Fig. 2

Associations of occasional (a) and regular (b) prenatal self-reported exposure to solvents with behavior traits at ages 2 and 6 (N = 715, PELAGIE cohort, France, 2002–2013), compared to non-exposure. Latent traits: Internal, internalizing behavior; External, externalizing behavior. All regression coefficients were adjusted for sex, education level, maternal age, breastfeeding duration, smoking during pregnancy, parity and mother-child interaction score. Results are presented in Table 3. Bold arrows indicate significant standardized regression coefficients (and confidence intervals). The association of latent traits at age 2 with prenatal exposure can be directly read on the graph arrows on the top of the graph. The direct association of latent traits at age 6 with prenatal exposure can be directly read on the broken arrows around the graphs. The indirect associations of each latent trait at age 6 can be derived by summing the associations through the two pathways pointing to it via internal and external traits at age 2. Each of these pathways generates an association with exposure. The intensity of the association is given by multiplying all the regression coefficients situated on the respective pathway. The total association, indicated in the shaded boxes under each latent trait at age 6, is obtained by summing the direct and indirect associations with exposure. Example of decomposition of the total association of occasional exposure to solvents with externalizing behavior at age 6: Direct association = 0.01; Indirect association via internalizing behavior trait at age 2 = (0.14*-0.06); via externalizing behavior trait at age 2 = (0.34*0.64). Indirect association = (0.14*-0.06) + (0.34*0.66) = 0.224. Total association = − 0.008 + 0.224 = 0.216