Postmenopausal Serum Sex Steroids and Risk of Hormone Receptor–Positive and -Negative Breast Cancer: a Nested Case–Control Study

Materials and Methods

EPIC is a multicenter prospective cohort study designed to investigate the relationships between diet, nutrition and metabolic factors, and cancer, consisting of 366,521 women and 153,457 men aged mostly between 25 to 70 years (16, 17). All participants were enrolled between 1992 and 2000 from 23 centers in 10 European countries: Denmark, France, Germany, Greece, Italy, Norway, Spain, Sweden, The Netherlands, and the United Kingdom.

Baseline anthropometric measurements and questionnaire data on habitual diet, reproductive and menstrual history, exogenous hormone use [oral contraceptive (OC) and HRT use], medical history, lifetime smoking and alcohol consumption history, occupation, level of education, and physical activity were collected. Blood samples were also collected for most participants. Details about the standardized procedures for recruitment, measuring baseline anthropometry (height, weight, waist, and hip circumferences), questionnaires and biological sample collection at study centers are given elsewhere (16, 17). Sweden and Norway were not included in the current analysis because independent studies were being completed on breast cancer risk, or because a blood serum sample was not available. All subjects gave written informed consent to use their questionnaire data and future analyses of their blood samples. The Internal Review Boards of International Agency for Research on Cancer (IARC) and all EPIC recruitment centers approved the analyses of serum samples.

Blood samples were collected according to standardized protocols. From each subject, 30 mL of blood was drawn and after centrifugation, blood fractions (serum, plasma, buffy coat, and red blood cells) were aliquoted in 28 plastic straws of 0.5 mL each (12 plasma, 8 serum, 4 erythrocytes, and 4 buffy coat for DNA), which were heat sealed and stored under liquid nitrogen (−196°C). Half of the 28 aliquots were stored locally and the other half centrally at the IARC, with the exception of Denmark blood samples which were stored locally in 1 mL tubes at −150°C.

In all countries (except for France, Greece, and Germany) incident breast cancer cases were identified using a combination of methods employing record linkage with cancer and pathology registries. Vital status was collected from regional or national mortality registries. In Greece, Germany, and France, active follow up of cancer was through health insurance records and direct contact of participants and their next of kin was used. Self-reported breast cancer cases were all systematically verified from clinical and pathologic records. The closure date for this follow-up period was the date of last complete follow-up for both cancer incidence and vital status, which ranged between 2003 and 2006, depending on the center. Cancer incidence data were classified according to International Classification of Diseases 10th Revision (ICD-10).

Information on receptor status of participants, as well as the available laboratory methods and quantification descriptions used to determine receptor status was collected by 20 centers using the same techniques to collect incident breast cancer cases. At the time of this study, insufficient information on HER2 status (n = 37) had been collected to be included in the current analysis. To standardize the quantification of receptor status received from the EPIC centers, the following criteria for a positive receptor status were used; ≥10% cells stained, any “plus-system” description, ≥20 fmol/mg, an Allred score of ≥3, an IRS ≥2, or an H-score ≥10 (18–22).

The present analysis was based upon postmenopausal female participants with a blood sample, after a priori excluding women with prevalent cancer at any organ site (except nonmelanoma skin cancer) at baseline examination. Women were considered postmenopausal at the time of blood collection if they had had no menstrual cycles in the last 12 months, were older than 55 years (if the menstrual cycle history was missing), or had reported a bilateral oophorectomy. All women included in this analysis were not using HRT at the time of blood donation.

This study expands on a previous case–control study nested within EPIC on postmenopausal breast cancer risk and endogenous hormone concentrations (10). Additional cases were women who subsequently developed breast cancer after blood donation and before the end of the study period. For each case subject, up to 2 control subjects with a blood sample were chosen at random among appropriate risk sets consisting of all cohort members alive and free of cancer at the time of diagnosis of the index case. An incidence density sampling protocol was used, such that controls could include subjects who became a case later in time, while each control could be sampled more than once. The matching criteria were the study recruitment center, age at blood donation (± 6 months), time of the day of blood collection (± 1 hour), and fasting status (<3 hours, 3–6 hours, >6 hours).

Among the cases included in the study conducted in 2004 (10), only 49% had information on hormone receptors and were included in the present analysis (329 cases plus 596 matched controls; ref. 10); this part of our previous study is now referred to as “study phase 1.” After completing a new round of follow-up in EPIC, all newly identified ER-negative cases, plus an equal number of ER-positive cases along with their matched controls were included (225 cases and 225 controls; this is now referred to as “study phase 2”). Overall, a total of 554 (382 ER-positive and 172 ER-negative) cases and 821 matched controls were thus included in this analysis.

Hormone assays in study phase 1 were carried out at IARC, whereas in study phase 2 hormone assays were done at the German Cancer Research Center (DKFZ), using the same assays whenever possible.

In study phase 1, estradiol concentrations were determined using a radioimmunoassay with a double-antibody system for the separation of free and bound antigen [Diagnostic Systems Laboratories Inc. (DSL)]. Because this assay was no longer produced by the company when study phase 2 started, estradiol was assayed using a similar double-antibody radioimmunoassay (DiaSorin). In both study phases, testosterone concentrations were measured with the same radioimmunoassay (Immunotech). A solid phase “sandwich” immunoradiometric assay (CIS Bio International) was used for the analysis of SHBG levels.

Serum samples were thawed once and all hormone measurements done on one day to avoid additional freeze-thaw cycles. Laboratory technicians were blinded to the case–control status of the study subjects. Cases and their individually matched controls were always analyzed within the same analytical batch.

Mean intrabatch coefficients of variation were 5.8% and 11.4% for estradiol, 10.8% and 8.2% for testosterone, and 8.0% and 3.1% for SHBG for study phases 1 and 2, respectively. Interbatch coefficients of variation were 13.1% and 13.4% for estradiol, 15.3% and 14.0% for testosterone, and 16.5% and 7.1% for SHBG for study phases 1 and 2, respectively. Serum concentrations of free testosterone and free estradiol (unbound to SHBG or albumin) were calculated from mass action equations using absolute concentrations of each steroid and SHBG, and assuming a constant serum albumin concentration of 43 g/L (23, 24).

In all analyses, levels of estradiol (pmol/L), free estradiol (pmol/L), testosterone (nmol/L), free testosterone (pmol/L), and SHBG (nmol/L) were log₂-transformed to normalize their variable distributions. Statistical significance of baseline case–control differences was evaluated using conditional logistic regression. Correlations between hormones and anthropometric indices, adjusting for age at blood donation and laboratory batch were calculated using Pearson's correlation coefficient. Statistical significance of case–control differences in geometric mean hormone levels were evaluated using paired t tests of case values versus the average of the two matched controls.

Conditional logistic regression models were used to estimate ORs and 95% CIs for breast cancer subclassified by single hormone receptor status (ER-positive/negative or PR-positive/negative) or joint hormone receptor status (ER+PR+, ER+PR-, and ER-PR-) at different serum hormone concentrations. There were too few breast cancer cases with the joint ER-PR+ receptor pattern (n = 11) to be considered as a separate outcome category. The risk associated with serum levels of the sex steroid hormones was examined both on the log₂ continuous scale and in tertiles.

To statistically account for the differences observed between study phase 1 and study phase 2, tertile cut-points were based on the study phase–specific hormone distributions in control subjects.

Likelihood ratio tests were used to assess linear trends in ORs with increasing exposure level with assigned quantitative scores 1, 2, and 3 for the tertile categories. Heterogeneity between breast cancer subtypes was assessed using a log likelihood ratio test to assess conditional logistic regression models with and without interaction terms for breast cancer subtype outcome (ER-positive ER-negative, PR-positive, PR-negative or ER+PR+, ER+PR-, ER-PR-). Interaction terms were created by multiplying breast cancer subtype with the linear trend over the tertile score of hormone levels.

The effects of potential confounders (other than those accounted for by the matching criteria) were examined by adding regression terms into the logistic regression models. The categorical variables included age at menarche (<12 years, 12 years, 13 years, 14 years, and >14 years), age at first childbirth (nulliparous, <23 years, 24–25 years, 26–28 years, >29 years), number of full-term births (nulliparous, 1 full-term birth, 2 full-term births, 3 full-term births, and 4 or more full-term births), history of breastfeeding (ever vs. never), OC (ever vs. never), hormone therapy use (ever vs. never), age at menopause, smoking status (current, former never), alcohol consumption (<1.5 g/d, 1.5–10 g/d, 10–20 g/d, and >20 g/d), physical activity (active, moderately active, moderately inactive, and inactive; ref. 25), education level (none, primary school, technical/professional school, secondary school, longer education including university degree, or not specified) and body mass index (BMI; continuous). Missing values (generally <2%) were accounted for by creating an extra category in each covariate.

To assess differences in risk estimates by subgroups, analyses based on the continuous log₂ scale were stratified by study phase, time between blood donation, and diagnosis less than and greater than 2 years, age at diagnosis (greater than/less than 64 years), and BMI above and below the median. Analyses by subgroups of BMI were done using unconditional logistic regression, adjusting for matching factors. Sensitivity analyses excluding past users of OC or HRT were also done. Heterogeneity between the subgroups was assessed using Cochrane's Q-statistic (26).