Data analyses The principle of intention to treat was applied, but in reality all participants who took part in the 16-week measurement had also completed the interventions. The primary analysis concerned differences in outcomes between the BEL and the CAU group and the stability of the same outcomes at the six-month follow-up. The instruments used produced ordinal scales and since equal distances between scale steps could not be assumed, mainly non-parametric statistics were used. To use the variation in an optimal way sum scores were used, and change scores were calculated as the difference between a later and a previous measurement (e.g., the follow-up score minus the baseline score). Two sets of scores were computed, namely differences from baseline to completion of the intervention and from baseline to the followup. The inferential statistics used to test differences in change scores between the BEL and CAU group were the Mann-Whitney U-test. For descriptive purposes, the raw change scores were transformed to Tscores; that is, all scales got the same mean (=50) and standard deviation (=10). The Wilcoxon test, performed on the respective samples separately, was used to shed further light on findings regarding change scores. In order to account for clustering effects, parametric statistics had to be used, and mixed linear model was employed to calculate intra-class correlations (ICC). The level for a statistically significant pvalue was set at p < 0.05, but all p-values