Distribution of anthropometric and behavioral characteristics at 13 and 17 years of age according to smoking and drinking categories.

P-values were calculated using chi-square of Fisher's exact tests, as appropriate (for the comparisons between proportions), or using one-way ANOVA (for the comparisons of means).*Behavioral clusters defined using the following variables: sports activities, fruit intake, sleeping hours, time spent in sedentary activities, tobacco and alcohol use [23]. Adolescents were classified in one of three following groups according to behavioral aggregation: healthiest (cluster 1), intermediate (cluster 2) and least healthy (cluster 3).