This study investigated the association between markers of folate-status in utero and risk of developing psychotic symptoms at age 12. Our analyses focused on the direct and moderating influences of maternal folate supplementation assessed at 18 and 32- weeks gestation, maternal dietary folate intake at 32- weeks gestation and maternal and child MTHFR C667T genotypes.
Overall, we found no support for the hypothesis that determinants of impaired perinatal folate-status increase risk of non-clinical psychotic symptoms, as is commonly hypothesised for schizophrenia. There was nominal evidence for higher odds of PLIKS in children when mothers supplemented their diet with folate early during pregnancy (18 weeks). These results were consistent in their direction of association with the observed maternal MTHFR
C677T effect (recessive model only); in particular boys had lower odds of expressing PLIKS when their mothers carried the TT genotype, which is associated with lower folate-status (Brattström et al., 1998; Friso et al., 2002
). However, there was no evidence that the studied risk factors acted jointly or through gene-nutrient interactions, and none of the reported overall findings was significant when adjusted for multiple testing. Although observed associations for folate supplements at 18 weeks and maternal TT genotypes may suggest risk effects of higher fetal folate-status, especially in boys, they are likely to represent chance findings.
The power of our study to detect genetic and environmental effects was good to excellent. For example, within the Child MTHFR
C667T sample the power approached 75% (recessive model) to approximately 100% (log-additive model) to detect a moderate recessive (OR = 1.44)(Gilbody et al., 2007
) or log-additive genotype-risk effect respectively assuming a MTHFR
677T allele frequency of 0.33 and a PLIKS population risk of 10% (Gauderman and Morrison, 2006
). Likewise, the power to detect a protective folate supplement effect (OR = 0.70) within the nutrient sample, assuming for example a 24% prevalence of maternal folate supplementation, as observed at 32 weeks, was high (91%). In view of the inherent study power, it is therefore unlikely that we have missed genetic or dietary/supplement effects that relate to fetal folate status during pregnancy, if they exist as hypothesised for schizophrenia. Their absence however might be related to several reasons: First, the relationship between internal folate-status and mental health outcomes at age 12 is likely to be complex. This may involve interactions between MTHFR
C677T genotype and folate status (Brattström et al., 1998; Friso et al., 2002
) but also between folate and other key determinants of the folate cycle such as those that are regulated by vitamin B12 (e.g. Selhub et al., 2009
). However, our power to detect effect moderation was low. Assuming an alleviation of the genotype risk through folate supplementation (OR = 0.50) and otherwise similar conditions for a recessive genetic effect as outlined above, our interaction sample had only 25% power to detect gene–nutrient interactions (Gauderman and Morrison, 2006
). Second, the effect of folate intake during pregnancy may depend on the embryogenetic phase as the epigenetic architecture of the fetal genome is established during time-sensitive periods (Dolinoy et al., 2007
). It is therefore possible that the effect of fetal folate status on later mental health outcome varies during the course of the embryonic development and that our measures of maternal folate intake were inadequate at measuring the true exposure during these specific periods. Third, it is not clear to what extent non-clinical psychotic symptoms as assessed in this study reflect the same pathology underlying chronic and severe psychotic disorders such as schizophrenia. Although most established risk factors for schizophrenia also show associations with non-clinical psychotic experiences (Schreier et al., 2009; van Os et al., 2009; Zammit et al., 2009
), findings across these phenotypes are not always consistent (Zammit et al., 2008
). Our findings may therefore have relevance for studying the aetiology of non-clinical psychotic symptoms during childhood and adolescence, and contribute to the identification of risk factors that distinguish psychosis-like symptoms from schizophrenia. Fourth, we had no direct measure of folate exposure in utero
. Although both, nutrition scores based on self-reported FFQs (Kleiser et al., 2009
) and MTHFR
genotypes (Frosst et al., 1995
), have been related to serum/plasma homocysteine levels, misclassification will be present to some extent. This may have led to underestimating associations if misclassification is assumed to be non-differential with respect to psychotic symptoms, as seems most probable.
No evidence was found for bias due to non-random genotyping dropout when examining differences in estimated mating-type probabilities in PLIKS cases and controls. MTHFR
C667T genotype distributions are furthermore unlikely to be confounded through socio-economic factors (Davey Smith and Ebrahim, 2003
). Nonetheless, residual confounding remains a possibility that may have affected the selected measures of folate supplementation and dietary intake although our analyses were adjusted for a wide range of confounders including maternal parity, maternal cigarette smoking and alcohol consumption during the first three months of pregnancy, maternal family history of depression, maternal education, occupational social class, and housing tenure.
In conclusion we found no evidence that markers of folate status in utero were associated with risk of developing psychotic symptoms in adolescence.