, 2,66-74 (1992). BRADFORD, M.M., A rapid and sensitive method for the quantitation PETERSON, F., Annual Report on the results of treatment in gynecologic of micropam quantities of protein using the principle of protein cancer. FIGO, Stockholm (1988). dye-binding.Anal. Biochem., 72,248-254 (1976). REYNOLDS, R.K., TALAVERA,F., ROBERTS,J.A., HOPKINS, M.P. and BRIITON,L.C., WILSON, T.O., GAFFEY,T.A., LIEBER,M.M., WIEAND, JARAIMMENON,K.M., Characterization of epidermal growth factor H.S. and PODRATZ, K.C., Flow cytometric DNA analysis of stage I rece tor in normal and neoplastic human endometrium. Cancer, 66, 196311974 (1990). endometrial carcinoma. Gynecol. Oncol., 34,317-322 (1989). DOWNWARD, J., YARDEN,Y., MAYES,E., SCRACE,G., T o m , N., SAINSBURY, J.R.C., FARNDON, J.R., NEEDHAM, G.K., MALCOLM,A.J. STOCKWELL, P., ULLRICH, A., SCHLESSINGER, J. and WATERFIELD,and HARRIS,A.L., Epidermal growth factor receptor status as predicM.D., Close similarityof epidermal growth factor receptor and v-erb-B tor of early recurrence of and death from breast cancer. Lancet, I, oncogene protein sequences. Nature (Lond.), 307,521-527 (1984). 1398-1402 (1987). EORTC, Breast Cancer Cooperative Group Revision of the standards SCAMBIA, G., BENEDEITI PANICI,P., BAITAGLIA, F., FERRANDINA, G., for the assessment of hormone receptors in human breast cancer. BAIOCCHI,G., GALLO,A., SICA,G. and MANCUSO, S., Effect of Europ. J. Cancer, 16,1513-1515 (1980). recombinant human Interferon-alpha 2b on receptors for steroid EPPSTEIN, D.A., MARSH,Y.V., SCHRYVER, B.B. and BERTICS,P.J., hormones and epidermal growth factor in patients with endometrial Inhibition of e idermal growth factor stimulated cell growth by a cancer. Europ. J. Cancer, 27,Sl-53 (1991). synthetic pept&.J. cell. Physiol., 141,420-430 (1989). SCAMBIA, G., BENEDEITI PANIC],P., BAITAGLIA, F., FERRANDINA, G., G., GREGGI,S., DE VINCENZO, R. and MANCUSO, S., HANCOCK, M.C., LANGTON, B.C., CHAN,T., TOY,P., MONAHAN, J.J., BAIOCCHI, MISCHAK, R.P. and SHAWVER, L.K., A monoclonal antibody against Significanceof epidermal rowth factor receptor in advanced ovarian the c-erbB-2 protein enhances the cytotoxicity of cis-diamminedichlo- cancer. J. clin. Oncol., 10,829-535 (1992). ro latinum against breast and ovarian tumor cell lines. CancerRes., 51, TAKETANI, Y. and MIZUNO,M., Evidence for direct regulation of 45754580 (1991). epidermal rowth factor rece tors by steroid hormones in human KAPLAN, E. and MEIER,P., Nonparametric estimation from incom- endometria! cells. Hum.Reprod), 6,1365-1369 (1991). plete observation.J. Amer. statist. Ass., 53,457481 (1958). UTAAKER, E., IVERSEN, O.E. and SKAARLAND, E., The distribution and LEAKE,R., CARR,L. and RINALDI, F., Autocrine and aracrine effects prognostic implications of steroid receptors in endometrial carcinomas. Gynecol. Oncol., 28,89-100 (1987). in the endometrium.Ann. N.Y. Acad. Sci., 622,145-188 (1991).