The intervention study was nested in the Danish Health Examination Survey 2007-2008 (DANHES) [24], a nationwide health study in Denmark. DANHES was carried out in 13 municipalities in 2007-2008 and comprised a Web-based questionnaire and a health examination test. All adults (18+ years of age) in 13 municipalities (N = 538,497) were invited by letter to take part in a Web-based questionnaire. The questionnaire assessed lifestyle-related health aspects with a focus on diet, smoking, alcohol use, and exercise. A representative sample in each of the 13 municipalities were also invited to a health examination test (n = 180,103). For details see Eriksen et al [24].

The DANHES was used to recruit participants and as a baseline assessment. The intervention study was conducted in 11 of the 13 municipalities participating in DANHES during May 2008–May 2009. We excluded 2 municipalities from the intervention study, as one served as a pilot study for DANHES and paper questionnaires were used in the other.

The main inclusion criterion for participating in the intervention study was being physically inactive during leisure time. This was defined by the participants’ answer to a 4-category question describing PA level in leisure time. We included participants in the lowest 2 categories, mostly sedentary or light activities, in the study and excluded participants in the highest categories, moderate and vigorous PA. Further exclusion criteria were presence of serious heart problems, not being able to perform everyday activities, or missing values in the International Physical Activity Questionnaire (IPAQ) and the leisure-time PA question.

We identified participants who met the inclusion criteria by a screening program and invited them to join the intervention study at the end of the questionnaire in DANHES. If willing to participate, each participant was randomly assigned by the registration program to either an intervention (website) or a no-intervention control group. The only incentive given to participants was the possibility of being assigned to the intervention group. Blinding was not feasible.

The participants in the website group received an email with a link to a PA website immediately after allocation to the website group. In addition, relevant data from the health survey were automatically transferred to the intervention website. To access the website, the participants were required to log on to the website, using the same personal username and password given in the health survey.

All participants gave informed consent before being enrolled in the study. The study was approved by the Danish National Committee on Biomedical Research Ethics (H-D-2008-035).

The intervention website was founded on the theories of stages of change [25] and of planned behavior [26]. First, we identified determinants in the theory: intentions, attitudes, self-efficacy, skills, social support, and knowledge. Second, we specified key objectives for each determinant and decided where in the intervention the objectives would be implemented (Table 1). The key objectives were used in the wider planning of the intervention to specify how objectives could be translated into action in the real-life intervention.

The website was structured as three major parts: (1) a personal page, which included individually tailored PA advice and a personal profile, (2) a page with training programs and general recommendations, and (3) a forum and discussion page for questions from participants.

The individually tailored PA advice consisted of three parts: (1) a general introduction, (2) normative feedback, which related the participant’s PA to the current PA recommendations and (3) general advice about using the tools on the website. The normative feedback was based on the summarized PA time from the participant’s answers in the IPAQ. Feedback was given in the domains of everyday activity, fitness training, and strength training. In each domain we defined categories in which the participants received tailored feedback according to their level of PA. The categories were partly based on PA recommendations from the Danish National Board of Health translated into minutes per week [27], together with analyses from answers in the IPAQ from 2 municipalities. The analyses showed that the categories needed to be wide because of overreporting in the IPAQ. The categories were everyday activity, being the total activity time from questions in the IPAQ summarized into low (<1200 min/week), moderate (1200-3500 min/week), and high levels (>3500 min/week); fitness training, being summarized time from vigorous and moderate-intensity activity from the transport and the leisure-time domain and moderate-intensity activity from the domestic domain in the IPAQ, at low (<40 min/week), moderate (40-350 min/week), and high (>350 min/week) levels; and strength training, being summarized time from the highest intensity level in the 4 domains in the IPAQ, at low (<100 min/week) and moderate to high (>100 min/week) levels. Participants who were 60 years of age or older were given extra advice regarding the importance of strength training. To monitor their progress during the study, the participants could register personal data, such as waist circumference and the result of a short fitness and strength test in the personal profile. We used this kind of biofeedback to keep the participants motivated during the intervention. Furthermore, current activity per day could be calculated with a short activity calculator, and it was possible to set goals for the following 4 weeks. We decided on 4 weeks for goal setting as a balance between time for seeing results and maintaining motivation. All participants were encouraged to make a personal profile to set their goals, monitor their progress, and implement their goals.

The part of the website that included the page with the training programs and general recommendations was structured in the same way as the PA advice in the 3 domains: everyday activity, fitness training, and strength training. The participants were encouraged to go through the different suggestions and programs and pick the ones that suited them best based on the individually tailored advice and goals set by each participant. General information about motivation and relevant links were also presented on the website.

On the forum and discussion page, a physiotherapist experienced in PA counseling answered all questions about PA and training from participants. In addition, participants could share experiences and give each other tips or search for training partners in a second forum.

We kept the tailored PA advice short so as not to overload the participant with information and, apart from training programs and general recommendations, as a means of following the theory of stages of change [25]. The theory describes how change is a process of progressing through a series of stages: precontemplation, contemplation, preparation, action, maintenance, and termination. Hence, participants ready for change and already in the action stage could make a personal profile, set goals, and find training programs. Participants in the precontemplation to preparation stage who did not log on to the website or did not make a personal profile were sent two email reminders to encourage them to become involved with the intervention. Participants in the maintenance stage who made a personal profile were sent reminders and encouraging emails to keep the profile updated after 4, 8, 12, and 16 weeks.

The content of the website was developed by the research team. Two professional Web companies did the graphic design and implementation. The intervention was pretested among experts and representatives of the target population. We tested screening of participants, invitation to the intervention, automatic generation of individually tailored advice, email generation, and general usability of the website. Furthermore, we used the first municipality participating in the study (n = 1298) for a pilot study. Comments and suggestions from these participants were used to fine-tune the website. Multimedia Appendix 1 shows recruitment of participants and the intervention website.

The baseline health examination included measurements of blood pressure, height, weight, body fat percentage, and grip strength. Aerobic fitness was estimated from a watt-max [28] or a sub-max test [29] on an ergometer bike. Body mass index (BMI) was calculated from measurements of height and body weight. For further details about methods and test procedures, see Eriksen et al [24].

From 3 selected municipalities in the intervention study, we invited 1190 participants to a follow-up health examination after 12 weeks. The participants were invited by email, which was sent 3 to 4 weeks prior to the examination. If the participant did not respond, a reminder email was sent 1 week after the first. The follow-up examination included the same measurements as the baseline health examination, and the same test procedures were followed. The follow-up examination was blinded to the examiners. After the health examination, the participants received their results to track developments from baseline to follow-up. We expected to see improvements in measurements included in the health examination, as participants in the intervention group were encouraged to engage in moderate- and high-intensity PA on a weekly basis. Improvements have been seen after 3 months in other studies depending on diet [30-32].

The baseline questionnaire included demographics, health, lifestyle, and health behavior. We used the long version of the IPAQ, which is known to be a valid and reliable instrument for assessing PA [33], both at baseline and at follow-up. It consists of 31 items that collect information on PA in the 4 domains work, transport, housework and gardening, and leisure time. Motivation for changing PA behavior was assessed by the question “Do you wish to be more physically active than you already are?” (yes, yes/maybe, or no).

After 3 months and after 6 months, we invited all participants by email to answer a follow-up questionnaire. The follow-up questionnaire included questions about use of the website for the website group. Due to a technical error, only half the participants were invited to answer the 3-month follow-up questionnaire.

The primary analysis in this intervention was overall level of PA based on self-reported PA from the IPAQ. The secondary outcome measures were blood pressure, height, weight, body fat percentage, and grip strength. The outcome measures were specified a priori.

As a post hoc outcome measure, a secondary analysis was carried out among active users of the intervention. Here we divided the participants of the website group into three groups according to user activity (no log-on, log-on once, and log-on more than once) and assessed level of PA. Furthermore, we calculated the odds ratios of being an active user of the website.

We assumed that a reasonable effect of the intervention on total PA time estimated by the IPAQ would be around 12% for the intervention group and 5% for the control group. With a power of 80% probability of detecting a 12% versus 5% difference as statistically significant at the 5% level, we calculated the minimum sample size to be 250 in each group. We expected that approximately 50% of the participants in DANHES were sedentary. Assuming that 80% accepted participation and 25% were lost to follow-up, this would still give us a large population and hence ensure sufficient power.

We analyzed the IPAQ results according to the Guidelines for Data Processing and Analysis of the International Physical Activity Questionnaire [34] with the exception that we included participants with a missing value in day or time in the follow-up analysis. Nonparametric analyses were used, since IPAQ data were not distributed normally (Wilcoxon rank sum test between two groups and Kruskal-Wallis test between more than two groups).

Results were primarily analyzed as intention-to-treat analyses with the use of the last observation carried forward to account for missing data at follow-up. We analyzed completer data including only participants who completed the follow-up health examination or questionnaire.

Website use was assessed by the follow-up questionnaire and combined with information provided by the company that was responsible for the website, which recorded whether a participant logged on.

Odds ratios of being an active user (several log-ons versus one or none) were calculated in relation to sex, highest education level (<10, 10–12, 13–14, or 15+ years), age group (18–44, 45–64, or 65+ years), and motivation to be more active (yes, yes/maybe, no) by the use of logistic regression.

For all statistical calculations and analyses, we used Stata version 11.2 (StataCorp LP, College Station, TX, USA). We performed chi-square tests to examine differences in proportions between the groups. We considered P< .05 as statistically significant.

In total, 53,956 persons participated in DANHES in the 11 municipalities. Of these, 28,054 participants met the inclusion criterion and were invited to participate in the intervention study (Figure 1). Among these, reasons for nonparticipation were refusal (8593) and not answering the invitation (6242). A technical error gave some participants in the control group access to the website and resulted in exclusion of 895 participants. A total of 12,287 participants were enrolled in the study, resulting in a 43.80% participation rate. The response rates in the 3-month questionnaire were 57.55% (2375/4127) in the intervention group and 66.41% (2175/3257) in the website group.

In the health examination, 32 participants were excluded due to participation in another health intervention, leaving 583 in the website group and 585 in the control group. In total, 434 (37.2%) participated in the follow-up health examination, with 215 in the intervention groupand 219 in the control group. Participants who were lost to follow-up in the health examination and questionnaire were not significantly different from those who completed follow-up in respect to baseline characteristics (data not shown).

Baseline characteristics did not differ significantly between the website and control groups as shown in Table 2. Mean age of the participants was 50 (SD 13.6) years and 64.82% (3925/6055) were women; overall, 18.95% (2329/12,287) were mostly sedentary and the rest reported light PA in their leisure time.

No significant differences were found in baseline characteristics between the website and control groups in the health examination (Table 3). The mean BMI among participants was 25 (SD 3.8) kg/m²and mean body fat percentage was 30.5% (SD 8.1). Aerobic fitness was somewhat low at 32 (SD 7.8) mL/min/kg.

Participants in the website group did not report a significantly different level of PA compared with the control group at the 6-month follow-up (Table 4).

The result was the same in the completer analysis and analysis at 3-month follow-up (data not shown). Analyzing participants who stated that they wanted to change their PA separately gave the same result (data not shown). In relation to other health and baseline characteristics, no significant differences were found at follow-up at 3 and 6 months (data not shown). Furthermore, no significant changes were found in either the website or the control group from baseline to follow-up (data not shown).

The results from the follow-up health examination showed no significant differences between the website and control groups (Table 5). From baseline to follow-up, we found no significant changes in either the website or the control group (data not shown).

When we divided the participants into three groups according to use of the intervention website (no log-on, log-on once, and log-on more than once), we found a significant difference between the groups in leisure-time PA and total PA (Figure 2). Dividing participants from the website group who participated in the follow-up health examination into the same three subgroups of website use did not show significant differences between the groups (data not shown).

We found that 22.0% (694/3156) of the website group logged on to the website once and only 7.0% (222/3159) logged on more than once (Table 6). The question about website use was answered by 3245 participants, of whom 73 (2%) reported that they had logged on to the website, even though reports from the website company responsible for the website showed that they did not log on to the website. We excluded these participants in the analyses of website users. According to 70.0% (710/1014) of the respondents in the website group, forgetfulness was the main reason for not logging on to the website (data not shown).

In the group of participants who did log on to the website, 31.4% (318/1014) believed that the intervention helped them to increase PA. When we analyzed this group separately, we did not find significant changes in PA from baseline to follow-up (data not shown). The website’s PA adviser received few questions and the participant forum was not used at all.

Analysis of the active users of the intervention in relation to age group, educational level, motivation, and sex showed that participants in the age group 45–64 years (odds ratio 1.6, 95% confidence interval 1.2–2.1) and 65+ years (odds ratio 2.0, 95% confidence interval 1.4–2.8) were more likely to log on more than once than were those in the age group 18–44 years. Motivated participants (odds ratio 1.5, 95% confidence interval 1.0–2.1) were more likely than nonmotivated participants to log on more than once.

Web-based research is still in an early stage, with many lessons to be learned. The question is how we can move forward and develop interventions that can change PA behavior. The finding in this study suggests that active users of a Web-based intervention can achieve a positive effect. However, for unmotivated users, single-tailored feedback may be too brief. Future research should focus on the step from intention to action and on developing more sophisticated interventions. This is seen in Web-based smoking cession interventions, which combine different types of media and have many contacts, and thereby have the potential to reach both motivated and unmotivated sedentary individuals.