Bibian van der Voorn

110 CHAPTER 8 transformation. Therefore, we decided not to correct for skewness and to present these as mean ± SD, in line with previous analyses in this cohort 23 . Given the sex specific differences in neurological outcome in previous analyses of this cohort 23,24 , all analyses were corrected for sex. Statistical analysis was done by stepwise linear regression, with IQ or behavior as dependent factor. First, the influence of antenatal glucocorticoid treatment on IQ and behaviorwas investigatedusingmultivariate regression, with antenatal glucocorticoid treatment and sex as independent factors. Second, the influence of the GR and MR polymorphisms on IQ and behavior were independently analyzed using multivariate regression, with the SNP concerned and sex as independent factors. Third, the interaction term, antenatal glucocorticoid treatment and the SNP concerned were added to the model. Fourth, we repeated all analyses, corrected for perinatal risk factors and socioeconomic status ( table 1 ). RESULTS The allele frequencies of thesepolymorphisms inour sample, of 4.0%, 2.6%, 48.5%and 9.9% respectively, were well in range with those observed in the normal population 17, 25, 26 . The genotype distributions were in agreement with the distribution predicted by the Hardy-Weinberg equilibrium, except for the -2G/C SNP. This has been described earlier in a normal Caucasian population sample 17 . Compound SNP carriage was not present for the GR polymorphisms, in contrast to theMR SNPs. In addition, intercourse compound variation (the combination of both a GR and a MR polymorphism) was observed in a small number of subjects ( table 2 ). Table 1 shows that perinatal variables, sex, low socioeconomic status and GR and MR SNPs were equally divided between treated and non-treated subjects. Table 3 presents the relation between antenatal glucocorticoid treatment versus IQ and behavioral problem scores. Antenatal glucocorticoid treatment was not associated with IQ-score. Antenatal glucocorticoid treatment was associated with unfavorable scores on self-reported and parent-reported internalizing behavior and with unfavorable scores on parent-reported externalizing behavior and total problem scores. Table 4 shows the relation between the GR polymorphisms versus IQ, YASR and YABCL scores. Carriers of the R23K SNP had a significantly higher IQ-score compared to non-carriers, as well as higher Z-scores of the intellectual subdomains logical