Circulating levels of sex steroid hormones and risk of endometrial cancer in postmenopausal women†
Article first published online: 21 OCT 2003
Copyright © 2003 Wiley-Liss, Inc.
International Journal of Cancer
Volume 108, Issue 3, pages 425–432, 20 January 2004
How to Cite
Lukanova, A., Lundin, E., Micheli, A., Arslan, A., Ferrari, P., Rinaldi, S., Krogh, V., Lenner, P., Shore, R. E., Biessy, C., Muti, P., Riboli, E., Koenig, K. L., Levitz, M., Stattin, P., Berrino, F., Hallmans, G., Kaaks, R., Toniolo, P. and Zeleniuch-Jacquotte, A. (2004), Circulating levels of sex steroid hormones and risk of endometrial cancer in postmenopausal women. Int. J. Cancer, 108: 425–432. doi: 10.1002/ijc.11529
Contents of this study are solely the responsibility of the authors and do not necessarily represent the official views of the National Cancer Institute.
- Issue published online: 21 NOV 2003
- Article first published online: 21 OCT 2003
- Manuscript Accepted: 23 JUL 2003
- Manuscript Revised: 10 JUL 2003
- Manuscript Received: 12 MAR 2003
- US National Cancer Institute. Grant Numbers: R01 CA81212-01, R01 CA81200-01, R01 CA34588, P30 CA16087
- Swedish Cancer Society
- Italian Association of Cancer Research
- endometrial cancer;
- cohort study
Experimental and epidemiological data support a role for sex steroid hormones in the pathogenesis of endometrial cancer. The associations of pre-diagnostic blood concentrations of estradiol, estrone, testosterone, androstenedione, DHEAS and SHBG with endometrial cancer risk were investigated. A case-control study was nested within 3 cohorts in New York (USA), Umeå (Sweden) and Milan (Italy). Cases were 124 postmenopausal women with invasive endometrial cancer. For each case, 2 controls were selected, matching the case on cohort, age and date of recruitment. Only postmenopausal women who did not use exogenous hormones at the time of blood donation were included. Odds ratios (OR) and their 95% confidence intervals (CI) were estimated by conditional logistic regression. ORs (95% CI) for endometrial cancer for quartiles with the highest hormone levels, relative to the lowest were as follows: 4.13 (1.76–9.72), ptrend = 0.0008 for estradiol, 3.67 (1.71–7.88), ptrend = 0.0007 for estrone, 2.15 (1.05–4.40), ptrend = 0.04 for androstenedione, 1.74 (0.88–3.46), ptrend = 0.06 for testosterone, 2.90 (1.42–5.90), ptrend = 0.002 for DHEAS and 0.46 (0.20–1.05), ptrend = 0.01 for SHBG after adjustment for body mass index, use of oral contraceptives and hormone replacement therapy. The results of our multicenter prospective study showed a strong direct association of circulating estrogens, androgens and an inverse association of SHBG levels with endometrial cancer in postmenopausal women. The effect of elevated androstenedione and testosterone levels on disease risk seems to be mediated mainly through their conversion to estrogens, although an independent effect of androgens on tumor growth cannot be ruled out, in particular in the years close to diagnosis. © 2003 Wiley-Liss, Inc.
The leading hypothesis about endometrial cancer pathogenesis postulates that exposure to estrogens not opposed by progesterone increases risk of developing endometrial cancer.1, 2 The proposed mechanism is estrogen-induced stimulation of endometrial cell proliferation, which raises the probability of occurrence and accumulation of mutations.2 The ‘unopposed estrogen’ hypothesis is based on epidemiologic data and the observation that endometrial cell division rates seem to be maximally stimulated by the estradiol levels found during the early follicular phase of the menstrual cycle and are effectively zero in the presence of luteal phase progesterone.3 Epidemiological evidence in support of the ‘unopposed estrogen’ hypothesis includes the increased risk of endometrial cancer observed in users of unopposed exogenous estrogens, such as sequential oral contraceptives (OCs) and estrogen-only replacement therapy. There is no increase in risk of endometrial cancer when a progestin is added to estrogen, either continuously or sequentially for at least 10 days, although the optimum schedule for progestin administration has not yet been defined and should take into account the effect of progestin on the breast.4 Premenopausal women with anovulatory syndromes, who have progesterone deficiency, and postmenopausal obese women, who have elevated circulating estrogen levels, are also at increased risk.5, 6, 7 There is evidence to suggest that before menopause endometrial neoplasia is especially related to progesterone deficiency, while after menopause, when ovarian progesterone production has ceased, cancer risk is directly related to estrogen levels.3, 6, 8
Androgens do not seem to have a direct stimulatory effect on endometrial cell proliferation,9, 10, 11, 12 After menopause, however, when the ovarian production of estrogens ceases, an association between circulating androgen levels and risk of endometrial cancer is expected because of the aromatization of androgens into estrogens in peripheral (in particular adipose) tissues. In addition, the persistence of a strong positive association between serum androstenedione levels and endometrial cancer risk after adjustment for circulating estrone observed in a large case-control study, led Potischman et al.8 to hypothesize that abnormal endometrial cells could produce estrogens in situ from the plasma pool of androstenedione, and thus gain a growth advantage independent of circulating estrogens. Indeed, several studies have shown that endometrial cancer tissues possess aromatase activity.13, 14, 15, 16
Numerous retrospective case-control studies have shown an increase in endometrial cancer risk with higher concentrations of endogenous estrogens, mostly with levels of estrone and total estradiol.5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17 We reported previously the results of the first prospective study on endometrial cancer in postmenopausal women, a case-control study nested within the New York University Women's Health Study (NYUWHS), and showed that elevated pre-diagnostic circulating levels of estrone, total estradiol and free estradiol were related to an increased risk, whereas sex hormone binding globulin levels (SHBG) and SHBG-bound estradiol were inversely related to risk.18
We have expanded our nested case-control study through: (i) additional follow-up of the NYUWHS; and (ii) inclusion of 2 other cohorts, the Northern Sweden Health and Disease Study (NSHDS) in Umeå (Sweden), and the Study of Hormones and Diet in the Etiology of Breast Cancer (ORDET) in Milan (Italy). We describe findings on the association of circulating levels of endogenous androgens: androstenedione, testosterone, and dehydroepiandrosterone sulfate (DHEAS) with subsequent risk of endometrial cancer and the updated results on endogenous estrogens and SHBG based on 2.2 times as many cases as in the previous study.
MATERIAL AND METHODS
|Cohort||Study setting||Recruitment nt period||Cohort size||Age range at enrolment||Last complete follow-up||Cases||Controls||Median age of subjects (range)|
|NYUWHS,New York,USA||Mammographic screening clinic||1985–1991||14,275||34–65||Feb. 98||69||131||60.4 (44.2–66.9)|
|NSHDS,Umeå,Sweden||General population||1986–2001||43,268||30–70||Dec. 00||46||88||60.0 (49.2–69.6)|
|ORDET,Milan,Italy||Healthy volunteers and women attending breast cancer prevention unit||1987–1992||10,788||35–70||Jan. 97||9||17||57.8 (50.2–66.6)|
In all 3 cohorts, demographic and anthropometric data were collected at enrollment, through self-administered questionnaires in the NYUWHS and the NSHDS, and questionnaires administered by trained nurses (who also carried out the anthropometric measurements) in the ORDET. Collection of data on reproductive history, use of OC, hormone replacement therapy (HRT), medical history (including diagnosis of diabetes mellitus and intake of drugs for diabetes), and smoking varied according to study cohort. In the NYUWHS, data on reproductive history were collected at enrollment whereas data on smoking, OC and HRT use up to the index date (date of diagnosis of the case) were collected through telephone interviews of cases (88%) and matched controls (83%) or using the most recent follow-up questionnaire for participants who did not complete the interview (resulting in data available on 93% of the cases and 96% of the controls). Data on medical history were collected at baseline and updated through follow-up questionnaires. In the NSHDS, medical history and smoking information was collected at enrollment. A questionnaire on reproductive life and sex hormone use was administered prospectively to 47% of the subjects and a similar questionnaire was sent out retrospectively to all cases and matched controls to complete and update the collected information at baseline (response rate = 95%). In ORDET, these data were collected at enrollment.
In all 3 cohorts, participants were asked to donate a venous blood sample. Subjects reporting hormone use at baseline were not recruited in the NYUWHS and ORDET cohorts, and potential case and control subjects from the NSHDS who reported use of exogenous hormones at the time of blood donation were considered ineligible.
The NSHDS and ORDET components of the study included Caucasian subjects, with very few exceptions in the NSHDS. In the NYUWHS, 88% of the subjects included in the current study provided information about ethnic origin: 91% indicated that they were non-Hispanic Whites, 5% Black, 4% Hispanic and 1% other ethnicity.
Identification of endometrial cancer cases and selection of control subjects
Case subjects were postmenopausal cohort members with primary invasive endometrial cancer diagnosed 6 or more months after the initial blood donation, who were identified within the parent cohort by the date of the last complete follow-up. Follow-up in the NYUWHS consists of periodic contact by mail and telephone, as well as record linkages with state-wide tumor registries (New York, New Jersey, Connecticut and Florida) and the U.S. National Death Index. In the NSHDS, follow-up for cancer incidence and vital status was carried out through record linkages with regional and national cancer registries, and registries for all-cause mortality, using the participants' unique national identification numbers as the identity link. In the ORDET study, records were linked to the local cancer registry (Lombardy Cancer Registry) to identify endometrial cancer cases, and to the regional residents' files to check the vital status of cohort members. A total of 124 endometrial cancer cases were eligible for the study from the 3 cohorts (Table I).
Of the 105 malignancies with histological information, 66 (63%) were endometrioid, 10 (10%) serous, 1 (1%) mucinous, 1 (1%) clear-cell type; 3 (3%) were of mixed type and 24 (23%) were carcinomas or adenocarcinomas not-otherwise specified.
For each case subject, 2 control subjects were selected at random from appropriate risk sets. The risk set for a given case included all postmenopausal subjects who were alive and had not had a hysterectomy by the index date, and who matched the case on cohort, date (±3 months) and age (±6 months) at blood donation. The menopausal status of study subjects with inconclusive questionnaire data and who were 45–55 years old at recruitment was confirmed by follicle-stimulating hormone (FSH) measurement. Women were classified as postmenopausal if their FSH measurement was >12.75 IU/L. The FSH cut-off point was selected based on data provided by the kit manufacturer and an analysis of the distribution of FSH levels according to menopausal status as assessed by questionnaire, as well as according to age (<42 and >55 years), using data from more than 300 women from the 3 cohorts. Only postmenopausal subjects were considered eligible. A total of 236 control subjects were identified and included in this study (Table I).
The Ethical Review Boards of New York University School of Medicine, the University of Umeå, the Istituto Nazionale Tumori in Milan and the International Agency for Research on Cancer (IARC), in Lyon, France, periodically reviewed and approved the present study.
The hormone analyses were carried out on serum samples obtained from the NYUWHS subjects and heparinized plasma samples obtained from the NSHDS and ORDET subjects. Samples from case subjects and their matched control subjects were always analyzed within the same batch, assay kit and on the same day. Laboratory personnel were unable to distinguish between case and control samples.
For the subjects included in the initial NYUWHS study18 (Study I, n = 156, 43% of the subjects in the current study), estradiol, estrone, androstenedione and testosterone were measured by the Clinical Studies Center of Quest Diagnostics Inc (Nichols Institute, San Juan Capistrano, CA). Serum samples were subjected to organic extraction and celite chromatography and the appropriate fractions analyzed by radioimmuno-assays (RIA). SHBG and DHEAS were measured in the laboratory of Dr. Levitz at NYU School of Medicine, using an immunometric chemiluminescent assay on an IMMULITE 2000 instrument (Diagnostic Products Corp., Los Angeles, CA). To assess laboratory precision, one aliquot from a common pool generated by using serum samples from a random sample of the healthy postmenopausal NYUWHS participants was included with every other matched set after labeling to prevent identification. Intra-batch coefficients of variation were 1.7% for testosterone, 8.9% for androstenedione, 4.6% for DHEAS, 4.8% for total estradiol, 9.5% for estrone and 4.2% for SHBG.
All subsequent laboratory analyses (Study II, 204 subjects, 57% of subjects in the current study) were carried out at the Hormone Laboratory at IARC, France. To control the quality of the measurements, samples from 3 standard sera were inserted randomly in each batch. Testosterone and DHEAS were measured by RIA with reagents from Immunotech (Marseille, France); androstenedione and estrone by double antibody RIA with reagents from Diagnostic System Laboratories (Webster, TX); estradiol by ultrasensitive double antibody RIA with reagents from Diagnostic System Laboratories; SHBG by immunoradiometric assay (IRMA) with reagents from Cis-Bio (Gif-sur-Yvette, France); and FSH by an immunoradiometric assay with reagents from Diagnostic System Laboratories. The mean intra-batch coefficients of variation were 8.2% for a testosterone concentration of 0.3 ng/ml, 6.3% for an androstenedione concentration of 0.5 ng/ml, 4.6% for a DHEAS concentration of 40.0 μg/d-ml, 3.5% for an estradiol concentration of 30 pg/ml, 5.3% for an estrone concentration of 20.0 pg/ml, 4.4% for an SHBG concentration of 40.0 nmol/l and 4.2% for an FSH concentration of 10 IU/L.
Sex steroid hormones and SHBG data were log2-transformed to reduce departures from the normal distribution and make interpretations of RR for continuous variables easier, as suggested by the Endogenous Hormones and Breast Cancer Collaborative Group.22 The paired t-test was used to compare mean hormone concentrations between cases and controls (the case subject value vs. the means of her matched control subjects).23
An analysis of covariance was used to investigate subgroup differences in mean hormone levels (by case/control status, study cohort, OC or HRT use, parity, history of diabetes, body mass index [BMI] and smoking categories) adjusting for age at sampling and sub-study (if appropriate). These analyses were carried out using the Generalized Linear Models (GLM) SAS procedure.24 Spearman partial correlations, adjusted for sub-study and age were calculated between hormone variables, SHBG and BMI.
Odds ratios (OR) for disease by quartile levels of the hormone variables were estimated by conditional logistic regression models using the SAS ‘PHREG’ procedure. Quartile cut-off points were determined on the basis of the variable distributions of the case and control subjects combined for each of the three cohorts separately to account for differences in hormone levels due to measurements in serum or plasma samples. Additionally, the NYUWHS data was divided in two sub-studies according to the laboratory method used. Likelihood ratio tests were used to assess linear trends in ORs over the quartiles, giving quantitative scores of 1–4 to the 4 levels. Two-sided p-values were calculated and considered statistically significant when <0.05. The potential confounding effects of ages at menarche and menopause, parity, BMI, use of OC and HRT, smoking and diabetes were examined by including these factors in the conditional logistic regression models. In the adjusted conditional regression models missing BMI values for 12 subjects were replaced by the cohort-specific median value of BMI. Analyses limited to the cases diagnosed 2 or more years after blood donation were also conducted and χ2 tests were carried out to assess the presence of modification of the effect of sex hormones (or SHBG) by lag time between blood donation and diagnosis.
Characteristics of the study participants are presented in Table II. Mean age at cancer diagnosis was 64.3 ± 6.1 years (median = 64.9 years). The time between recruitment and cancer diagnosis for the case subjects ranged from 6.9 months to 13.3 years, with an average of 5.4 years (median 4.5 years). Ninety-six percent of the case subjects were diagnosed at least one year after cohort recruitment and 81% were diagnosed after more than 2 years. Case women were substantially heavier and with higher BMI than controls, tended to report less frequently use of OCs, but more often use of HRT. There were no significant differences between case and control subjects in the mean age at menarche or menopause, ever having had a full-term pregnancy, ever smoking, or ever having been diagnosed with diabetes.
|Variable||Cases (n = 124)||Controls (n = 236)||p for case-control difference|
|Height (cm)1||160.9 (159.6–162.2)||160.7 (159.7–161.7)||0.883|
|Weight (kg)1||70.8 (68.4–73.3)||66.4 (64.4–68.3)||0.0013|
|BMI (kg/m2)1||27.4 (26.5–28.3)||25.7 (25.0–26.4)||0.0013|
|Age at menarche (years)1||13.0 (12.7–13.3)||13.3 (13.0–13.5)||0.143|
|Never||22 (18)||39 (16)||0.534|
|Ever||97 (78)||188 (80)|
|Missing||5 (4)||9 (4)|
|Never||99 (80)||174 (74)||0.094|
|Ever||21 (17)||58 (24)|
|Missing||4 (3)||4 (2)|
|Age at menopause1||50.5 (49.6–51.4)||50.0 (49.3–50.7)||0.503|
|Never||75 (61)||169 (71)||0.114|
|Ever||44 (35)||63 (27)|
|Missing||5 (4)||4 (2)|
|Never||80 (72)||167 (80)||0.124|
|Ever||27 (28)||37 (20)|
|Never||53 (43)||87 (37)||0.524|
|Current||21 (17)||39 (16)|
|Ex-smokers||23 (18)||52 (22)|
|Missing||27 (22)||58 (25)|
|No||111 (90)||218 (92)||0.414|
|Yes||9 (7)||11 (5)|
|Missing||4 (3)||7 (3)|
In the control group, age was inversely correlated with circulating estradiol (r = −0.30, p < 0.0001), androstenedione (r = −0.24, p < 0.0003) and DHEAS (r = −0.17, p < 0.01), but was not related to estrone, testosterone and SHBG concentrations. Table III presents age-adjusted Spearman correlation coefficients between hormone variables, SHBG and BMI. BMI was positively correlated with circulating estrogens, weakly correlated with testosterone and DHEAS, inversely correlated with SHBG and not related to androstenedione levels. There were positive correlations between circulating concentrations of sex steroid hormones. In general, the estrogen-androgen correlations were weaker than androgen-androgen or estrone-estrogen correlations.
No significant differences in mean hormone values were observed according to ever-use of OC or HRT, smoking or ever having had a full-term pregnancy after adjustment for age at blood sampling, study cohort and case-control status. Hormone levels did vary according to the assays used in the original (Study I) and second study (Table IV).
|n||Median (10th–90th percentiles)||n||Median (10th–90th percentiles)|
|Study I||53||8 (4–22)||103||6 (4–13)||0.05|
|Study II||69||28.1 (16.3–54.1)||127||21.0 (11.7–44.0)||0.0001|
|Combined||122||21.0 (5.0–42.3)||230||13.3 (4.5–31.1)||0.0001|
|Study I||52||27 (14–54)||101||22 (14–38)||<0.05|
|Study II||70||22.7 (12.0–48.9)||129||17.4 (11.1–31.6)||0.008|
|Combined||122||24.0 (12.3–52.0)||230||19.9 (11.9–35.4)||0.001|
|Study I||53||0.56 (0.26–1.06)||103||0.49 (0.31–0.82)||0.30|
|Study II||71||1.15 (0.49–2.00)||133||0.97 (0.38–1.93)||0.02|
|Combined||124||0.84 (0.36–1.73)||236||0.63 (0.35–1.59)||0.01|
|Study I||53||0.24 (0.13–0.45)||103||0.24 (0.11–0.45)||0.81|
|Study II||71||0.29 (0.09–0.66)||133||0.22 (0.05–0.50)||0.03|
|Combined||124||0.28 (0.10–0.55)||236||0.23 (0.08–0.49)||0.04|
|Study I||53||67.3 (29.3–155)||103||58.8 (27.4–131)||0.14|
|Study II||71||114.8 (63.6–245)||133||79.9 (46.3–185.2)||0.002|
|Combined||124||99.3 (34.8–210)||236||70.7 (34.4–162)||0.001|
|Study I||53||45.3 (23.0–95.2)||103||59.8 (30.6–97.7)||0.05|
|Study II||71||35.4 (21.6–86.1)||133||45.6 (24.2–88.3)||0.05|
|Combined||124||42.3 (21.6–87.6)||236||50.1 (27.2–94.0)||0.006|
Three hundred eighty four postmenopausal NYUWHS subjects, who served as controls in studies of breast, endometrial and ovarian cancer, have provided a second blood sample from 12–60 months after the baseline blood donation. Sex steroid hormones and SHBG were measured in the initial and in the second blood sample of these women (a total of 768 samples) at the Hormone Laboratory at IARC, France. Intra-class correlations between repeated hormone measurements were: 0.66 (95% CI = 0.61–0.73) for estradiol, 0.58 (0.52–0.66) for estrone, 0.63 (0.57–0.70) for androstenedione, 0.64 (0.58–0.70) for testosterone, 0.92 (0.90–0.93) for DHEAS and 0.87 (0.85–0.90) for SHBG, indicating that hormone and SHBG levels are fairly stable over time in a given individual and that a single measurement can be used to characterize an individual's average level.
Median hormone levels in case and control subjects are presented in Table IV. Case subjects had higher median levels of all sex-steroid hormones measured, the differences being more pronounced for estradiol, estrone and DHEAS. The difference in median levels of estradiol and circulating androgens were more pronounced in the newly added subjects, however, there was no statistically significant heterogeneity of the quantitative relationship of risk associated with hormone levels by laboratory method or study cohort. Case subjects had significantly lower levels of SHBG than control subjects.
Odds ratios for endometrial cancer (Table V) were consistent with the differences in mean hormone levels in that an increase in risk was observed with higher levels of estradiol, estrone, DHEAS, androstenedione and testosterone, whereas SHBG levels were inversely related to risk. Adjustment for log-BMI, OC and HRT use reduced the strength of the associations of estradiol, estrone, DHEAS and SHBG with cancer risk, but they remained strong and significant, whereas the association with testosterone became marginally significant (Table V). The association of androstenedione with cancer risk became only slightly stronger in the adjusted model. Among the adjusting variables, BMI had the greatest effect, reducing the strength of the associations of the hormone variables with endometrial cancer, whereas adjustment for HRT increased the regression coefficients. Adjustment for OC use only slightly reduced the point estimates.
|Hormone||Quartile||p for trend4|
|Crude model2||1.00||1.32 (0.65–2.69)||2.19 (1.07–4.47)||5.39 (2.50–11.6)||0.0001|
|Adjusted model3||1.00||1.24 (0.59–2.62)||1.88 (0.88–4.01)||4.13 (1.76–9.72)||0.0008|
|Crude model2||1.00||1.46 (0.72–2.96)||2.10 (1.07–4.10)||4.55 (2.28–9.08)||0.0001|
|Adjusted model3||1.00||1.39 (0.66–2.93)||1.81 (0.88–3.71)||3.67 (1.71–7.88)||0.0007|
|Crude model2||1.00||1.24 (0.64–2.40)||1.47 (0.72–2.99)||2.04 (1.05–3.98)||0.03|
|Adjusted model3||1.00||1.42 (0.69–2.94)||1.61 (0.75–3.45)||2.15 (1.05–4.40)||0.04|
|Crude model2||1.00||1.65 (0.85–3.20)||2.22 (1.17–4.23)||2.06 (1.06–4.03)||0.02|
|Adjusted model3||1.00||1.62 (0.82–3.20)||2.30 (1.16–4.55)||1.74 (0.88–3.46)||0.06|
|Crude model2||1.00||1.55 (0.78–3.05)||2.29 (1.17–4.48)||3.03 (1.53–5.98)||0.0001|
|Adjusted model3||1.00||1.49 (0.73–3.02)||2.11 (1.05–4.24)||2.90 (1.42–5.90)||0.002|
|Crude model2||1.00||0.65 (0.36–1.18)||0.38 (0.21–0.71)||0.35 (0.18–0.70)||0.0006|
|Adjusted model3||1.00||0.73 (0.38–1.39)||0.41 (0.21–0.81)||0.46 (0.20–1.05)||0.01|
Restricting the analyses to the 101 women diagnosed 2 or more years after blood donation reduced somewhat the strength of the associations of estrogens and SHBG with endometrial cancer risk but they remained strong and statistically significant (Table VI). The test for homogeneity showed no evidence of effect modification by lag-time between blood donation and cancer diagnosis for estradiol (p < 0.33), estrone (p < 0.80) and SHBG (p < 0.33). The associations between endometrial cancer risk and androgens, although still positive, were weaker than in analyses including all cases and no longer statistically significant for androstenedione and testosterone. The tests for homogeneity were statistically significant for all 3 androgens (p < 0.03 for androstenedione, p < 0.04 for testosterone and p < 0.02 for DHEAS), indicating a stronger association of androgens with endometrial cancer risk in the 2 years before diagnosis than in preceding years.
|Hormone||Quartile||p for trend4|
|Crude model2||1.00||1.92 (0.87–4.28)||2.95 (1.28–6.78)||4.57 (2.00–10.4)||0.0002|
|Adjusted model3||1.00||1.82 (0.79–4.15)||2.96 (1.20–7.29)||4.06 (1.66–9.93)||0.001|
|Crude model2||1.00||1.23 (0.56–2.69)||1.26 (0.59–2.70)||3.72 (1.76–7.87)||0.0004|
|Adjusted model3||1.00||1.26 (0.55–2.87)||1.18 (0.52–2.70)||3.36 (1.46–7.69)||0.004|
|Crude model2||1.00||0.95 (0.48–1.88)||0.77 (0.36–1.67)||1.65 (0.81–3.38)||0.18|
|Adjusted model3||1.00||1.15 (0.54–2.50)||0.96 (0.42–2.20)||1.93 (0.87–4.22)||0.11|
|Crude model2||1.00||1.50 (0.73–3.06)||1.63 (0.80–3.35)||1.52 (0.75–3.10)||0.24|
|Adjusted model3||1.00||1.44 (0.69–3.03)||1.78 (0.84–3.75)||1.53 (0.74–3.14)||0.21|
|Crude model2||1.00||1.00 (0.49–2.05)||1.50 (0.75–3.01)||1.75 (0.85–3.62)||0.07|
|Adjusted model3||1.00||1.06 (0.50–2.25)||1.59 (0.77–3.27)||1.92 (0.90–4.10)||0.05|
|Crude model2||1.00||0.70 (0.36–1.35)||0.32 (0.15–0.65)||0.45 (0.22–0.95)||0.005|
|Adjusted model3||1.00||0.78 (0.38–1.60)||0.34 (0.15–0.73)||0.61 (0.25–1.49)||0.05|
BMI was directly related to endometrial cancer risk with ORs of 1.04 (0.53–2.06) for BMI between 23.1 and 25.0; 2.17 (1.12–4.19) for BMI between 25.0 and 28.3; and 2.44 (1.29–4.61) for BMI greater than 28.3 when compared to women with BMI < 23.1 (ptrend < 0.008).
Adjustment of estrogen-endometrial cancer models for levels of circulating androgens or SHBG slightly reduced the strength of the associations, but all models remained statistically significant (Table VII). In contrast, adjustment for estrogens resulted in a decrease in the effect of androstenedione and testosterone (ranging from 17–24%) and a loss of statistical significance. Adjustment of DHEAS models for estrogens weakened the association with endometrial cancer risk, but the estimates remained significant after adjustment for estrone and very close to being significant after adjustment for estradiol (Table VIII).
|Estradiol||2.06 (1.47–2.89)||1.94 (1.37–2.76)||2.00 (1.39–2.88)||1.87 (1.32–2.66)||1.95 (1.39–2.75)|
|Estrone||1.84 (1.31–2.58)||1.69 (1.15–2.48)||1.78 (1.21–2.62)||1.52 (1.06–2.19)||1.69 (1.19–2.38)|
|Androstenedione||1.52 (1.08–2.14)||1.22 (0.85–1.75)||1.19 (0.81–1.75)|
|Testosterone||1.26 (1.02–1.57)||1.05 (0.83–1.32)||1.04 (0.81–1.33)|
|DHEAS||1.51 (1.16–1.96)||1.30 (0.99–1.72)||1.40 (1.04–1.88)|
|SHBG||0.66 (0.49–0.90)||0.74 (0.54–1.01)||0.71 (0.52–0.99)|
Restricting the analyses to women with endometrioid tumors (59 case-control sets), strengthened the associations of estradiol, estrone, androstenedione and testosterone with endometrial cancer risk, did not influence the association with SHBG and abolished the statistical significance of the association with DHEAS. Excluding subjects with FSH values <30 IU/L (a more conservative cut-off point to assess menopausal status) did not alter the observed associations. The associations of estrogens with endometrial cancer risk were slightly stronger in the group of women who never used HRT than in the group of ever-users, however, the associations were not statistically heterogeneous.
With this study, we confirmed our previous observations of a direct association of pre-diagnostic circulating estrogen levels and an inverse association of SHBG concentration with endometrial cancer risk in postmenopausal women.18 Similar results were reported by the 2 largest retrospective case-control studies of post-diagnostic endogenous sex-steroid hormone levels and endometrial cancer.8, 17 These associations persisted in analyses limited to cases diagnosed 2 or more years after blood donation.
In overall analyses circulating androgen levels were also directly related to endometrial cancer, although less strongly than circulating estrogen levels. Adjustment of androstenedione and testosterone models for estradiol or estrone levels resulted in a decrease in the regression coefficients and loss of statistical significance, whereas adjustment of the estrogen models for androgen levels only slightly decreased the estimates and the associations remained significant. These observations suggest that estrogens are a major determinant of endometrial cancer risk, while circulating androgens contribute to endometrial cancer development mainly as precursor hormones for the synthesis of estrogens. Some independent effect of androgens on tumor growth cannot be ruled out, however, in particular in the years close to diagnosis. The association of androgens with endometrial cancer risk was weaker in analyses limited to cases diagnosed 2 years or more before diagnosis than in overall analyses, and we observed a significant interaction between androgens and lag time between blood donation and diagnosis, indicating a stronger association of androgens with endometrial cancer risk in the 2 years before diagnosis than in the preceding years.
Based on the results of their case-control study, showing a strong positive association between androstenedione levels and endometrial cancer that persisted after adjustment for circulating estrone concentrations, Potischman et al. proposed that early in the neoplastic process abnormal endometrial cells may acquire the ability to produce estrogens locally from the plasma pool of androgens and thus gain a growth advantage that is independent of circulating estrogen levels.8 This hypothesis is supported by in vitro studies that showed an increased aromatase activity of endometrial neoplastic cells, but not of normal endometrial cells.13, 14, 15, 16 Such a mechanism could explain our results as well as the results of the study by Potischman et al.,8, where hormonal assays were carried out on serum specimens collected from the case subjects after endometrial cancer diagnosis. If, indeed, such an increased capacity for aromatization is a particular characteristic of endometrial neoplastic cells, then the association of circulating androgens with risk would be expected to be stronger in retrospective case-control studies than in prospective studies in which androgen levels are measured before tumor development and diagnosis of the disease. Furthermore, in prospective studies, the association would be expected to be stronger in the years immediately before diagnosis than in the years further away from it, as we observed in our study. Further epidemiologic investigations with larger numbers of subjects and studies on specimens obtained from normal and neoplastic endometrium would be necessary to establish the role of elevated androgen concentrations in endometrial cancer development and progression.
Pre-diagnostic concentrations of the exclusively adrenal androgen DHEAS were also directly and strongly associated with risk and this association persisted after adjustment for estrogens. DHEAS can be involved in endometrial cancer pathogenesis through several mechanisms. DHEAS is the most abundant, albeit weak, androgen and is the major source of circulating DHEA.25 It has been proposed that DHEA not only exerts an androgenic but also estrogenic effects in postmenopausal women,26, 27 due to the binding of either DHEA itself or its metabolite 5-androstene-3β, 17β-diol to vacant estrogen receptors.27, 28 It is possible that the persistence of the association of DHEAS with endometrial cancer risk after adjustment for levels of estrogens might be due to this additional indirect estrogenic effect of DHEAS. Additionally, DHEAS has a long half-life in the circulation and can be used as a marker of adrenal sex-steroid synthetic activity, which may be of particular importance after menopause and reduction of the ovarian sex steroid synthesis.
Mean levels of sex-steroid hormones varied according to the type of hormonal assay applied: indirect methods (including a purification step) or direct (no-extraction) methods. A validation study conducted at the Hormonal Laboratory at IARC (where all estradiol, estrone, androstenedione and testosterone measurements by direct kits for this study were carried out) showed high correlations between measurements carried out by an indirect (reference) and direct methods with preservation of the relative ranking of the subjects.29 Additionally, a recent study by Dorgan et al. concluded that although absolute concentrations may differ, measurements by mass spectrometry and RIA after an extraction step correlate well for the hormones measured in the present study30 and a large pooled study of breast cancer did not show any important differences in risk estimates calculated from hormone measurements obtained by direct or indirect methods.22
We carefully selected the direct assay kits for the hormonal analyses in this study on the basis of: (i) high correlation with the hormone values obtained by the reference indirect method; (ii) good performance of the direct kits in studies of hormone reproducibility over time; and (iii) mean absolute hormone concentrations measured.29 To minimize variations or bias due to assay type or batch, the samples of all members of a given matched set were always analyzed in the same batch using identical assays, and, for the statistical analyses hormone quartiles were defined on the basis of laboratory method and cohort study. Thus, we believe that differences in hormone levels according to the laboratory methods applied and blood sampling procedures in the parent cohort studies should not have influenced substantially the associations of sex steroid hormones with cancer risk observed in our study.
In conclusion, the results of our pooled prospective study on endometrial cancer showed a strong direct association of circulating estradiol and estrone levels and an inverse association of SHBG levels with endometrial cancer in postmenopausal women. The association of endometrial cancer risk with androstenedione and testosterone concentrations in postmenopausal women seems to be primarily due to their role as precursor hormones for estrogen synthesis, although they may have some independent effect, especially during endometrial tumor progression. Elevated concentrations of DHEAS also conferred an increased risk of endometrial cancer that may be mediated through its conversion to more active sex-steroids or to its correlation with the activity of the adrenal glands for synthesis of sex-steroids.
Y. Afanasyeva, L. Quinones and D. Masciangelo provided technical assistance in the NYU Women's Health Study; Ågren, H. Sjodin and L. Marklund helped with the management of the Swedish Biobank database; D. Del Sette, E. Meneghini and E. Mugno for helping with the ORDET database; D. Achaintre and J. Bouzac contributed to the laboratory analyses and J. Dehedin in the manuscript preparation.
- 7VainioH, BianchiniF, editors. Weight control and physical activity. Lyon, France: IARC Press, 2002.
- 24SAS Institute. SAS.STATR User's Guide Version 6. SAS manual 4(6). Cary, NC, USA: SAS Institute Inc., 1990.