This RCT is part of the PAM project, which is described extensively elsewhere [22]. Mainly office workers from 20 to 40 years old, all apparently healthy, with differing levels of education were recruited from eight worksites in the surrounding areas of Amsterdam, the Netherlands. For all worksites, the same recruitment protocol was used. In the recruitment procedure, we informed the participants about the beneficial health outcomes of regular PA (ie, increased cardiovascular health, and reduction of the risk of overweight, type 2 diabetes, depression, and some types of cancer) by providing a brochure and through individual communication. Inclusion criteria were ability to walk without aid, Dutch speaking, and not being pregnant. First, PA levels were monitored for 2 weeks by means of a PAM and a PA questionnaire. Based on these 2 weeks, the study population (N = 302) was divided into “active” (most active 50% of the population) and “inactive” (least active 50% of the population) groups. The inactive adults were invited to participate in the RCT and were told that they were eligible because of their low level of PA. To be able to detect a between-group difference of 20% in PA level (80% probability and a significance level of .05), two groups of 50 participants were required. Of the 152 invited adults, 102 (67% response rate) completed the baseline measurement and were randomly assigned to either the intervention (n = 51) or the control group (n = 51). Randomization to the intervention or control group was performed at the individual level by choosing sealed envelopes after the baseline measurements. The flow of subjects through the RCT and the distribution of nonresponders are shown in Figure 1. This study was approved by the Medical Ethics Committee of VU University Medical Center and was conducted between September 2004 and November 2005. All participants gave their informed consent.

Figure 1

Flowchart of the intervention (I) and control (C) subjects in the RCT. Note that for analysis of the primary outcome measure at t1, 2 out of 48 (4%) and 1 out of 50 (2%) participants in the intervention and control group respectively, were excluded for analysis due to impossible values for physical activity (not shown in figure, see text under "Primary Outcome Measure").

After randomization, participants in both the intervention and control groups were advised to increase their PA levels. The control group received a single written information brochure with brief general PA recommendations. This print brochure is published by the Netherlands Heart Foundation and contains brief general information on the health benefits of PA and the PA recommendation. Everyone can obtain the brochure free of charge.

The intervention group received the PAM and was provided with Web-based tailored PA advice (PAM COACH) [23] for a 3-month period. Figure 2 and Table 1 show the sitemap of the PAM COACH website.

Figure 2

Functionalities of the PAM COACH website

All measurements took place at the worksite during working hours, at baseline, immediately after the 3-month intervention, and at an 8-month follow-up, which corresponds to 5 months after the end of the 3-months intervention period. The measurements are described in detail elsewhere [22]. Gender, age, and education level were obtained at baseline. Education level was categorized into low and high. High education level comprised higher vocational education or a university degree. All other levels of education were defined as low.

To compare baseline values, the chi-square test was used for gender, education, and awareness distributions. Nonparametric testing (Mann-Whitney U test) was used for PA data. Independent samples t-test was used to analyze all other demographic variables, determinants of PA, aerobic fitness, and body composition. The effect of the intervention was estimated based on the intention-to-treat principle, including all participants who had attended at least one follow-up measurement. Logistic regression analysis was used for the dichotomous outcome measure, awareness of meeting the PA, and sports recommendation (0 = nonrealists, 1 = realists). For all other outcome measures, standard linear regression analysis was used to test the differences between intervention and control groups at follow-up. The follow-up measurements were defined as dependent variables. Baseline values of the particular dependent variable were always included as covariates. The parameters of interest are the regression coefficients (β), indicating the effect of the intervention of interest compared to the control group. In a second analysis, gender (0 = male, 1 = female), age (continuous), education (0 = low, 1 = high), and BMI (continuous; not added in the analyses for body composition outcome measures) were considered as possible confounders or effect modifiers. If the interaction between group allocation and the variable concerned had a p-value below .10, then subgroup analyses were performed. Additionally, we adjusted for program adherence on the outcomes by performing regression analyses among intervention participants, including the log-in frequency. Analyses were performed using SPSS version 14.0 (SPSS Inc, Chicago, IL, USA).

Table 2 shows baseline characteristics of both the intervention and control groups; 61 out of 102 participants were female (60%), and 66 out of 102 (65%) had a high level of education. Participants in the intervention and control groups were comparable except for age (intervention group was, on average, 1.3 years older; P = .05) and intent to participate in sports at baseline (higher in control group; P = .01). The majority of participants (68 out of 102, 67%) met the PA recommendation at baseline, with 78% (n = 32) of the men and 59% (n = 36) of the women meeting the recommendation.

In the total study sample, our 3-month intervention did not significantly affect PA levels (Table 3). However, because of effect modification by education, we conducted subgroup analyses. These analyses showed that the 3-month intervention resulted in a relative lowering of light-intensity PA (2-4 METs) among higher-educated participants (adjusted difference between intervention and control group in min/week, β = −349, 95% CI = −632 to −66, P = .02). This effect was not sustained at the 8-month follow-up, 5 months after the intervention. A higher adherence to the program did not result in increased levels of PA (data not shown).

Of the PAM users, 35 out of 48 (73%) reported to have worn the PAM “regularly” to “often” (Table 5). This finding was supported by the log-in frequency (almost once a week) of the PAM data to the PAM COACH website. Just over half of the PAM users (24 out of 46, 52%) set a personal goal, and 33 users (72%) entered their favorite activities on the website. Main reasons for not using these items were lack of interest and not being able to find them on the website. The tailored advice was read by 34 out of 46 (74%) PAM users, of whom 14 did not find the advice appealing. Main reasons were as follows: the advice was not personal or specific enough (n = 9), the advice was not applicable to their daily situation (n = 6), little variety in the advice (n = 3). Overall, the participants rated the PAM and the PAM COACH website as sufficient.

The results of this study must be interpreted in light of its limitations. First, our primary outcome and some of our secondary outcomes were based on self-report and therefore prone to misreporting. However, since we looked at changes in PA behavior, at least the bias associated with systematic errors is cancelled out. Nevertheless, we compared the self-reported total PA data with the objective uploaded PAM data among participants of the intervention group (data not shown). The PAM data confirm the decline in total PA as assessed by the self-report (change in median min/week: −147) after the 3-month intervention.

Second, our control group is not a truly non-intervention group because they received an information leaflet on PA. However, we do not expect changes in PA by providing such brochures only. Furthermore, our findings may reflect ceiling effects associated with a relatively active sample at baseline. Although the study was aimed at inactive employees, 35 out of 51 (69%) participants in the PAM intervention group and 33 out of 51 (65%) participants in the control group already met the PA recommendation at baseline. Even though this information was based on self-report, it seems likely that the participants selected for the RCT were in general more active and health conscious than the general Dutch population. This is supported by the facts that the percentage of subjects who were acquainted with the PA recommendation was high in both groups, and the percentage of subjects who were aware (realists) of their compliance with the PA recommendation and their regular participation in sports was high in both groups. The fact that we did not collect information about the participants’ willingness to become physically active can be considered as a limitation. In addition, our study results are mainly applicable to people who are employed at a workplace that allows personal Internet use, which limits the generalizability of our study.

Finally, the practical advice given on the website was partly based on the objectively monitored PAM score. Accelerometers are insensitive to certain types of movements, in particular, nonambulatory physical activities with arm and or limb movements, such as cycling and weightlifting. This limitation of the accelerometer may have reduced the accuracy and relevance of the advice given by the PAM COACH website, particularly for subjects who cycle a lot, which is common in the Netherlands. Although activities that are not accurately measured by the PAM can be included manually on the PAM COACH website, a study has shown that recipients of negative or unexpected feedback responded by doubting the accuracy and credibility of the feedback information [34]. This phenomenon may have discouraged our participants from achieving their personal PA goal.

Strengths of the study are its design, the easy to implement intervention, and the low dropout rate. This RCT was set up as a short-term minimal intervention strategy in order to make it easily applicable in real-life settings. During the intervention, PAM users received short personalized PA advice together with supportive practical advice to reach their personal PA goal. In order to reach this PA goal, the activity preferences of the user were taken into account so that PA could be more easily implemented into their daily life. After registering, the user could decide when and how often to log in to the PAM COACH website. In spite of the minimal contact during the intervention between the researcher and the participant, adherence was moderate to high. PAM users logged in 10 times on average during the 3-month RCT, which is almost once a week and is comparable with frequencies of website log-ins from previous studies (range 0.7 to 1.5 times per week) [35,36]. Moreover, during the intervention we observed a low dropout, only 3 (6%) and 1 (2%) out of the 51 participants for the intervention and control group, respectively. The recurrent visits and low dropout during the intervention suggest that participants were interested in new information and were acquainted with the technology.

The appreciation of the intervention materials differed largely among participants in the intervention group; most participants expected more varied and concrete advice and found the advice not applicable to their daily life. This occurred in spite of our aim to tailor the PA advice to a certain extent based on the users’ actual PAM score in relation to their PAM goal. We strived for simple and concise PA advice and a variety of motivational tips on the PAM COACH website.

To conclude, we hypothesized that the combination of wearing a PAM combined with tailored PA advice delivered through the Internet would be potentially successful in increasing awareness of personal activity levels and actual PA levels. However, we did not observe any significant effect on awareness, PA level, determinants of PA, aerobic fitness, or body composition among the total group of young healthy Dutch employees. This may be explained by the fact that we conducted a minimal intervention in a study population that largely (67%) met the PA recommendations at baseline. Moreover, a large part of the intervention population did not find the advice appealing. Hence, the results of the present study do not give cause for wider implementation of this minimal intervention among healthy adults. Since we observed a tendency for a positive intervention effect on body weight among low-educated adults, more research may be necessary to investigate the effectiveness of this type of intervention among people who are overweight or of low socioeconomic status. In this, attention should be given to the quality and appropriateness of the tailored advice.