The French E3N Prospective Cohort Study (Etude Epidémiologique auprès de Femmes de la Mutuelle Générale de l'Education Nationale)

Who is in the cohort? In June 1990, a questionnaire was sent to 500 000 women who had been born between 1925 and 1950 and were insured by the Mutuelle Générale de l'Education Nationale (MGEN), a national health insurance plan that primarily covers teachers. The questionnaire was sent along with a leaflet explaining that an Inserm research team was launching a study of cancer risk factors and that participation would require filling in questionnaires every 2-3 years as well as the submission of a signed consent form providing permission to obtain information about each participant's vital status, address changes and medical expense reimbursements from the insurance plan. Nearly 100 000 women volunteered.

How often have they been followed-up? Until now, nine follow-up questionnaires have been sent every 2-3 years from the baseline. Approximately half of the answers were obtained after the first mailing. Two reminders were sent thereafter. The questionnaires were accompanied with newsletters that informed participants about the major results obtained to date. The participation rate remained high (77-92% according to the questionnaires) and the lost to follow-up rate was minimal because of the ability to trace non-respondents through their insurance plan files. The questionnaires were accurately filled in, with few missing or unacceptable answers.

What has been measured? To date, 11 self-administered questionnaires have been sent. The collected data are sociodemographic factors, anthropometric measurements, reproductive factors, hormonal treatments, health behaviour and lifestyle. Each follow-up questionnaire also recorded the participant's health status. The questionnaires are available at www.e3n.fr .

The questionnaires are anonymous and identified with an identification number and pin code that can be rapidly scanned to identify the respondents. The questionnaires are optically scanned and all answers are checked on screen. The scanned images are saved to allow data entry at a later time, including information regarding the addresses of medical doctors (18 000 to date) or drug names (pre-listed to avoid errors). The longitudinal data (repeats of identical questions for the purpose of updating information about topics such as menopause or smoking) are routinely homogenized. Several validation studies (e.g. dietary and anthropometrical data studies) have been performed and have revealed very satisfactory results.

Self-reported cases of cancer are validated and coded after reviewing the pathology reports obtained from medical practitioners, and nearly 90% of all cancer cases are histologically confirmed. Other diseases are also validated (e.g. diabetes, myocardial infarction, stroke, Parkinson disease) by requesting additional information about the participants (e.g. glycosylated haemoglobin levels, fracture-related circumstances, drug names) and sending questionnaires to medical doctors.