A number of individual risk factors for childhood obesity have been identified, but only some of these are amenable to prevention. To assess the amount of cases in a general population attributable to these risk factors, adjusted population-attributable fractions were estimated.

Cross-sectional study.

Obligatory school entry examination in 2001/2002 in six Bavarian communities (Germany).

5472 children at age 5–6 years.

Anthropometric measures were ascertained by public health nurses, and measures concerning sociodemographics, lifestyle and child behaviour such as child's daily meal frequency were obtained with self-administered parental questionnaires. Obesity was defined according to sex- and age-specific body mass index cut-off points proposed by the International Obesity Task Force. Adjusted population-attributable fractions were calculated based on logistic regression.

A combination of the risk factors low meal frequency, decreased physical activity, watching television >1 h day− 1, formula feeding and smoking in pregnancy accounted for 48.2% of obese children. This combination yielded a maximal achievable prevalence reduction of 1.5% for obesity (3.2% observed prevalence).

A modification of five known risk factors for childhood overweight and obesity could reasonably lower obesity prevalences at school entry. These risk factors should be particularly considered in decision making on preventive measures.