Forty-six studies of 61 obesity prevention programmes were included (n=23,172). There was random assignment of participation in 51 programmes.

Thirteen interventions evaluated in 14 trials showed a statistically significant effect of the intervention. The mean effect size was small (r=0.04), but was significantly higher than zero (p=0.01). The range of effect sizes was -0.24 to 0.50. There was statistically significant heterogeneity between the studies (p<0.001).

Univariate analysis.

An extensive investigation of potential factors moderating the effect size for interventions was reported. This suggested that interventions were more effective in trials enrolling children over 11 years and adolescents (r=0.07, p<0.05) than in those enrolling pre-adolescents (r=0.03, p=0.07), and that effect sizes were also larger in trials with female participants (r=0.13, p<0.01) than in those conducted with mixed or male populations (r=0.02, p=0.06). Interventions with a shorter duration (less than 16 weeks) also appeared to be more effective (r=0.06, p<0.01), as were interventions solely focused on weight change (r=0.09, p<0.001). Trials with self-selecting recruitment also showed larger effect sizes (r=0.14, p<0.001) than population-based interventions, for which effect sizes did not differ from zero (r=0.02, p=0.10).

Multivariate analysis.

Variables found to be significant predictors of effect size were entered into a multivariate model. Only age and self-selected recruitment were found to be statistically significant in the multivariate model.