0367) so that weight gain was seen in workgroups with high BMI levels. Quadratic effects showed that smoking cessation was indeed predicted by the percentage of smokers in the group, in that smoking cessation happened in the workgroups with the largest share of smokers (p = 0.0258). However, change in LTPA was not associated with the average activity level in the group. The purpose of this study was to investigate the importance of workgroups with regard to health behaviours and lifestyle changes. We investigated whether workgroups would account for part of the variation within health behaviours

and lifestyle changes. We found evidence for cluster BAY 73-4506 mw effects regarding current health behaviours; part of the variation in BMI, smoking status and amount smoked was explained by workgroups (2.62%, 6.49% and 6.56%, respectively). Workgroups Ku-0059436 manufacturer explained little of the variation in LTPA. With regard to changes in lifestyle, we found no significant effect of workgroups on variation in smoking cessation, smoking reduction, change in BMI, or change in physical activity. We did find that workgroup weight change depended on the average level of BMI in the group. Also, workgroup smoking cessation was seen in groups with larger shares of smokers. However, the average LTPA level did not predict change in LTPA level. Christakis and Fowler, 2007 and Christakis and Fowler, 2008 found clustering effects for obesity

and smoking cessation. Other researchers (Cohen-Cole and Fletcher, 2008a, Cohen-Cole and Fletcher, 2008b and Lyons, 2011) have suggested that the association could be explained by shared environmental factors and a tendency of forming relationships with people who have similar characteristics (homophily). Subsequent sensitivity analyses of the original studies found that the findings regarding obesity and smoking were reasonably robust to latent homophily and unmeasured environmental factors (VanderWeele, 2011). Another study using the methods of Christakis and Fowler found that attributes such as acne, height Linifanib (ABT-869) and headaches also seemed

to spread through social ties (Cohen-Cole and Fletcher, 2008a). This has led some authors to question the interpretation of the original findings (Cohen-Cole and Fletcher, 2008a) while others conclude that the original findings of contagion effects cannot be dismissed (VanderWeele, 2011). A potential advantage of our study is the use of a different methodology. Similar to Christakis and colleagues, our baseline might be influenced by homophily, but in our design, clustering of change could not have been explained by homophily. Since we only found significant effect of workgroup on baseline health behaviour, our study cannot rule out homophily as an explanation of the clustering of health behaviours. To reduce the risk of residual confounding we controlled for occupational position, lifestyle factors, and age, gender and cohabitation.