Protocol
- Home
- Scientific area
- Protocol
Constances, a unique research resource
One of the great innovations of modern epidemiology has been the development of large multi-purpose prospective population-based cohort studies, that function as research platforms. Constances is such a platform. This website aims to inform international researchers of the availability of this resource for their research needs.
Constances is a prospective cohort platform that combines:
- large numbers of participants from the general population of France;
- breadth of data covering many classes of health and medical variables, many classes of personal and environmental characteristics of participants, and many types of cofactors needed to disentangle causal relationships;
- prospective repeat data collection to monitor changes in these various variables;
- high quality data from many sources to ensure the validity of observed results.
Origin and purpose of Constances
This program was conceived and developed by a group of investigators affiliated with French universities and research institutes in the early 2000’s; the recruitment of participants and initial data collection began in 2012 and was completed in 2020. The cumulative number of active participants in the database is 220,000.
From the outset, the purpose was to provide the research community (French and international) with a resource that would be almost impossible for any single researcher to create, but which could be used by scores of researchers to address and answer myriad research questions regarding health and its determinants. It was also conceived as a resource that could be used by public health authorities in France to address questions that are of particular relevance to France. But this latter desideratum in no way compromises the ability to address the generic scientific issues.
Funding and institutional support
Constances is funded by a French government-supported program entitled Investments for the Future managed by the French National Research Agency and it is supported by two other public agencies, the National Health Insurance Fund (CNAM) and the National Retirement Fund (CNAV). In addition to these supports, Constances directors receive salary support from some French universities. Constances is recognized officially as a National Biology and Health Infrastructure by the French Ministry of Research.
Recruitment of Constances participants
In constituting the Constances sample of subjects, the ideal target population was the adult general population of France. For practical reasons, selection of eligible subjects to solicit was based on the following operational criteria:
- Registered with the French Social Security system (French acronym “CNAM”). Over 85% of French residents are registered with CNAM.
- Residents of 21 French geo-administrative units that are called “departments”. These were selected to constitute a geographic and socio-cultural-industrial cross section of the entire French population. In each of these departments, there is a clinic associated with the national Social Security system (called CES, by the French acronym) where baseline interviews and health examinations can be carried out on behalf of Constances.
- Aged 18 to 69.
Invitation to participate: Using the list of residents registered with the CNAM in each targeted department, a random age and SES stratified sample was selected in numbers sufficient to account for anticipated non-participation rates. Selected subjects were contacted by postal mail and invited to participate in Constances.
In very general terms, participation consisted of agreement to undergo medical and health examinations, to answer questionnaires, and to authorize national French administrative agencies to transmit data to Constances. Further supplementary participation consisted of providing biospecimens as solicited by Constances.
Designation of “participant”: Participants are people who agree to fill in questionnaires and who undergo a comprehensive health examination at a CES, where they sign an informed consent form.
Approximately 8% of solicited people agreed to participate. Recruitment of participants took place for the most part from 2012 to 2019, and ended in 2021, by which time the cohort numbered around 220,000 volunteer participants.
Some characteristics of cohort participants
The Constances cohort includes a sample of subjects that is diverse in terms of age, gender and socio-occupational category, with a slight majority of women (54%). The 220,000 participants enable the study of numerous health problems and risk factors. For example, by the end of 2023, more than 25,000 cancer diagnoses had been recorded among cohort members, of which 15,000 had occurred before inclusion and 10,000 after inclusion. With response rate of between 60% and 70% to the annual follow-up questionnaires and less than 0.5% of participants wishing to withdraw, the cohort benefits from strong participant loyalty. The proportion of subjects lost to follow-up is extremely low, thanks in particular to linkage with the national administrative databases where subjects continue to be monitored even if they no longer participate in active data collection.
All information collected by or obtained by Constances is of course subject to the most rigorous rules of anonymity and confidentiality. An independent trusted third party manages the subjects' identifying information (names, addresses, etc.) and acts as a link between the participants and the cohort staff. This ensures that nobody has access to the data of an identifiable person.
A great deal of relevant data was collected at baseline and continues to be collected on an ongoing basis, on a great variety of topics, including health status, morbidity, mortality, recognized and potential disease risk factors and socio-demographic characteristics.
At baseline
Data were collected via questionnaires completed at home or administered by physicians in the CES:
- Socio-economic characteristics.
- Lifestyle factors.
- Personal and family medical history.
- Health behaviors.
- Job history and occupational exposures.
- Health conditions specific to women.
- Clinical and paraclinical examinations, including intensive cognitive and cognitive and physical tests for participants over 45.
- Residential history, with each address being geo-coded.
During follow-up
Active follow-up with questionnaires: During follow-up, an annual questionnaire is sent to participants, in paper form or via internet, depending on the participant’s choice. A repeat health examination at the CES clinic is conducted every 4 years.
Passive follow-up: From their national administrative databases, CNAV provides Constances with information on each participant’s socio-occupational status (employment or social benefit giving entitlement to retirement) and vital status, going back to the first job, and CNAM provides health data in the comprehensive and universal National Health Data System (SNDS) since 2007.
Ongoing data enrichment
The CONSTANCES database is regularly enriched with a wide range of data drawn from a variety of sources. Socio-economic data concerning the residential neighborhood context (deprivation index, access to healthcare, etc.) and environmental factors (pollution, pesticides, light, noise, etc.), thanks to geocoded residential diaries. Job histories linked to "job-exposure matrices" enable to infer likely past occupational exposure to various chemical, biological and physical agents.
A biobank (blood and urine samples) contains material from around 58,000 participants. This can be used to obtain surrogate biological measures of various exposures and to characterize the person’s genetic profile.
Quality control of data
Systematic quality control is conducted in the CES clinics to assess the accuracy, reproducibility, concordance, internal and external validity of the data collected, and to investigate reasons for outliers. Epidemiological research assistants make regular site visits to CES to check compliance with standard operating procedures. Data from national administrative databases undergo targeted validations to ensure the accuracy of certain medical diagnoses. Quality control includes systematic analysis of non-response rates and cleaning of questionnaire data.
Participants may differ from non-participants according to health and social characteristics. Depending on the types of analyses (descriptive, analytic) and the nature of the difference between participants and non-participants, this could lead to biased estimates of descriptive data and to biased estimates of comparative parameters like relative risks.
To palliate the potential selection bias in descriptive analyses of prevalence of diseases or potential risk factors, which is a major preoccupation for public health authorities, the Constances team has developed an innovative set of reweighting algorithms to derive population estimates of prevalences that pertain to the general French population. The reweighting algorithm uses data from a reference cohort, made up of around 435,000 people who were randomly selected from among all non-participants who were invited to participate but ended up not participating for one reason or another (refusal or non-response to the solicitation). This anonymous cohort of non-participants is nevertheless monitored (passively) via the SNDS and CNAV databases. This representative cohort is used to calculate weighted prevalences of diseases and risk factors. During follow-up, administrative data from the SNDS and CNAV ensure that even non-active participants are tracked, enabling regular adjustments for attrition. This use of anonymized non-participants has of course been approved by the strict bioethics oversight authorities.