Objective Aim of this work is to provide a meta-analytic framework for the analysis of obesity and its determinants in children, where complex inter-relationships are observed among risk factors. To properly address the difficulties in managing such interactions and still to allow accessible interpretation of the results, a Bayesian Network model is developed.

Methods Seven risk factors have been considered as being potentially related with obesity in children (0–14): physical activity, residence (urban, central city, suburban), TV viewing, PC/playstation usage, parental education, smoking status of the parents, snacking behaviour. Influence on obesity has been evaluated reviewing the relevant literature, as emerging from a PUBMED search for “obesity” and “risk factor”. A Bayesian Network (BN) has been developed on this data for modelling the interactions among each factor. A sensitivity analysis has been performed to assess model validity and accuracy.

Results The BN model allowed us to estimate the individual risk of developing obesity. Each risk profile is thus associated with a specific pattern in each of the risk factors considered, e.g. a female child, doing regular exercise, watching TV, using computers and snacking has a probability of becoming obese of 18%, whereas the same child, where no physical activity is done, has a probability greater that 48%.

Conclusions Bayesian Networks are a useful tool for investigating and summarizing evidence when complex relationships exist among risk factors, in particular, as in the case of obesity, when there’s a concurrent incidence of several of them, interacting in complex ways.