Tamoxifen and toremifene treatment of breast cancer and risk of subsequent endometrial cancer: A population-based case-control study

Abstract

A population-based case-control study was performed to evaluate the risk of endometrial cancer related to tamoxifen or toremifene treatment. All patients with breast cancer diagnosis since 1980 in Finland who subsequently developed an endometrial cancer by the end of 1995 and 3 matched controls were identified among the 38,000 breast cancer patients of the Finnish Cancer Registry database. Detailed information on treatment of breast cancer and potential confounders was collected from hospital records. The OR for tamoxifen treatment (59 cases), adjusted for significant cofactors (increased risk associated with obesity, low parity and PR positivity) was 2.9 (95% CI 1.8–4.7). The OR for toremifene (3 cases) was 0.9 (95% CI 0.3–3.9). The OR related to adjuvant tamoxifen treatment reached its maximum 2–5 years after the beginning of treatment (OR 5.1, 95% CI 2.1–13), while the OR for tamoxifen used for palliative treatment of advanced breast cancer was especially high after a lag of over 5 years (OR 9.5, 95% CI 2.5–36). The risk increase due to tamoxifen was slightly higher if the age at initiation was below 55, and risk was more pronounced among patients with well-differentiated endometrial cancer than patients with cancers of clinical grades 2 or 3. According to our results, treatment with tamoxifen increases the risk of endometrial cancer. Due to the rare use of toremifene up to the mid-1990s, the risk assessment concerning it was inconclusive. © 2002 Wiley-Liss, Inc.

Already in the mid-1980s, studies from population-based cancer registries documented that the history of breast cancer denotes an increased RR of endometrial cancer, the RRs being 1.4 in Connecticut,1 1.3 in Finland2 and 1.7 in Sweden.3 More recently, numerous reports about the carcinogenic potential of tamoxifen, which is widely used in adjuvant treatment for breast cancer, have produced some concern.4, 5, 6, 7, 8 It has been estimated that tamoxifen causes a 2–3-fold increase in the incidence of endometrial cancer.9

Tamoxifen is a nonsteroidal antiestrogen that blocks the growth-promoting effects of estrogen in breast tissue, mainly through competitive inhibition of the ER mechanism.10 Although tamoxifen is an inhibitor of breast cancer growth, its effects throughout the human body vary and could be characterized as mixed estrogenic and antiestrogenic properties. Most of the estrogenic properties are desirable, e.g., preservation of bone mineral density in postmenopausal women or decrease of low-density lipoprotein cholesterol.

Toremifene is a synthetic analogue of tamoxifen and, like tamoxifen, is bound specifically to ER.11 So far, toremifene has not been found to be mutagenic or carcinogenic in experimental studies.12, 13, 14 While tamoxifen has been available in Finland since 1980 for the treatment of metastatic breast cancer and later for adjuvant treatment, toremifene has been in clinical trials since 1982 and came into clinical use about 10 years later. To our knowledge, there are no population-based studies concerning the association of toremifene and endometrial cancer.

Our aim was to assess the risk of subsequent primary endometrial cancer associated with the use of tamoxifen or toremifene as breast cancer therapy since 1980 in Finland. The analysis included all subsequent endometrial cancers diagnosed by the end of 1995 among the 38,000 breast cancer patients diagnosed between 1980 and 1995.

BMI, body mass index; CI, confidence interval; ER, estrogen receptor; FCR, Finnish Cancer Registry; HRT, hormone replacement therapy; LFU, length of follow-up; OR, odds ratio; PR, progesterone receptor; RR, relative risk.

MATERIAL AND METHODS

The is a case-control study nested in a population-based cohort of female breast cancer patients. The FCR has collected information on all cancer cases diagnosed in Finland since 1953. All women diagnosed with breast cancer since January 1980 with a subsequent diagnosis of primary endometrial cancer before the end of 1995 served as cases, if there was no cancer other than the first breast cancer (or basal cell carcinoma of the skin) diagnosed before the endometrial cancer and if the time from the breast cancer diagnosis to the diagnosis of the subsequent endometrial cancer was at least 90 days. LFU was defined as the number of months from breast cancer diagnosis to the diagnosis of endometrial cancer.

Controls were randomly chosen among breast cancer patients who had the same year of birth and the same year of breast cancer diagnosis as the respective case patient and who were alive and free of cancer (except the first breast cancer and possible basal cell carcinoma of the skin) after LFU months of follow-up. Hysterectomized women were not taken as controls. We excluded the majority of such women at the initial control selection phase through a record linkage with the national Hospital Discharge Register (including dates of hysterectomies since 1986) and with the Finnish Mass Screening Registry (including self-reported dates of hysterectomies among participants of the organized Pap smear screening program, n = 18). Hysterectomy after LFU months was not an exclusion criterion.

After the initial exclusions, hospital records were searched for 166 cases and 3 controls for all except 1 of the cases; for 1 case there were only 2 suitable controls. The basic data extracted from the FCR database included name, date of birth, closing date (i.e., date of breast cancer diagnosis + LFU months) and the following characteristics of the breast cancer: date of diagnosis, histology, pathologic specimen numbers and hospitals which the woman visited in the context of her primary breast cancer. In addition, the date and histology of subsequent endometrial cancer were extracted for cases.

Hospitals that diagnosed the first primary breast cancer, identified by the FCR, were used as index hospitals for data collection. Data collection was continued to the subsequent hospitals and health-care centers that had participated in diagnosis, treatment and follow-up of the breast cancer. To guarantee an identical data-collection procedure for cases and controls, hospital record information related to the subsequent endometrial cancer was not used in collecting data on exposure or confounder variables. However, the date of diagnosis and clinical characteristics of the endometrial cancers were verified from hospital records. The recorded hospital data included validation of the FCR data (dates and diagnoses of the cancers, possible unregistered malignancies of the patient), height and weight, family history of breast cancer, parity, hormone use (e.g., HRT), menopausal status at the time of the first breast cancer diagnosis, date of menopause and date of hysterectomy. All treatments received for the initial breast cancer and all further treatments (including surgery, radiotherapy, cytostatics and endocrine treatments) and their dates, durations and doses up to the month preceding the diagnosis of endometrial cancer of the case or the respective month of the controls were recorded. Use of the antiestrogens tamoxifen and/or toremifene was validated against patient, hospital pharmacy or health insurance records if available.

Patient records induced exclusion of cases if there was a cancer, other than the first breast cancer, with histologic verification or with administered cancer treatment diagnosed before the endometrial cancer diagnosis (incompleteness of the FCR) or if there was an error of more than 12 months in the FCR information on date of diagnosis of either breast or endometrial cancer. Altogether 22 cases were excluded.

Controls were excluded if (i) the respective case was ineligible, (ii) there was a cancer (with positive histology or administered cancer treatment) other than the first breast cancer diagnosed during LFU months, (iii) there was an error of more than 12 months in the FCR in the date of diagnosis of breast cancer, (iv) a hysterectomy (unknown to the Hospital Discharge Register and Finnish Mass Screening Registry) was performed during LFU months or (v) the patient died or emigrated during LFU months (delay in routine follow-up system of the FCR).

In the final analysis, there were 144 cases and 368 controls, a mean of 2.6 controls per case (Table I). Of the 512 cases and controls, 147 (29%) had been treated with tamoxifen, either as adjuvant treatment or as palliative treatment for advanced breast cancer, and only 9 (1.8%) had been treated with toremifene (Table I). All 3 cases and all except 1 of the 6 controls who had used toremifene also had preceding tamoxifen treatment. The proportion of patients with unknown data on these variables was only 3% for both cases and controls. The proportion of unknown data was also low for most cofactors (Table II). From the major cofactors, the proportion was high only for estrogen/progesterone status (26% for cases and 29% for controls) and HRT (38% and 61%, respectively).

Table I. Antiestrogen use among cases and controls

Variable	Cases (n = 144)	Controls (n = 368)
Adjuvant tamoxifen
Yes	37	60
<2 years	8	19
2–5 years	22	31
>5 years	6	5
Unknown duration	1	5
No	103	300
Unknown	4	8
Palliative tamoxifen
Yes	22	31
<2 years	8	15
2–5 years	5	11
>5 years	9	4
Unknown duration	—	1
No	118	326
Unknown	4	11
Toremifene
Yes	3	6
Adjuvant	—	2
Palliative	3	4
No	137	351
Unknown	4	11

Table II. Patients characteristics of cases and controls at breast cancer diagnostics.

Variable	Cases (n = 144)	Controls (n = 368)
1 NOS, not otherwise specified.
Size of tumor
<2 cm	61	164
2–5 cm	58	160
>5 cm	10	17
Locally advanced	9	9
Unknown	6	18
Stage
Localized	88	222
Regional lymph nodes	45	109
Distant metastases	6	12
Unknown	5	25
Histologic grade
Grade 1	22	47
Grade 2	49	97
Grade 3	24	73
Unknown	49	151
Histologic type
Ductal carcinoma	95	233
Lobular carcinoma	15	28
Tubular or mucinous carcinoma	6	27
Invasive carcinoma NOS1	19	46
Intraductal carcinoma NOS	4	22
Unknown	5	12
Menopausal status
Premenopausal	21	45
Perimenopausal	6	13
Postmenopausal	117	292
Unknown	—	18
ER/PR status
ER^−/PR^−	12	51
ER^+/PR^+	73	135
ER^+/PR^− or ER^−/PR^+	22	77
Unknown	37	105
Parity
0	33	55
1	31	89
2	43	72
3	21	44
≥4	13	51
Unknown	3	57
BMI
≤30	98	277
>30	44	71
Unknown	2	20
HRT
<5 years	13	24
≥5 years	19	13
No	57	115
Unknown	55	216
Castration
Yes	4	1
No	135	357
Unknown	5	10
Radiotherapy
Yes	95	211
No	45	147
Unknown	4	10
Chemotherapy
Yes	13	26
No	127	332
Unknown	4	10

Ethical Aspects

The study followed the recommendations for biomedical research involving humans of the Declaration of Helsinki.15 Permission to identify patients from the FCR and to collect data from hospital records was given by the Finnish Ministry of Health, the heads of the clinics and the ethical committees in the hospitals involved.

Analysis

The primary outcome of the study was the risk of subsequent endometrial cancer among breast cancer patients in relation to tamoxifen or toremifene treatment. The LFU for a control was equal to that of the respective case, and no events beyond this follow-up were taken into account.

Duration of antiestrogen therapy (toremifene and/or tamoxifen) was calculated from the first prescription to the date when the last prescription was expected to be finished (discontinuation of study drug) or to the end of follow-up, whichever came first. For subjects with multiple episodes of antiestrogen therapy, total duration was calculated by summing the months contributed by each continuous period of use. Subgroups of patients who received antiestrogen for <2, 2–5 and >5 years were analyzed. Daily doses of antiestrogen were recorded and cumulative doses calculated.

Conditional logistic regression (SAS Statistical Software; SAS, Cary, NC) was used to estimate ORs and 95% CIs by the duration, cumulative dose and recency of tamoxifen or toremifene therapy. It was a priori decided that exposure to antiestrogens would be considered positive when the cumulative treatment time was at least 90 days. However, in the analyses by length of use, the category of shortest use was also considered (thresholds 0, 3, 24 and 60 months). The cumulative dose thresholds for tamoxifen were a priori set to 14.6 g and 36.5 g and for toremifene 43.8 g and 109.5 g, based on the approved doses of breast cancer treatment (for tamoxifen 20 mg/day and for toremifene 60 mg/day) for 2 and 5 years of treatment. Stratification by the characteristics of the initial breast cancer and the subsequent endometrial cancer were done when appropriate.

Potential confounding factors such as parity, height, weight, BMI = [weight (kg)]/[height (m)]² at the time of breast cancer diagnosis, use of radiation therapy, chemotherapy and HRT were considered in the analyses and included in the final model if statistically significant (p < 0.10).

RESULTS

The unadjusted OR for tamoxifen treatment (any lag, any dose, adjuvant or palliative) was 2.3 (95% CI 1.5–3.6) and that for toremifene 1.4 (95% CI 0.4–5.7). Of the potential confounders, only PR status, parity and BMI were significant (p < 0.10). PR-positive patients had an elevated OR only when the treatment had started less than 5 years before the diagnosis of endometrial cancer of the case (or the respective closing date of the control). Obese women (BMI > 30) had an about 2-fold excess risk (Table III). The OR for those having 3 or more children was decreased.

Table III. ORs for endometrial cancer related to adjuvant/palliative tamoxifen and toremifene treatment of preceding breast cancer, adjusted for each other and for significant (p < 0.10) confounders (the ORs of which are also shown)

Variable	Cases/controls	OR	95% CI	p
Tamoxifen
No	81/269	1	Ref.
Yes	59/88	2.9	1.8–4.7	<0.0001
Toremifene
No	137/351	1	Ref.
Yes	3/6	0.9	0.2–3.9	0.83
PR status
−	31/119	1	Ref.
+	76/144	1.7	1.0–2.9	0.06
Parity
0	33/35	1	Ref.
1	31/89	0.5	0.2–0.9	0.03
2	43/72	0.9	0.4–1.6	0.62
3	21/44	0.6	0.3–1.2	0.13
≥4	13/51	0.3	0.1–0.7	0.01
BMI
≤30	98/277	1	Ref.
>30	44/71	2	1.2–3.3	0.01

After adjusting for these factors, the OR for tamoxifen treatment was 2.9 (95% CI 1.8–4.7, Table III). For patients with a time interval of ≥5 years between breast cancer and endometrial cancer, the OR was 3.7 (95% CI 1.6–8.6), based on 13 cases and 22 controls. The OR for toremifene was 0.8, however, based on 3 cases and 6 controls only (95% CI 0.2–3.8); none of the cases and only 1 control had used toremifene ≥5 years before the closing date.

In addition to the confounders mentioned above, use of HRT for more than 5 years appeared to be associated with the risk of endometrial cancer, the OR for long-term HRT use being 3.4 (95% CI 1.3–8.6). Inclusion of this factor into the model did not essentially change the ORs associated with antiestrogen treatment. The OR of 2.9 for tamoxifen shown in Table III increased to 3.4 (95% CI 2.0–5.7) if the use of HRT was added into the model. However, the duration of HRT was known for only 62% of the cases and 41% of controls; i.e., the ORs may be biased. Therefore, this variable was excluded from the final models.

The OR for adjuvant tamoxifen treatment increased only moderately with the duration of treatment. However, the OR was 10.4 in relation to over 5 years' treatment of advanced breast cancer (Table III). Due to the standard daily doses of tamoxifen and toremifene, duration of the treatment directly gave a good estimate of the cumulative dose received by each patient. Duration had fewer unknown values and was therefore included in the models.

The OR for adjuvant tamoxifen treatment was highest in relation to use 23–59 months before the endometrial cancer diagnosis, while for the treatment of advanced disease the risk appeared to be especially attributable to treatment given ≥5 years before (Table IV).

Table IV. ORs for endometrial cancer related to tamoxifen treatment of preceding breast cancer, by type, duration and timing (how many years before the endometrium diagnosis) of tamoxifen use, adjusted for toremifene use, PR status, parity and BMI

Variable	Cases/controls	OR	95% Cl	p
Duration of tamoxifen use (years)
Adjuvant
No use		1.0	Ref.
<2	8/19	1.7	0.6–4.7	0.31
2–5	22/31	3.4	1.6–7.1	0.001
>5	6/5	3.1	0.9–12	0.08
Palliative
No use		1.0	Ref.
<2	8/15	3.0	1.1–8.3	0.03
2–5	5/11	1.3	0.4–4.3	0.62
>5	9/4	10.4	2.3–47	0.003
Timing of tamoxifen use (months)
Adjuvant
No use		1.0	Ref
0–23	5/15	1.1	0.3–3.8	0.85
24–59	19/23	5.1	2.1–13	0.0004
60+	13/22	2.2	0.9–5.2	0.09
Palliative
No use		1.0	Ref.
0–23	7/15	2.6	0.9–7.5	0.07
24–59	4/11	1.0	0.3–3.7	0.99
60+	11/5	9.5	2.5–36	0.001

The histologic type or size of breast cancer did not modify the ORs. Young age at breast cancer diagnosis appeared to increase the RR of having tamoxifen-induced endometrial cancer. With a threshold age of 55 years, the OR for the younger stratum (16 cases and 18 controls using tamoxifen, adjustments as in Table II) was 4.6 (95% CI 1.4–14, p = 0.01), while the OR for the older stratum (43 cases, 70 controls) was 2.6 (95% CI 1.5–4.5, p = 0.001).

The ORs in relation to tamoxifen use appeared to be somewhat higher for well-differentiated grade 1 endometrial cancer (OR 4.1, 95% CI 1.7–9.8, p = 0.002) than grade 2/3 cancers representing moderate or poor differentiation (OR 2.7, 95% CI 1.3–5.6, p = 0.007). The significance of the cofactors varied by grade: PR positivity carried an OR of 2.8 (95% CI 1.2–6.7) in the model specific to grade 2/3 endometrial cancers but an OR of 1.0 (95% CI 0.4–2.2) in grade 1 cancers. Nulliparity appeared to be a stronger risk factor for cancers of grade 1 than of grades 2 and 3.

DISCUSSION

Long-term safety is crucial in the treatment of relatively young women. Tamoxifen has proven to be effective in the adjuvant treatment of breast cancer by increasing disease-free and overall survival, while the experience concerning long-term effects of toremifene cannot (yet) be abundant. In clinical studies, a dose of 60 mg/day toremifene is equally effective as 20 mg/day of tamoxifen in the treatment of postmenopausal women with advanced breast cancer. Both tamoxifen and toremifene decrease serum cholesterol concentration and prevent bone loss, which may reduce the risk of ischemic cardiovascular disease and osteoporosis in postmenopausal women. Both drugs can be considered potential candidates in the long-term treatment or prevention of breast cancer if the risk of subsequent malignancies or other side effects is not intolerably high.

Tamoxifen, however, is genotoxic and carcinogenic in rats and increases the incidence of endometrial cancer in women on long-term treatment. Toremifene is not genotoxic or carcinogenic in rats, but long-term experience among humans has been limited. One of the aims of our study was to assess whether toremifene treatment would involve a lower risk of endometrial cancer than tamoxifen. This attempt failed due to a smaller proportion of toremifene users (1.6% among controls) in the early 1990s than was estimated a priori (5%). Although the point estimate of the OR for toremifene users was about 1.0, suggesting no excess risk, its CI was too wide to make any conclusions. Toremifene treatment in Finland increased in the second half of the 1990s, and in a few years it should be possible to study the short-term risks of its use with feasible power.

The OR for tamoxifen use varied from 2 to 4 for most strata, which is well in accordance with earlier studies.9 Long-term tamoxifen treatment has been associated with increased incidence of subsequent endometrial cancer.4–6, 16 Whether this is due to estrogenic cancer growth-promoting action or to genotoxic carcinogenic action of the drug, as suggested by some studies,7, 12, 17–19 remains to be clarified.

We observed a 10-fold risk (95% CI 2.3–47) related to over 5 years' tamoxifen treatment for advanced breast cancer but no such increase in relation to duration of adjuvant use. In another case-control study, a similar RR was observed among adjuvant patients treated with tamoxifen for more than 5 years.20 In 2 other case-control studies,21, 22 the RR for endometrial cancer also clearly increased with duration of tamoxifen treatment, but these studies did not separate palliative and adjuvant therapy groups.

When our study was first designed, it was believed that the excess risk of endometrial cancer is only typical among patients over age 50 years at the time of the initiation of tamoxifen treatment. Shortly after, it was reported that younger patients also may develop endometrial cancers.5 In our results, younger patients had higher ORs than those above 50 years at the beginning of tamoxifen treatment.

The nationwide FCR is known for its high coverage and accuracy.23 The accuracy of coding multiple cancers has been verified through numerous comparisons with clinical data, including large-scale international studies.24, 25 The hospital records of the cases and controls revealed only few inaccuracies in the FCR data. Follow-up to emigration or death through the Finnish population registration system is complete; therefore, all controls were truly followed up for the same time as their respective cases.

Bernstein et al.21 addressed the possible combined effects of tamoxifen, weight at diagnosis of breast cancer and estrogen-replacement therapy on the risk of endometrial cancer. We had data on those factors, and adding BMI into the model significantly improved the fit. In the present study, the ORs associated with obesity (BMI > 30) in different multivariate models varied from 1.8 to 2.4, which corresponds well with the RR estimates published earlier in numerous studies. Also, the low ORs in relation to high parity observed in our study are in line with the confirmed protective effect of childbirths in the etiology of endometrial cancer, thus indicating high-quality data.

Our data collection was based on active search through hospital records, thus avoiding selective response activity or recall bias. Data collection was successful for most of the numerous data items included in the study protocol. Only HRT was poorly documented, being unknown for 38% of the cases and 59% of the controls.

Our study was designed to provide excellent data quality and nonselectivity, and the ORs are thus reliable. There was not time enough to assess the risk of subsequent endometrial cancer associated with toremifene use. The main contribution of our study is to shed more light on the cancer risk in relation to tamoxifen use. According to our results, 35–40 endometrial cancer cases in the entire Finnish female population of 2.6 million during our 16-year follow-up period were attributable to tamoxifen, affecting 0.1% of the 38,000 breast cancer patients in the study base (or 0.4% of the 9,000 patients treated with tamoxifen, as estimated from the proportion of tamoxifen-treated patients among the controls). Tamoxifen increases the risk of endometrial cancer, but this excess should be considered small in comparison to the gains for breast cancer patients in prevention of recurrences and contralateral breast cancer and good palliation in advanced breast cancer.

Acknowledgements

We are grateful for Dr. S. Maijanen for the entire collection of data from the clinical records in all hospitals and heath-care centers. Data collection for cases and controls was funded by Orion Pharma, the pharmaceuticals division of the Orion Group and the developer and manufacturer of toremifene (Fareston). We also thank Mr. J. Ellmén and Mr. E. Mäntylä from Orion Pharma for providing consultation on the molecular biology of antiestrogens. The research team designed the data collection, coding and analyses as well as wrote the Results and Discussion independently from Orion Pharma.