Univariate and multiple linear regression analysis.

*Dominant models were applied for these SNPs, hence coefficients reflect the difference in methylation level for carriers of the minor allele compared to major allele homozgyotes (reference group).†Females were compared to males (reference group).‡Additive models were applied for these SNPs, hence coefficients reflect the difference in methylation level for each additional copy of the minor allele compared to major allele homozygotes (reference group).ΦRecessive models were applied for these SNPs, hence coefficients reflect the difference in methylation level for minor allele homozygotes compared to carriers of the major allele (reference group).łReduced numbers in multiple regression models are due to limited maternal genotype data and removal of outliers, consequently, these reduced numbers may in part account for the lack of significance seen with some predictor variables. Note also that mean methylation levels were utilized for multiple regression modelling despite not always demonstrating the strongest effect size with individual predictors. Standardised beta coefficients are obtained by first standardizing all variables to have a mean of 0 and a standard deviation of 1, they denote the increase in methylation for a standard deviation increase in the predictor variables. Multiple regression analysis was not performed for ZNT5 associations as mean methylation was not considered across this locus.