Risk factors for eating disturbances: A meta analysis

The purpose of this study was to assess the relationships between antecedent risk factors and later disturbed eating outcomes. In order to accomplish this, 203 effect sizes from 36 independent studies were pooled for analysis. These effect sizes represented data for 19 predictors constructs and 5 disturbed eating outcomes. Preliminary findings suggest that eating disturbances can be predicted by antecedent behavior, characteristics or experience. However, estimates of the magnitude of the relationship of these antecedent conditions with later disturbed eating show considerably more heterogeneity than can be attributed to sampling error. A series of regression analysis indicated that the strength of the relationship between predictor and outcome was influenced by the predictor construct chosen and two methods construct; whether or not the sample was drawn from the United States and the proportion of study authors that were females. Based on statistical and conceptual considerations, however, data were not adjusted to account for the variance attributable to these two constructs. Type of disturbed eating outcome also modestly influenced the magnitude of the effect size. Predictors were further grouped into 13 and each unique predictor - outcome relationship was assessed. Overall, effect size estimates for nearly all of these relationships were statistically significant although the magnitude of some was quite small. These estimates were largest when the predictors included dietary restraint, body dissatisfaction, thin ideal internalization, affect and body mass were assessed. Particularly small effect sizes were observed for individual maturation, substance use and interpersonal distrust. In terms of outcome, larger effect sizes tended to be associated with the outcomes bulimic symptoms and dietary restraint. The strength of the relationships between the variables supports the hypothesis described in the dual pathways model of bulimia nervosa (Stice, 2001), although specific developmental pathways could not be assessed. Because initial cases of disorder were included in the analysis, conclusions regarding risk status should be made with caution