Univariable and multivariable regression analyses comprising DNA methylation sites that were significantly correlated with AS3MT haplotype in either Argentina or in Bangladesh.

*The β–coefficient refers to the slope of the regression line (change in methylation level) with increasing numbers of copies of the haplotype.†β and p-value from univariable regression. The models look as follows: Methylation level = α+β1×copies of AS3MT haplotype.‡β and p-value from multivariable regression (Argentina). The model look as follows: Methylation level = α+β1×copies of AS3MT haplotype+β2×total urinary arsenic (natural ln transformed)+β3×age+β4×use of coca leaves (yes/no).§β and p-value from multivariable regression (Bangladesh). The model look as follows: Methylation level = α+β1×copies of AS3MT haplotype+β2× total arsenic in maternal urine around gestational weeks 6–8 (natural ln transformed)+β3×maternal age+β4×sex of child.