Participants were part of the NutriNet-Santé Study, a large Web-based prospective observational cohort. It is implemented in a general population targeting Internet-using adult volunteers aged 18 years and older. The design, methods, and rationale have been described elsewhere [29]. Eligible participants were recruited by different means. Initially, a vast multimedia campaign (television, radio, national and regional newspapers, posters, and Internet) called for volunteers and provided details about the study’s website [12]. The multimedia campaigns are repeated every 6 months. Recruitment information was maintained on a large number of websites (national institutions, city councils, private firms, Web organizations) and is regularly updated via professional channels (eg, doctors, pharmacists, dentists, business partners, municipalities). The key message delivered in the call for volunteers was:

During each multimedia recruitment campaign and during the enrollment process, participants are informed that follow-up over at least 10 years is planned.

Previous findings showed that most of the participants enrolled after hearing about the study on television because this medium entails the widest reach [12]. In particular, television announcements permitted the recruitment of members of population subgroups that are not typically well represented in population-based epidemiological research, those belonging to lower socioeconomic strata. The radio, newspapers, Internet, and advice from acquaintances also proved to be substantial means of disseminating information about this epidemiological study to encourage participation.

To be included, participants have to fill in on the website an initial set of questionnaires assessing dietary intake, physical activity, anthropometrics, lifestyle, and socioeconomic conditions along with health status. Participants were informed by email that, after inclusion, they would be asked to complete the same questionnaires each year as part of their follow-up. In addition, they are invited to fill in a complementary questionnaire each month. Aspects related to convenience of participation (ie, ≤20 min each month) and confidentiality were also emphasized. In addition, a system of boosting motivation and retention was implemented. In order to forge a sense of community that helps advance research, participants receive a NutriNet-Santé membership card at inclusion and a certificate on completion of each follow-up year/wave. They also receive monthly email with scientific information regarding health and nutrition, and invitations to press conferences about the study results. For purposes of retention, free screening tests for cholesterol, triglycerides, and diabetes are offered to participants (the results are sent back with a special notice in case of abnormal test results).

All baseline questionnaires were first pilot-tested and compared with traditional administration methods (paper-and-pencil versions or interviews by a dietitian) [30-32]. Health events are monitored through questionnaires about hospitalizations and medication use as well as via a linkage with the national vital statistics database. In April 2011, 15,000 randomly selected participants among 86,652 individuals included at the time of the study were invited to complete a Web-based questionnaire regarding their motives for participation in the study.

This study was conducted according to guidelines laid down in the Declaration of Helsinki, and all procedures were approved by the Institutional Review Board of the French Institute for Health and Medical Research (IRB Inserm no: 0000388FWA00005831) and the Commission Nationale Informatique et Libertés (CNIL no: 908450 and no: 909216). Written electronic informed consent to participate in the study was obtained from all participants.

The present analyses focused on data from a random sample of participants in the NutriNet-Santé cohort who had completed the questionnaire assessing their participation motives and who had no missing sociodemographic, lifestyle, anthropometric, or health status data. These characteristics were compared between participants included in our analysis and those who had stopped participating within 6 months after their enrollment (calculated from the date of the last connection on the website), using a chi-square goodness-of-fit test. The possible reasons for participation were grouped into the following 4 categories: (1) to help advance public health research on chronic disease prevention, (2) to help advance nutrition research, (3) in response to the call from the media, from a close friend/associate or from a medical provider, and (4) other motives (ie, to receive regular scientific information about health and nutrition, out of curiosity, to belong to a group, and other). Perceptions/attitudes toward the public funding of the study were categorized into 3 groups: very important, important, and not important.

According to French recommendations [33], male drinkers were categorized as abstainers and irregular consumers (<once a week), moderate (0-30 g alcohol/day), or heavy drinkers (>30 g alcohol/day), and females as nondrinkers (0 g alcohol/day), moderate (0-20 g alcohol/day), or heavy drinkers (>20 g alcohol/day). Normal weight, overweight, and obesity were defined according to the World Health Organization classification for BMI, as BMI <25 kg/m², 25≤ BMI <30 kg/m², and BMI ≥30 kg/m², respectively [34]. Gender, age (≤25, 26-35, 36-45, 46-55, and >55 years), marital status (married or living with a partner, single/divorced/widowed), having at least 1 child at home, education (elementary school, secondary, college graduate, and advanced degree), occupational category (managerial staff, farmers/self-employed, intermediate profession, employees/manual workers, and never-employed/homemaker), type of area of residence (rural, semiurban <20,000 inhabitants, urban 20,000-100,000 inhabitants, urban ≥100,000 inhabitants, Paris metropolitan area), smoking status (never smoker, former smoker, current smoker), alcohol consumption, BMI, self-reported type 2 diabetes, hypertension, and hypercholesterolemia are presented in a frequency/percent format for the whole sample. Multivariate associations between the individual characteristics and the motives for participation (participation motive related to the advantages of the Internet, reasons for participation, and attitudes about public funding) were assessed using multinomial logistic regression. Reference category used in the analysis of the associations between participation motive related to the advantages of the Internet and individual characteristics was yes. For the analysis regarding reasons for participation, the reference category was other motives, and for those concerning attitudes about public funding the reference was not important. In each multivariate model, the explanatory variables were adjusted for each other. Odds ratios (OR) and 95% confidence intervals (CI) are reported. Significance tests were 2-sided with a P value set at <.05. A more conservative P value of.01 was also used for estimating the robustness of the results. Statistical analyses were performed using SAS software version 9.3 (SAS Institute Inc, Cary, NC, USA).

Profiles of participants in a Web-based epidemiological cohort have rarely been studied [8,10-16] and motives to participate in such cohorts have not yet been explored. The present study highlighted specific sociodemographic and health status (ie, weight status) characteristics of participants in a large Web cohort according to the perceived importance of the Internet, the reasons for participation, and the importance of public funding for research. Our results could guide the development of novel strategies for the retention of diverse population samples in Web-based cohorts, particularly during key periods of data collection.

Our findings revealed that almost half of the participants reported that Internet use was a decisive reason for participation. In fact, this element exerted a stronger influence among men, persons younger than 35 years, individuals of higher socioeconomic status, those without children at home, moderate alcohol consumers, and obese persons. Our results are concordant with previous studies that compared sociodemographic and lifestyle characteristics of participants who used Web-based questionnaires with those of participants who used paper-and-pencil instruments [8,14]. There is clear evidence that men, young people, and single persons are less likely to participate in epidemiological studies than are women, older, or married individuals [9,35-37]. Next, several studies have shown that individuals who presented risk behaviors, such as smoking, heavy alcohol use, or obesity, were often underrepresented among research participants [9,38,39]. Thus, our study suggests that the Internet allows for reaching a large population, but also provides access to hard-to-reach individuals given their social or behavioral status and for whom the Internet seems to be a more attractive and more convenient mean for participation compared to traditional means. Individuals belonging to low socioeconomic strata are often poorly represented when relying on traditional methods [40]. Our findings showed that the Internet appeared to be a less important motive for participation among individuals in low occupational categories, compared to managerial staff. Therefore, further exploration of measures that can be adopted in epidemiological Web-based studies to specifically increase opportunities for participation among low socioeconomic groups is necessary.

In addition, our study indicated that participation in an exclusively Web-based nutrition cohort was driven mainly by a desire to help advance research on chronic disease prevention or nutrition, especially among older participants, those with normal weight, and those who lived with a partner. Our results are also consistent with existing knowledge regarding motives for participation in volunteer-based cohorts on health and nutrition, which do not use the Internet [24,26-28]. These reasons for participation may reflect altruistic tendencies, but also a vested personal interest [27]. As explained by Williams et al [28], individuals may be more willing to participate if they believe that the potential benefits of their participation are large (eg, life versus death), highly likely to materialize, quickly attained, or likely to benefit themselves or someone important to them. Our results suggest that the desire to contribute to chronic disease risk prevention should be used as a key lever for participants’ retention in Web-based cohorts, particularly during decisive periods of exposure data collection.

Two-thirds of our sample found the use of Internet for completion of the questionnaires to be a benefit, given its flexibility, whereas less than 1% found it to be complex. Also, one-quarter of participants felt more comfortable filling in the questionnaires on the website rather than face to face with an investigator. On the other hand, only 22% visited sections of the website of the study regarding news and progress of the study. Thus, in Web-based studies, the reduced participant burden (eg, quick, easy and convenient access, increased distance between participant and investigator allowing participants not to feel judged) [41] should help minimize attrition.

Participation in the study for altruistic reasons may be reinforced by the public nature of the research. Indeed, two-thirds of the participants considered the exclusive public funding as important or very important, with the link being particularly pronounced in men, older adults, and individuals of higher socioeconomic status. This finding is not surprising in a European context in which the majority of cohort studies are funded by public organizations. Indeed, 94% of French participants in an opinion survey conducted in the general French population reported that a large part of biomedical research needs to be funded by public funding, and 80% of responders feared that the increased participation of private funding in public research could undermine the independence of research and is damaging to certain research areas deemed less profitable [42]. In turn, the use of Internet for epidemiological studies could be viewed with suspicion by some participants because of fear that personal information might be diffused or sold to third parties or that their responses might not be anonymous [43]. In addition to reassuring participants that their personal information is kept private and safe, it seems important to point out the fact that the study is exclusively funded by public sources to investigate important public health issues.

Our study has several limitations. First, responders were older and belonged to higher socioeconomic strata than nonrespondents, which might have led to an underestimation of the observed associations. Moreover, our result suggests that the influence of weight status on participation is open to question because participants in a long-term cohort are likely to be particularly health conscious and interested in nutritional issues. In addition, results may reflect the motives of participants accustomed to the study rather than their motives for enrollment in the Web-based cohort because the questions about motives were asked approximately 2 years after baseline. However, key information on exposure and potential confounding factors was collected during the first 2 years of participation in the cohort. Thus, a focus on the motives of those participants who actively participated 2 years after their inclusion is useful in terms of retention strategies during decisive periods of data collection in Web-based cohorts. In addition, the percentage distributions of the given reasons for participation could be biased due the use of a predefined list of response options. However, the participants had the opportunity to choose the “other” response option and to specify the exact reason for participation. Another limitation was the lack of information on reasons for declining participation because the call for participation was not delivered to a predefined list of individuals. Finally, anthropometric status was assessed by self-reporting and may have led to misclassification. However, in a separate validation study that used data from a subsample of 2513 participants, we compared self-reported and measured height and weight (and the resulting BMI) [44]. In particular, these participants had completed the anthropometric questionnaire 3 days before the clinical examination. The intraclass correlation coefficient ranged from .94 for height to .99 for weight. BMI classification was correct in 93% of the cases; the weighted kappa statistic was .89. Of 2513 participants, 23.5% were classified as overweight (BMI ≥25) with Web-based self-report versus 25.7% based on measurements by study staff, leading to a sensitivity of 88% and a specificity of 99%. For obesity, 9.1% versus 10.7% were classified as obese (BMI ≥30), respectively, leading to sensitivity of 83% and specificity of 100%.

Our study highlighted that the reliance on the Internet, the willingness to help advance public health research, and the exclusive public funding of the study were key motives for participation in this exclusively Web-based cohort. In addition, these motives for participation differed by sociodemographic profile and BMI, but not by lifestyle or health status. These findings can help improve retention of diverse population samples, particularly during important data collection periods.