Background
Globally more than one billion adults are overweight (i.e., having a Body Mass Index (BMI) > 25 kg/m
2) and the numbers are still rising [
1]. In the Netherlands nearly half of the adult population is overweight [
2]. For those who are overweight, weight management (i.e., weight loss and/or prevention of weight gain) is important to alleviate overweight related health problems and to reduce chances of developing cardiovascular diseases and diabetes [
3].
Few people appear to make use of professional help for weight management [
4]. The reasons for this sparse use are not known, but clinicians not referring to professional help [
5,
6], financial costs, lack of time and personal preferences [
7] could play a role. The work setting provides an opportunity to introduce a large group of adults to a weight management programme. Worksite interventions so far used various combinations of activities and the optimal design is not clear [
8].
Weight loss programmes in the health care setting usually rely on lifestyle modification to change dietary intake and physical activity [
9]. These strategies are known to produce weight loss [
10,
11]. Typically lifestyle modification is supported by (individual or group) face-to-face counselling, requiring multiple visits to a treatment facility. This may be less appealing to working adults, who are often constrained by lack of time for such programmes. Behaviour counselling by phone and e-mail (i.e., distance counselling) could be more feasible in the work setting. In other settings distance counselling has been applied to weight loss, dietary behaviours and physical activity. Phone counselling trials for weight loss, including trials primarily aimed at changes in diet and/or physical activity, showed mixed results [
12‐
17]. The majority of phone counselling studies for physical activity and dietary behaviour found behaviour changes [
18]. Few trials have investigated e-mail counselling for weight control or lifestyle behaviours. Those that did, found positive effects on body weight, mixed effects on diet [
19,
20] and no effect on physical activity [
19‐
21]. Only one study recruited participants from a work setting [
19]. We found no studies that directly compared the impact of phone counselling with e-mail counselling.
The main purpose of this study was to ascertain effects on body weight of a lifestyle programme with 10 biweekly counselling sessions by phone as well as by e-mail compared to self help materials, in overweight workers, at six months. Secondary purposes were to determine effects on waist circumference, diet and physical activity and to compare the effects of counselling by phone with the effects of counselling by e-mail.
Discussion
Our study shows that a lifestyle programme combined with a maximum of 10 counselling sessions in six months, aimed at overweight workers, is effective for reducing body weight by 1.5 kg if counselling is done by phone and 0.6 kg if counselling is done by e-mail, compared to self help materials. Distance lifestyle counselling is also effective for producing clinically relevant weight loss. No effect was found for avoiding a 3% weight gain. The weight reduction from counselling by phone was higher than weight loss found at six months in two other studies [
15,
16]. Nevertheless, results in both intervention groups seem lower than those seen in other distance counselling studies [
17,
19,
20]. An explanation for the larger effect on weight loss in the studies by Tate et al. [
19,
20] is their explicit recommendation of a maximum daily intake of 1500 kcal, while we focused on a healthy diet. Furthermore, these studies offered more frequent contact, ranging from daily to weekly phone calls or e-mails, than we did. Their effects are in line with results from a meta-analysis [
11], showing that increasing the counselling intensity significantly increases weight reduction. Increasing intensity also raises the costs of a behavioural intervention programme. Future research should study the cost-effectiveness of different intensities.
We have also shown that the lifestyle programme with distance counselling is effective for reducing waist circumference by 1.9 cm in the phone group and 1.2 cm in the internet group, compared with self help materials. Tate et al. [
19] found larger waist circumference reductions from e-mail counselling, but these reductions are probably associated with the higher weight loss that was produced in their study.
Phone counselling resulted in an intra-group reduction of 2.7 fat points, representing 6–8 grams of fat and a reduction of 54–72 kcals per day. In an average diet of 2250 kcals per day this would constitute a 2.4–3.2% reduction in energy from fat. Another study, emphasizing much lower fat intakes than we did, showed less reduction [
17], while a second study showed a larger reduction in the intake of total fat than we achieved [
15]. This study was performed in cardiac patients that were counselled to lower their blood cholesterol. Maybe they were more motivated to change fat-intake than our overweight subjects. Nonetheless, the effect we find on fat consumption in the phone group is substantial and constitutes a meaningful contribution to weight management.
We found no intervention effects on the consumption of vegetables or fruit at six months. With regard to vegetable consumption this could be explained by a ceiling effect. Mean intake at baseline was already close to the, in The Netherlands, recommended minimum intake of 150 g/day. Alternatively, in our programme fruit and vegetables were recommended as 'healthy' choices, but their importance for weight regulation was not discussed. Whether emphasizing the role of fruit and vegetables for weight control increases their consumption should be further studied.
Physical activity levels increased as a result of the intervention, but only the phone group showed a significant difference compared with the control group. This is in agreement with studies that found increased physical activity from phone counselling [
37,
38] and no effect from internet counselling [
21].
Attendance to the counselling sessions was satisfactory in individuals with complete data, but low in those with missing data. Attendance has been found to be associated with weight loss [
14,
16,
19], so improving attendance could increase weight loss. However, the question remains if attendance to counselling sessions is responsible for successful weight change, or rather if it is a representative of an underlying motivational construct that also influences behaviour change.
A secondary aim of the study was to determine differences in the effects of phone counselling and e-mail counselling. With regard to fat intake and physical activity, the phone group appears to perform better than the internet group, because only in this group significant changes in comparison with the control group were seen. In addition, changes in the phone group are larger than in the internet group but direct comparisons between the phone and internet group showed no statistical differences.
Several potential limitations in this study need to be considered. First, for 29% of the participants no follow-up data on body weight at six months were available. This is comparable to other distance counselling studies [
16,
17,
19] and lower than in some studies in the work setting [
39,
40]. Missing data has implications. Results from completers-analyses and from analyses for which the baseline value is carried forward, are only valid if data are missing completely at random [
41]. The comparison between completers and non-completers showed that missingness was associated with observed data like baseline body weight and counselled modules. We therefore based our imputation model on missing at random (MAR) assumptions and included all variables that were related to the variables with missingness in our imputation model. An advantage of multiple imputation over single imputation methods is that it allows for the uncertainty of the values that are used to substitute the missing values [
41]. The results we found after multiple imputation differed from the completers-analyses, especially for the internet group, but are more credible because of the MAR assumption and the use of multiply imputed datasets.
A second restriction is that analyses of waist circumference and of the behavioural outcomes were limited to complete cases. Loss to follow-up was non-differential. However, in the intervention groups, participants that completed follow-up measurements had also completed more modules compared to the participants with missing follow-up. As argued before, attendance to the sessions could be indicative of adherence to behaviour change. Thus non-responders and dropouts in the intervention groups would have fewer or no change in their diet and physical activity behaviour than responders. Although non-responders and dropouts in the control group can be assumed to be equally (non)adherent to these behaviour changes, effects in all participants are probably attenuated compared to the complete-case-analysis.
A further consideration is whether the effects we found on body weight are meaningful. From the individual viewpoint additional weight loss of 1.5 kg or 0.6 kg (i.e., the mean weight losses in the phone and internet group compared to self-help materials) is not the amount wished for. However, as Rose has argued, small changes in a large group can have a huge impact on public health [
42]. A modelling study showed that reducing BMI by 2 points in a moderate to high risk group (BMI ≥ 24) has considerable effect on the population burden of diabetes [
43]. The type of programme we studied can be used to reach a large group of overweight employees; we managed to engage about 25% of the overweight working population. We therefore consider our results to be of relevance for public health. Further research should elicit if they are sustainable and cost-effective.
Other limitations of our study are that behavioural outcomes are all based on self-report and that we only measured a few of the dietary changes associated with weight control. We found that an exhaustive food questionnaire increased our questionnaire to unacceptable length. For that reason we focused on fat, fruit and vegetable intake. More objective measurement of lifestyle behaviours was not feasible because of the trial size. Self-report is vulnerable to social desirability bias which especially at follow-up might have led to more favourable outcomes.
Lastly, the study population does not represent the general Dutch working population (40% high educated, 57% men). This is related to the fact that we mostly included companies that employ white collar workers. Also, self-selection of more health oriented workers probably took place judged by baseline adherence to public health guidelines which is higher than found in the general population and by the proportion of smokers which was lower than expected on the basis of education level and age. This is a common phenomenon in lifestyle interventions, demonstrating that it is hard to engage those who, from the public health perspective, are most in need of change. When an intervention like ours is implemented in the work setting, efforts should be made to recruit lower-educated and high-risk individuals, and effects from the intervention in this population should be evaluated.
Strengths of our study include objective measurement of body weight, broad inclusion criteria, size of the group studied, use of multiple imputation for missing data, recruitment of individuals who previously had not been engaged in weight loss programmes and the design of an intervention suitable for the occupational setting. We are therefore confident that the programme we developed and the results we found are transferable to the occupational health practice.
Acknowledgements
This research was supported by The Netherlands Organization for Health Research and Development (2100.0096) who funded this study within the Prevention programme, The Netherlands Heart Foundation (2001B167) and Body@Work TNO-VUmc (2720).
Prof. dr. Jos Twisk (VU University Medical Center and VU University, Amsterdam) is acknowledged for his statistical advice. Dr. Martijn W Heymans (VU University Medical Center and VU University, Amsterdam) applied the multiple imputation method and we are thankful for his advice on the analysis of the datasets.
We recognize our counsellors Maaike Gademan, MSc, Maartje van Stralen, MSc, Marijne van der Wal, BSc, RD and Marijke Hollander, BSc, RD. We also acknowledge our research assistants David Samoocha, MSc and Ellen Paap, MSc, and the numerous data-entry assistants. We are grateful for the assistance of the trainees from the Hogeschool Holland in performing the follow-up measurements.
We also thank the two reviewers for their help in improving the manuscript.
Finally, we greatly acknowledge the companies and employees that participated in the study.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
WM and GAMA conceived of the study. JCD and MFW acquired the data. MFW performed the statistical analysis and interpretation of data and drafted the manuscript, under supervision from WVM, GAMA and JCD. All authors participated in the design of the study, in the revision of the manuscript and read and approved the final manuscript.