The present study was designed to test the 12-and 24-month effectiveness of the NHF-NRG In Balance-project, with regard to changes in body weight, BMI, sum of skinfolds and waist circumference. The results indicate that with regard to changes in sum of skinfolds and waist circumference the project was indeed effective at both 12 and 24 months. Even though changes in weight and BMI between the intervention and control group were not significantly different, they did change in the desired direction. Overall, the intervention of the NHF-NRG In Balance-project had a positive effect on the body composition measures of the individuals in the intervention group. The interpretation of effect sizes of Cohen's d imply effects of medium magnitude for the changes in skinfold thickness and waist circumference both after 12 and 24 months (Cohen's d between 0.33 and 0.55). Such changes in body composition indicators may have important health implications, as it has been demonstrated that the health risks associated with obesity derive primarily from fat rather than weight . Moreover, it is not only the total amount of fat that is important, but also the distribution of fat in the body , with central fatness being most related to health risks . The reduction in skinfold thickness and waist circumference observed in the present study reflects a reduction in central fatness [22, 25]. The decrease in waist circumference is most relevant, as a large waist circumference is independently associated with health risks [26, 27] and mortality [28, 29]. On a population level it has even been shown that there is a more significant trend of increases in waist circumference over time than BMI .
The observed changes in anthropometric measures could be a result of changes in participants' food intake and/or physical activity behaviour. With regard to changes in waist circumference it has been demonstrated that an increase in fibre intake was associated with a reduction in waist circumference in men . A strong dose-response relationship has also been observed between the amount of exercise and measures of central obesity . Interestingly, changes in physical activity can lead to changes in body composition, which may be reflected in changes in waist circumference, while body weight remains stable through increased muscle mass [33, 34]. This is in line with the findings of the present study.
Stratified outcome analyses were interesting. It appeared that the intervention only had an effect on the changes in skinfold thickness in women and not in men. It would be interesting to see if this is a result of the engagement in different energy balance-related behaviours of men and women.
The process evaluation of the environmental interventions showed that two worksites formed a worksite linkage-board, who implemented several environmental interventions throughout the two year period. When taking baseline characterises into consideration, the individuals in these two worksites appeared to show better results with regard to changes in waist circumference and sum of skinfolds than individuals in worksites with fewer components to the intervention both after 12 and 24 months (data not shown). Although the study was not powered to significantly detect these between-worksite differences, this finding does underscore the importance of intervening on both the individual and the environmental level. Moreover, it showed that the context of the worksites did not affect the uptake of the intervention, as one of these two worksites had predominantly white-collar workers and the other blue-collar. This finding as worksite-health promotion programs are often less likely to result in health behaviour change in blue-collar workers .
The NHF-NRG In Balance-project is one of few worksite obesity prevention programmes, which 1) is primarily aimed at weight gain prevention through changes in both food intake and physical activity, 2) contains both individual and environmental components and 3) assesses longer-term follow-up effectiveness. A recent review of papers on lifestyle interventions aimed at prevention of overweight and obesity, with primary programme objective weight management, prevention of weight gain or moderate weight loss among adults, included four additional studies to the present study, in which workplace interventions were evaluated. Two of these studies included behavioural goals that were aimed at both diet and physical activity; three included both cognitive and environmental goals and two studies assessed effectiveness after a 12 month follow-up. Significantly smaller increases in BMI in the intervention conditions were observed in one study; no treatment effect for weight or BMI changes was found in the others. Two of the studies also included measurements on percent body fat, both of which observed significantly positive effects . These findings are in line with those observed in the present study. To date, there has been an increase in the number of worksite obesity prevention studies that are testing environmental or combined environmental-and individual-level worksite interventions over a longer period of time, e.g., through the National Heart, Lung, and Blood Institute . However results regarding effectiveness have not yet been published.
In the present study, we perceived several benefits of implementing the intervention within a worksite setting. Firstly, the worksites provided access to a large number of adults with different educational backgrounds. Moreover, the employees within the worksites are able to play an important role in diffusing the intervention throughout the worksite by impacting social norms, which in the long-term may influence the behaviours of co-workers who did not change their behaviour initially . Difficulties were perceived with regard to enhancing facilitators of environmental changes, as only two of the six worksites set up a worksite-linkage board. As the linkage boards play a crucial role in the adoption, implementation and institutionalization of the environmental components, strategies should be developed to mobilize support and commitment for the formation of such boards.
There are a number of limitations of this study, including those concerning the generalizability. The first is related to the recruitment of companies, as only 9% of the approached companies were willing to participate. An important reason for companies not to participate in the NHF-NRG In Balance-project proved to be the randomized evaluation design of the programme, implying that companies were not willing to take the risk of being excluded from the intervention . We were therefore forced to drop the original randomization design of the programme and assign worksites to the experimental and control group based on matching. As a result of which it is possible that selection bias occurred, weakening the internal validity of the results. Moreover, external validity was weakened by the fact that participating worksites were most likely not representative of the average worksite, in that the participating worksites probably showed a higher interest in health promotion than worksites in general. Implementing the project in less interested worksites might not have generated the same results. A second limitation of the present study is the recruitment of participants. Even though the aim of the project was to prevent weight gain in young adults, there was a relatively high response of older and overweight individuals, in line with observations of other studies [27, 28]. This may have resulted in a selection bias, in which individuals who were more interested to change the targeted behaviours were oversampled. Moreover, there was a high response of participants with a tertiary education. The third limitation concerns the statistical analysis, although sophisticated multilevel analyses were executed in this study, the statistical procedures may not fully account for all potential dependencies that were introduced as a result of the research design. For example, our statistical model contained only one random component for worksite, implying that every worksite is assumed to have exactly the same response to the intervention (if in intervention) or to the control situation (if in the control condition). The fourth limitation pertains to the process evaluation; unfortunately we were unable to perform an in-depth analysis regarding the uptake of interventions by the individuals. The fifth limitation is related to the absence of a significant difference in weight changes over time between both groups. Even though the drop-out rate after two years was above 20% (27.5%), the required sample size to detect a medium effect (n = 372) was still met (n = 401). However, weight changes observed in the control group were smaller than those expected, with smaller weight change differences between the groups (0.5 kg at 12 months). The smaller increase in weight in the control group is most likely a result of measurement effects. However, it could also be a result of a selection bias; the control group might have consisted of more motivated individuals who are susceptible to change. Moreover, it is possible that those individuals who dropped-out were those with a higher BMI.