Generalized Estimating Equation (GEE)—Poisson regression spline function between natural ln-transformed urinary arsenic concentrations (∑As) and expected respiratory symptoms at 4 and 7 years of age.

The ∑As is in μg/L. Case-complete approach (i.e., participants with missing values in the dependent, independent, or adjustment variables were not included in the analysis). N at 4 years of age equal to 339 including 0 counts = 216 (63.7%), 1 count = 73 (21.6%), and ≥2 counts = 50 (14.7%). N at 7 years of age equal to 339 including 0 counts = 170 (50.1%), 1 count = 101 (29.8%), and ≥2 counts = 68 (20.1%). ∑As = iAs + MMA + DMA adjusted for specific gravity. Fig. A) Generalized Estimating Equation (GEE) using family equal Poisson and splines functions. Respiratory symptoms include asthma, asthma medication, wheeze, wheeze medication, sneeze, and eczema. The grey line shows the crude model. The dark blue line shows the adjusted models for child sex (boys or girls), and maternal smoking status (“Have you ever smoked?”—binary) and level of education (primary, secondary, or university studies), cohort (Asturias, Gipuzkoa, Sabadell, or Valencia), and calorie adjusted consumption of vegetables (g/day), fruits (g/day) and fish/seafood (g/day) at 4 and 7 years of age. The blue shade shows the 95% confidence interval of the adjusted model. The black dashed line shows the average expected counts (0.73). To facilitate interpretation, the crude and adjusted estimates (95% confidence intervals) at ln-transformed urinary ∑As 1.57 (average concentration), and 4.00 (99th percentile concentration) were 0.62 (0.48, 0.80)– 0.63 (0.36, 1.10) and 1.24 (0.63, 2.42)– 1.33 (0.61, 2.89), respectively. Fig. B) Density function of ln-transformed ∑As.