We have confirmed that a FGGRS is strongly associated with fasting glucose in a population based cohort and therefore may be used to determine the effect of glucose exposure on a phenotype of interest. It was not only strongly associated with fasting glucose levels but also with levels of HbA1c
. Not all SNPs reported to be associated with fasting glucose
 showed an association with this exposure in our analysis but three of the variants with the strongest effects in earlier studies did (i.e. those in ADCY5, G6PC2 and GCK). Genes associated with HbA1c in our study include two described in a previous GWAS of HbA1c (MTNR1B and G6PC2), although the variants examined were different
, and a SNP in WFS reportedly linked to HbA1c in a prospective epidemiological study conducted in France
The T2DGRS, including all variants associated with T2D, some of which were also associated with fasting glucose, was not associated with plasma fasting glucose or HbA1c. Neither were the maternal and the offspring FGGRS nor the offspring T2DGRS associated with IQ in the ALSPAC population. In contrast, the maternal T2DGRS was associated with an increase of 0.03SD in child’s IQ after adjustment for offspring T2DGRS and population stratification.
Additionally, we uncovered associations with IQ for a few individual SNPs in mothers and children, the majority of which were T2D but not fasting glucose-associated variants. Only the variants in PROX1 and G6PC2 had been associated with levels of fasting glucose before
[18, 40]. Since there was some indication that the PROX1 SNP was in HW disequilibrium in mothers and children this association should be carefully assessed. However, evidence for the association of IQ with individual variants disappears after a Bonferroni correction for multiple testing (cut-off p-value: 0.001).
We did not detect an association of the FGGRS with diabetes/glycosuria in the mothers, yet this outcome was positively associated with the T2DGRS. SNPs independently associated with maternal diabetes/glycosuria had been earlier related to plasma fasting glucose regulation and T2D risk.
Overall, there was very little overlap between the SNPs associated with fasting glucose and HbA1c levels, maternal diabetes/glycosuria and IQ scores in ALSPAC.
Given the results described above, we have not found sufficient evidence that fasting glucose concentration either in mothers during gestation or in children later in life affects cognitive ability at mean age 8, but we lacked statistical power to reliably establish the absence of association. Despite the FGGRS being strongly associated with fasting glucose in plasma its effects are quite small therefore a larger sample size will be required to carry out an adequately powered analysis.
On the other hand, we have uncovered some evidence that children of mothers who carry T2D susceptibility alleles exhibit higher IQ scores than children of non-carriers. This appears to be in disagreement with recent studies, and our study as well, that have revealed that maternal pre-pregnancy diabetes and gestational diabetes, especially with poor metabolic control, have an adverse effect on offspring cognition
[9, 41, 42], although not all studies have detected a reduction in IQ scores
[43, 44]. We speculate that the presence of T2D risk alleles (which could be contributing to maintain higher glucose levels in blood) may be beneficial during pregnancy as women have to fulfil the baby’s glucose requirements besides their own. However, having a fully established disease phenotype, such as gestational diabetes, can in turn be detrimental, either because glucose levels end up being excessively high or because of the existence of factors related to the disease, other than hyperglycemia, that unfavourably influence neurodevelopment and are not captured by these genetic variants and scores. Additionally, the negative association between diabetes and cognition could be the result of confounding by lifestyle and other factors such as poor diet, smoking, alcohol intake and a lower educational achievement. Further studies are definitely needed to examine this issue in greater detail.
Strengths and limitations
Among the strengths of our study is having access to a large birth cohort such as ALSPAC to explore the role of maternal fasting glucose status during pregnancy and their offspring own fasting glucose levels in relation to cognitive ability in children. Then, the validation in the ALSPAC population of a robust instrument such as the FGGRS, which can be used in future Mendelian randomization studies to determine the causal relationship between fasting glucose (or HbA1c) levels and specific phenotypes. Lack of power to obtain conclusive results is likely the major weakness of this study. Additionally, because most individuals had fasting glucose and HbA1c levels within the normal range, it may have not been possible in this setting to unravel an association that manifests itself only under very high or very low glucose concentrations.