The purpose of this study was to analyze the effects of a worksite environmental intervention on determinants of dietary behavior regarding eating more fruit and vegetables and eating less fat and on actual (self-reported) fat, fruit and vegetable intake.
The results of this controlled trial showed that this environmental intervention only had a modest effect on determinants of dietary behavior. A significant effect was found on the perceived social support from colleagues regarding eating less fat. This determinant significantly increased at the short-term and borderline significant at long-term. However, also counterintuitive effects were found. First, at 12 months the attitude toward eating less fat decreased in the intervention group and decreased even more for subjects with a higher BMI at baseline. Second, self-efficacy towards eating less fat at work decreased significantly in the intervention group. The intervention was ineffective in significantly increasing fruit, vegetable intake and decreasing fat intake of the intervention group. An interesting finding was, however, that in the intervention group at short term the subgroup of workers who did not take their lunch to work every day significantly increased their fat intake compared to those in the control group.
Just as in our study, in a controlled trial of Steenhuis et al  a similar lack of results on self-reported fat, fruit and vegetable intake was found. In that trial, the effectiveness of two environmental programs in worksite cafeterias of seventeen worksites was evaluated. In the first environmental program a larger variety of low fat products, and fruit and vegetable were offered in the canteen. In the second program low fat products were labeled. In contrast to our environmental intervention, both programs were combined with an educational program and were compared with just an educational program alone and a control condition. No intervention effects of the combined intervention programs were found on self-reported fruit, fat and vegetable intake. In addition, in the Steenhuis study, no effects were found also on determinants of behavior regarding eating less fat, and more fruit and vegetables. In contrast, our intervention was effective in significantly increasing social support regarding eating less fat. However, in our study as a result of the intervention the attitude and self-efficacy scores became more negative. This can be interpreted as a re-evaluation of their food habits by the subjects in the intervention group as a result of the food information provided in the company canteen. Because of this intervention the subjects might have perceived it as more difficult to eat less fat (at work), in contrast to previous beliefs.
Other worksite health promotion programs (WHPP's) did show positive results on self-reported fruit-vegetable and fat intake. These trials [24–29] were included in our review on the effectiveness of WHPP's with environmental components . It concerned trials that combined education, counseling or other individual strategies, with environmental changes. These environmental changes mostly consisted of extending the availability of healthy products and food labeling. Besides the fact that these trials applied combined interventions, another major difference with our study was that in these studies a more heterogeneous (blue and white collar) population was approached.
This difference in study population is an important point that might explain our poor results. In our study a primarily white-collar and highly educated population participated. White-collar populations are known to have in general more favorable food patterns (i.e. they eat more fruit-vegetables and less fat)  Therefore, a possible ceiling effect might have prevented the fruit and vegetable intake to increase, which might explain the slight decrease in mean vegetable intake observed at both worksites. When comparing vegetable intake in our population at baseline (i.e. 150 to 165 grams per day) to the general Dutch vegetable consumption recommendation (i.e. at least 150–200 grams of vegetables per day), it can be concluded that the baseline values were already relatively adequate, leaving little room for improvement. This seems a valid argument, when comparing these baseline values to the mean vegetable intake in the Dutch population, which was 134 grams per day in 1997 . Baseline median fruit intake values in our study were also relatively high, with 1.8 to 2 pieces of fruit per day for the men and women, respectively. These figures correspond with the Dutch fruit intake recommendation (i.e. two pieces of fruit per day). Another contributor to possible ceiling effects in our study was the fact that a year before the intervention began; the canteen management had already changed their policy towards a healthier diet in the company canteen. For example, some 'bad' snacks were sold on only one day of the week and all 'bad' snacks were made more expensive. In contrast, fruit and vegetables were subsidized. This policy change at the intervention company should be regarded as a 'natural' environmental co-intervention.
Another explanation for the lack of positive results could be that in our study the same questionnaires as in the study of Steenhuis et al  were used. However, these questionnaires were not specifically developed to measure fruit-vegetable and fat intake in worksite canteens. By excluding the fat items regarding hot meals that are generally consumed at home, an attempt was made to limit the contribution of products consumed at home to the total fat score. In addition, our intervention focused also on vending machine products, but the questionnaire did not include questions on this issue. Nevertheless, these questionnaires were used to measure fruit-vegetable and fat intake, because of a lack of a validated short food frequency questionnaire, which are applicable to measure Dutch worksite food patterns.
A weak point in this study was that a relatively large proportion of the study population was not a regular visitor to the company canteen (about 40%). Because of this, the food intervention did not have the full impact it could have had. However, at follow-up no interaction was found between whether or not being a regular visitor to the canteen, and fruit-vegetable and fat intake. Also, the food intervention might have been too modest to sort any effect. As mentioned in the method section, only one product group at the time was highlighted by means of larger information sheets near the products included in the selected group. No information was put directly on the products and no clear-cut distinction between healthy or unhealthy products was made (for instance labeling products with either red or green colors), like in a study of Larsson et al . Larsson et al used a food-marking symbol (the 'Green Keyhole') to make it easier for consumers to select low-fat and high fiber alternatives. This symbol was used on products that were an alternative to high-fat or low-fiber products. Perhaps in our study a comparable and a more obvious distinction between products should have been made. Instead of focusing on all products within pre-selected product groups (and one contrasting unhealthy alternative), the focus of the intervention should have been more on giving information solely on more healthy products. In our study it was hypothesized that, when giving information about the caloric value of a healthier product and the unhealthier alternative (e.g., high and low fat cheese), the subjects would choose a healthier alternative more often.
A limitation of this study might be the fact that no randomization was performed. Bias introduced by possible differences between worksites, might have been prevented if a randomization at the level of the individual could have been performed. However, due to the nature of the intervention this kind of randomization was not possible. Moreover, the main reason for not performing randomization at the level of the worksite was that, at the moment that the FoodSteps research proposal was approved, one worksite had already agreed to participate. In order to speed up the preparations of the intervention, this worksite was chosen as the intervention worksite. During this preparation period the control worksite still had to be found.