In sensitivity analyses, we further adjusted for paternal smoking obtained at the first survey (i.e., 6 months of age of children) as an indicator of environmental tobacco exposure. At the first survey, the parents were asked whether they moved or built an extension since 1 year before birth (to the 6 months of age), and thus restricted the analyses to children whose parents answered they did not move or build an extension and repeated the analyses. To reduce exposure misclassification, we restricted the participants to children who were born in the municipalities with an area less than 400 km2 (square of 20 km). Finally, we repeated multilevel logistic regression analyses adjusting for both individual and municipality-level variables considering that participants were nested in each municipality instead of each prefecture.