healthmedicinet

The E3N cohort study

The E3N (Etude Epidémiologique auprès de femmes de la Mutuelle Générale de l’Education Nationale) study is an ongoing prospective cohort involving 98,995 French female volunteers,
born between 1925 and 1950 and members of a national teachers’ health insurance plan
16]. The E3N study was initiated to identify female cancer risk factors, especially dietary,
hormonal and reproductive factors. The study began in 1990 when participants returned
the first questionnaire and gave signed informed consent. Since then women have completed
self-administered questionnaires sent by mail every 2 to 3 years about their lifestyle,
health status and medical history. The E3N study protocol was approved by the French
National Commission for Data Protection and Privacy.

Study population

Women included in our study population were participants of the E3N study, who had
completed the E3N diet history questionnaire (DHQ) in 1993 and were followed until
June 2008. Among the 74,522 women with dietary data, women having under- or over-reported
their consumption, that is, those in the bottom and top 1% of the energy intake to
basal metabolic rate ratio computed on the basis of age, height and weight, were excluded
(n?=?1,360). In addition, we excluded women who reported cancer diagnosis (except
non-melanoma skin cancer) before completing the E3N DHQ (n?=?4,705) along with those
who developed in situ breast cancer (n?=?483) and those for whom follow-up information was unavailable
(n?=?822). We further excluded 3,302 women with missing body weight (BW) in 1993 and
20 women who had never menstruated. Finally 63,830 women were included in our study
population.

Breast cancer cases

From the third self-administered questionnaire to the ninth, data on health status
and medical history of women, including cancer occurrence, were collected and updated.
A total of 3,465 women reported primary invasive breast cancer between 1993 and 2008
and 92% of them were confirmed by pathology reports. Information on estrogen receptor
(ER) and progesterone receptor (PR) were extracted from these reports, and invasive
breast cancer cases were classified as ER+/PR+ (n?=?1,596), ER+/PR- (n?=?561) or ER-/PR-
(n?=?414). Because of the small number of cases, the ER-/PR+ invasive breast cancers
were not considered in the analyses (n?=?107). Because the proportion of false-positive
self-reported breast cancer in the E3N cohort was 5%, breast cancers not confirmed
by pathology report (n?=?285) were considered as cases. Deaths were identified through
family contact or from insurance files, and causes of death were obtained from the
French National Service on Causes of Death 17].

Assessment of dietary dioxin exposure

Consumption data

Dietary data were collected from a validated self-administered DHQ, specifically developed
for the assessment of the previous year’s usual diet of the E3N population and sent
in 1993 18]. The E3N DHQ covered the daily consumption of 208 food items by collecting food frequencies
and portion sizes for 8 meals and snacks during the day. The questionnaire comprised
both a quantitative and a qualitative part. In the quantitative part, women were asked
to indicate the frequency of consumption and to estimate portion sizes of food groups,
food items and beverages. Portion sizes were estimated with a validated photo booklet
19]. The qualitative part provided detailed information on women’s relative consumption
frequency of specific food items within one of the food groups of the first part.
Furthermore, a food composition table was generated from a French national database
20], providing information on fat content of most of the food items assessed in the E3N
DHQ.

Contamination data

An assessment of the level of food contamination by dioxins (PCDDs and PCDFs) in France
was published in 2000 by the French High Council for Public Health (Conseil Supérieur d’Hygiène Publique de France (CSHPF)) 21]. These contamination data were based on food samples collected from 1996 to 1998
which was the closest period to the E3N DHQ. Food samples analyzed in this report
(n?=?444) included dairy products, meats, seafood products, cereal products, fruits
and vegetables, eggs and fats (Additional file 1: Table S1). They were collected from monitoring and control plans mostly performed
by the Directorate for food of the French Agriculture ministry (DGAL) and the General
Directorate for Competition Policy, Consumer Affairs and Fraud Control (DGCCRF). Measurements
of dioxin contamination in these samples were performed in one single laboratory (Carso
Laboratory, Lyon, France) with a certified method using high-resolution gas chromatography/high-resolution
mass spectrometry (HRGC/HRMS). Dioxin concentrations were calculated with the lower-bound
method, assuming that non-detected congeners were equal to zero. Results were expressed
in pg TEQ/g lipid weight, except for foods from vegetal origin (that is, cereal products
and fruit and vegetables) for which results were expressed in pg TEQ/g fresh weight.

Exposure assessment

Dietary dioxin exposure in our study was estimated for each woman, by combining consumption
data from the E3N DHQ and lipid levels from the food composition table developed by
the E3N team with food dioxins contamination data from the CSHPF report (Additional
file 1: Table S1), using the following formulas recommended by the World Health Organization
(WHO) 22]:

Equation 1:

(1)

Equation 2, for cereal products and fruit and vegetables:

(2)

where E_i is the dietary dioxin exposure for the woman i (in pg TEQ/kg BW/day), C_i,k is the consumption of the food k by the woman i (in g/day), F_k is the proportion of lipids of the food k (in %), L_k is the average dioxin contamination of the food k (in pg TEQ/g lipid weight, see equation 1), P_k is the average dioxin contamination of the food k (in pg TEQ/g fresh weight, see equation 2) and BW_i is the body weight of the woman i (in kg) reported in the E3N questionnaire of 1993.

The intake of PCDDs and PCDFs was calculated in toxic equivalent (TEQ) using toxic
equivalence factors (TEF) first developed by the North Atlantic Treaty Organization
(NATO) in 1989 23]. In 2001, the WHO established a dioxin-tolerable daily intake of 2.3 pg TEQ/kg BW/day
24],25].

Statistical analysis

Participants contributed person-years of follow up from the date they completed the
E3N DHQ to the date of diagnosis of primary invasive breast cancer, date of death,
date of return of the ninth questionnaire or June 2008, whichever occurred first.
Estimated dietary dioxin exposure was categorized according to quartiles of its distribution
in the study population. Baseline characteristics of the participants were described
by quartiles of estimated dietary dioxin exposure using mean and standard deviation
(SD) for quantitative variables and frequency and percentage for qualitative variables.

Hazard ratios (HR) and 95% confidence intervals (CI) for breast cancer were estimated
for an increase of 0.43 pg TEQ/kg BW/day (that is, 1 SD) and for quartiles of estimated
dietary dioxin exposure with first quartile used as reference, using Cox proportional
hazard models, stratified by 5-year-interval birth cohorts (from 1925 to 1950). Women’s
age was used as the timescale. To evaluate potential dose-response associations, tests
for linear trend across quartiles were derived from the Wald test of the models with
the semi-continuous variable. All models were adjusted for known breast cancer risk
factors and potential confounders: height (cm), body mass index (BMI) before and after
menopause (cut off: 25, 30 kg/m²; considered as a time-dependent variable), total energy intake excluding alcohol
(kcal/day), alcohol intake (cut off: 0, 6.9 g/day), education (undergraduate/postgraduate
with a 1- to 4-year university degree/postgraduate with a 5+ year university degree),
physical activity at baseline (cut off: 34, 46, 62 metabolic equivalent task-hour
per week (MET-h/w)), smoking status (never smoked/ex-smoker/current smoker; time-dependent
variable), age at menarche (cut off: 12, 14 years), previous use of oral contraceptives
(never/ever), previous use of progestin before menopause (never/ever), menopausal
status combined with use of menopausal hormone treatment (MHT) (premenopausal/postmenopausal
non-using of MHT/postmenopausal using of MHT; time-dependent variable), age at menopause
(cut off: 47, 54 years, among postmenopausal women only), age at first full-term pregnancy
combined with number of live births (no child/?2 children and age at first birth 30 years/?2
children and age at first birth ?30 years/?3 children), breastfeeding (never/ever),
previous family history of breast cancer (yes/no), previous history of personal benign
breast disease (yes/no; time-dependent variable), previous mammography (yes/no; time-dependent
variable). Further adjustments for food group consumption (including fruit and vegetable,
fish and seafood, dairy products or total dietary flavonoid intake) were made. All
adjustment variables had less than 5% missing data that were replaced by their modal
or median value in the population. Information on estimated dietary dioxin exposure
was available for all women of the study population.

We also estimated risk of invasive breast cancer according to hormone receptor status
(ER/PR). We excluded cases for whom joint hormone-receptor status (n?=?787) was missing
from the corresponding analyses. Interactions between BMI and estimated dietary dioxin
exposure were tested among premenopausal and postmenopausal women in relation to breast
cancer risk because of opposite associations between BMI and risk of breast cancer
before and after menopause 26] and the storage of dioxins in adipose tissues. Analyses were also stratified by weight
change during follow up (cut points??2 kg/5 years; 2 kg/5 years) and dietary patterns
(“Western” vs. “Healthy” patterns 27]). Because of previously observed negative associations, we studied interactions between
estimated dietary dioxin exposure and smoking status, and breastfeeding 28],29] while interaction between estimated dietary dioxin exposure and alcohol consumption
was studied due to potential positive association 30]. All tests for interaction were derived from the likelihood-ratio test comparing
the models with and without an interaction term. Sensitivity analysis was performed,
excluding breast cancer cases diagnosed less than 2 years after completing the E3N
DHQ, to eliminate possible preclinical tumors. All P-values were two-sided and the significance level was set at 0.05. We used the SAS
statistical software version 9.3 (SAS Institute Inc., Cary, NC, USA) for data analysis.