METHODS

Patients

Four postmenopausal patient groups were selected. The first group (group I) consisted of patients with breast cancer using tamoxifen 20–40 mg daily as adjuvant treatment. Menopause was defined as no menstrual periods in the preceding 12 months and oestradiol concentrations of < 0.10 nmol/litre. These patients attended the gynaecology outpatient clinic of the University Hospital Groningen from January 1995 until December 1998, as part of a gynaecological screening programme (n = 91). Outpatient hysteroscopy and uterine curettings or biopsies were performed in patients with an endometrial thickness of > 5 mm on ultrasound, which occurred in 62 cases. Endometrium was obtained in 32 patients (22 asymptomatic patients, 10 patients with vaginal bleeding), whereas in 27 patients insufficient tissue was obtained for assessment. In three patients with thickened endometrium on ultrasound, a hysterectomy was performed for other reasons. A total of 35 benign endometrial samples was included in group I. Group II comprised non-malignant endometrium from the uteri of 24 postmenopausal women with stage I cervical cancer, who had never used tamoxifen and were currently not using hormonal treatment. All cases were retrieved from the database files of the department of pathology, University Hospital Groningen, from 1995 until 1998. Group III consisted of 15 patients with breast cancer who developed endometrial carcinoma during or after tamoxifen use. Cases were obtained from the database of the Northern Dutch Cancer Registry from 1989 until 1996. In group IV, cases were retrieved from the database of the department of gynaecology, University Hospital Groningen, from 1991 until 1998; the group comprised patients with endometrial carcinoma who had never used tamoxifen and were not using hormonal treatment at time of diagnosis (n = 51).

Methods

Sections (4 μm thick) of formalin fixed, paraffin wax embedded tissue were stained with standard haematoxylin and eosin. For immunohistochemistry, sections were mounted on 3`3`-aminopropylethoxysilane (Sigma-Aldrich, Diesenhofen, Germany) coated glass slides. After antigen retrieval by an autoclave method,²¹ the heat stable monoclonal antibody MIB-1 (Immunotech, Marseille, France), directed against an epitope of the Ki67 antigen, was used as a marker for proliferation.¹⁷ For the detection of ER and PR, primary monoclonal antibodies (Immunotech) were incubated overnight at 80°C in 0.01M Tris/HCl (buffer pH 9.0). All slides were then incubated with a second rabbit antimouse peroxidase labelled polyclonal antibody, using diaminobenzidine as chromogen, and counterstained with haematoxylin. Normal colonic mucosa was used as positive control for MIB-1. Positive controls for ER and PR were breast cancer specimens with known steroid receptor status. Negative controls underwent the same procedure omitting the primary antibodies.

Scoring

Histopathology was reviewed by an experienced gynaecopathologist (HH). Mitoses were counted in 10 fields of vision at ×40 magnification in epithelium and stroma, and the mitotic index was defined as the percentage of positive cells of the cells counted. The expression of MIB-1, ER, and PR was evaluated both in glandular epithelium and stroma. For counting MIB-1 positive epithelial cells, we used an eyepiece grid at ×40 magnification and counted the epithelial cells crossing 10 consecutive horizontal lines. In cases of normal endometrium, 10 consecutive fields were counted. In cases of tamoxifen induced endometrial pathology, counting was performed in the field with the highest epithelial density. In the carcinoma cases, we selected well fixed fields of tumour, whereas in cases of apparent tumour heterogeneity we counted 10 fields in the area with the lowest MIB-1 index and 10 fields in areas with the highest MIB-1 index. The overall score is given as the mean of these two indices. Fields with squamous metaplasia were excluded from counting. An average of 707 epithelial cells (207–1585) was counted for each sample. An average of 820 stromal cells (408–1364) was counted for each sample. Endometrial epithelial and stromal cells were considered to be positive for MIB-1, ER, or PR if nucleoplasm or nucleoli were unequivocally stained. The MIB-1 index was defined as the percentage of positive cell nuclei. A specimen was considered positive for ER or PR if staining of > 5% of nuclei of epithelial or stromal cells was seen. All samples were scored without knowledge of the clinical data, by one investigator (KtH). At random, 10% of the samples were read by two other investigators (MJM, HH). The paired differences between the different investigators in the MIB-1 index were small: mean, 0.7%; SD, 4.9%; 95% confidence interval, 1.3% to 2.8%.

Statistics

Data were analysed using the SPSS software for Windows (8.0) (SPSS Inc, Chicago, Illinois, USA). For statistical analysis of the Ki67 data the Mann-Whitney U test was used. For ER and PR analysis a χ² test was performed, corrected for continuity. Only p values < 0.05 were considered significant.

RESULTS

Patient characteristics

Table 1 summarises the patient characteristics. All patients were postmenopausal and the median ages in the four patient groups were similar. In group I, all patients had been using tamoxifen at the time of tissue sampling, with a median duration of 27 months (range, 5–84 months). In group III, the median interval between the start of tamoxifen and the detection of endometrial carcinoma was 18 months (range, 3–90 months).

Histopathology

Table 2 shows the histopathology results. The 35 endometrial samples in group I comprised 18 endometrial polyps, one sample with complex non-atypical hyperplasia, and 16 with cystic atrophy, including the three hysterectomy specimens. The histology of the polyps showed a striking resemblance to the cystic atrophy of the endometrium (fig 1).¹⁵ Both were composed of cystically dilated glands, covered with monolayered or pseudostratified metaplastic epithelium. The nuclei of the epithelial cells were small, round to oval, with small nucleoli and few (if any) mitotic figures. In the stroma, no mitotic activity was found. The stroma was cellular, especially around the glands, with focal oedema and formation of collagen.^13,15 The best descriptive diagnosis of the polyps was “adenofibroma”. Mitotic figures were seldom found (< 0.5%) in both the stroma and the epithelium of tamoxifen users and non-users, although no statistical analysis was performed on these data. Group II showed atrophic endometrium in 17 patients, endometrial polyps in three, simple hyperplasia in one, and proliferative endometrium without atypia in three patients. In groups III and IV, two of the 15 tumours and two of the 51 tumours were papillary serous carcinomas, respectively; the other tumours were all endometrioid adenocarcinomas, with a tendency for higher grade tumours in tamoxifen users (group III) (table 2).

MIB-1, ER, and PR expression in epithelial cells

Table 3 shows the MIB-1 index, and the expression of ER and PR in epithelium. The mean MIB-1 index in benign endometrial epithelium was higher in tamoxifen users (mean, 13% (SD, 13%) v mean 2% (SD, 2%); p < 0.05). In endometrial cancer, the mean MIB-1 index was higher, but did not differ between tamoxifen treated and untreated patients. However, in endometrial stroma, the MIB-1 index in all groups was very low (< 0.5%). Figure 2 shows MIB-1 staining in cystic atrophy in a tamoxifen user.

The expression of ER in benign epithelium was high and did not differ between tamoxifen users and non-users, whereas in tamoxifen users the expression of ER was significantly different in benign and malignant endometrium (p = 0.005). The expression of ER in endometrial cancer was lower in tamoxifen treated than in untreated patients (60% v 88%; p = 0.05; fig 3). Scoring of endometrial stroma revealed a high expression of ER in benign stroma, and a lower expression of ER in the malignant counterpart (both p < 0.05), although there was no significant difference between those patients exposed to tamoxifen and those who were not (table 4).

The expression of PR in the epithelium was similarly high in all groups. However, in stroma, the expression of PR was upregulated in tamoxifen users compared with non-users in benign (84% v 54%; p < 0.05) but not in malignant endometrium (table 4; fig 4). The expression of PR in stroma was higher in benign endometrium, both in tamoxifen users and non-users (both p < 0.05).

DISCUSSION

Our study shows a higher proliferative state, as determined by MIB-1 index, of benign endometrial epithelium in postmenopausal tamoxifen users when compared with untreated women. However, this was not reflected by the histomorphology of the epithelium or by mitotic figures in standard haematoxylin and eosin staining. On the contrary, the epithelium in tamoxifen users remained thin whereas the stroma became cell rich. However, the MIB-1 index in the stroma was also very low. The increased MIB-1 index in the endometrial epithelium might be indicative of an increased proliferative inclination in this subgroup of postmenopausal tamoxifen users. The fact that the control groups (II and IV) comprise endometrial tissue from women without breast cancer could be a source of bias, because breast and endometrial cancer share several risk factors. Another potential source of bias is the fact that there were more symptomatic patients in the tamoxifen users with benign endometrium (group I) than in non-users with cervical cancer (group II), because abnormal bleeding is usually associated with increased proliferation. The endometrial cancer risk in tamoxifen users is increased two to threefold and therefore it is very likely that this higher proliferative state has played a role in its development.^4,7,22 Recently, two studies on the modulation of Ki67 expression by tamoxifen in endometria of premenopausal and postmenopausal women have been published.^23,24 These studies show results similar to ours. They are both small studies and in one study controls were hormone replacement users²⁴ and there were only six postmenopausal women in the other.²³ Our study is the first in which the MIB-1 index was studied in a larger group of postmenopausal women, and the first in which the effects of tamoxifen are described both in endometrial epithelium and stroma. Because tamoxifen has an agonistic effect on postmenopausal endometrium and ER and PR expression may be upregulated, we have related the MIB-1 index to ER and PR expression. In the benign endometrium most of the samples were ER positive, irrespective of tamoxifen use, consistent with the findings of others.^22,25 However, in benign endometrium, we observed an increased expression of PR in stromal cells of tamoxifen users compared with non-users. In malignant endometrium this difference was not significant. This higher stromal PR expression in the endometrium is consistent with an agonistic effect of tamoxifen on the stroma, mostly independent of the histological diagnosis. Whether or not the upregulation of PR in stroma by tamoxifen is related to the relatively higher incidence of stromal tumours^22,26 and malignant mixed Müllerian tumours²⁴ in tamoxifen induced endometrial cancer still is not clear. However, our findings differ from those of Schwartz et al in tamoxifen users, who found increased expression of PR only in glandular epithelium and decreased expression of ER in the stroma.²⁷ An explanation for this discrepancy could be the small size of their study.

Progestagens have an antimitotic effect on endometrial epithelial cells. Despite the upregulation of the expression of PR in tamoxifen exposed endometrium, clinical studies have not yet shown a beneficial effect of systemic progestins on the prevention of tamoxifen induced endometrial proliferation,^11,28,29 although withdrawal bleeding has been described.²⁹ In a recent randomised trial, a levonorgestrel releasing intrauterine device resulted in an 85% stromal decidualisation of tamoxifen primed endometrium, although endometrial ultrasonographic appearance and thickness remained unchanged and an excess of bleeding appeared in the treatment group.³⁰ Only data from a large prospective trial can answer the question of the efficacy of progestin treatment in the prevention of endometrial cancers and polyps in longterm tamoxifen users.

“In benign endometrium, we observed an increased expression of the progesterone receptor in stromal cells of tamoxifen users compared with non-users”

In the group of tamoxifen users with endometrial cancer, we found that the percentage of ER positive tumours was lower (60%) than in non-users with sporadic cancer (88%), as has been described previously.³¹ The presence of ER and PR was linked to well differentiated endometrial carcinomas with a good prognosis, whereas the absence of receptors was considered to predict invasive growth and a poor prognosis.³² Some studies describe more adverse histology for tamoxifen induced endometrial carcinomas,^7,22,33 and a lower expression of ER compared with non-users,²² although there is some controversy over this subject.^3,34,35 In our study, the histology of the endometrial carcinomas of tamoxifen users (group III) and non-users (group IV) showed only two papillary serous tumours in both groups and, as a consequence, in group III the adverse histology constitutes a greater part of the carcinomas than in group IV. Moreover, there is a tendency for higher grade tumours in the tamoxifen group (table 2).

In conclusion, using the MIB-1 index as a measure of proliferation, we found a tamoxifen associated higher proliferative state in the glandular epithelium, which might play a role in the observed higher incidence of endometrial cancer in tamoxifen users. In stromal cells, we found a tamoxifen associated higher expression of PR, suggesting an oestrogenic effect on the stroma that is independent of histological diagnosis. Therefore, our data support the theory of an oestrogenic effect of tamoxifen on postmenopausal endometrial epithelium in addition to stroma.