Is maternal diabetes during pregnancy associated with neurodevelopmental, cognitive and behavioural outcomes in children? Insights from individual participant data meta-analysis in ten birth cohorts

Abstract

Background: Growing evidence shows that dysregulated metabolic intrauterine environments can affect offspring's neurodevelopment and behaviour. However, the results of individual cohort studies have been inconsistent. We aimed to investigate the association between maternal diabetes before pregnancy and gestational diabetes mellitus (GDM) with neurodevelopmental, cognitive and behavioural outcomes in children.

Methods: Harmonised data from > 200 000 mother-child pairs across ten birth cohorts in Europe and Australia were available. Mother-child pairs were included for analysis to determine whether GDM was recorded (yes or no) and whether at least one neurodevelopmental, cognitive and behavioural outcome was available in children aged 3 to 13 years. Confounder-adjusted regression models were used to estimate associations between maternal diabetes and child outcomes using two-stage individual participant data (IPD) meta-analysis. Model 1 included a crude estimate. The full adjustment model (model 2) included adjustment for child sex, maternal age, pre-pregnancy BMI, pregnancy weight gain, maternal smoking during pregnancy, plurality, parity and maternal education.

Results: Children (aged 7-10 years) born to mothers with GDM had higher attention-deficient hyperactive disorder (ADHD) symptoms compared to non-exposed controls (model 2, regression coefficient (β) 3.67 (95% CI 1.13, 6.20), P = 0.001). Moreover, children (aged 4-6 years) born to mothers with GDM exhibited more externalising problems than those born to mothers without GDM (model 2, β 2.77 (95% CI 0.52, 5.02), P = 0.01). A pre-existing maternal history of type 1 and type 2 diabetes mellitus was associated with ADHD symptoms at 4-6 years (model 1, β 8.82 (95% CI 2.21, 15.45, P = 0.009) and β 7.90 (95% CI 0.82, 14.98, P = 0.02), respectively). The association was no longer apparent in further adjustments.

Conclusions: This study found that children between 4 - 6 and 7-10 years of age born to mothers with GDM have a greater likelihood of developing externalising problems and ADHD symptoms, respectively. Externalising problems often co-exist with ADHD symptoms and precede formal ADHD diagnosis. Overall, this large-scale multi-cohort study suggested that a dysregulated metabolic environment during pregnancy may contribute to ADHD symptoms and externalising problems in young children.

Keywords: Attention deficient hyperactive disorder; Autism spectrum disorder; Behavioural; Cognitive; Externalising problems; Gestational diabetes mellitus; Neurodevelopmental.

Fig. 1

Association between GDM and offspring’s ADHD symptoms at 4–6 and 7–10 year of age. The forest plot shows the Regression Coefficient (β) and random effect (RE) for ADHD percentiles. Model 1 (a) and (d) include crude estimates, model 2 (b) and (e) full adjustment for child sex, maternal age, plurality and parity, BMI, pregnancy weight gain, and maternal smoking, and sensitivity adjustments (model 3 (c) and (f)) include full adjustment and EU-SILC. The Sensitivity adjustment excludes the Raine Study

Fig. 2

Association between GDM and offspring’s externalising problems at 4-6 and 7-10 year of age. The forest plot is showing Regression Coefficient (β) and random effect (RE) for externalising problems in children 4–6 and 7–10 years of age exposed to GDM versus children not exposed to GDM. Model 1 (a) and (d) include crude estimates, model 2 (b) and (e) full adjustment for child sex, maternal age, plurality and parity, BMI, pregnancy weight gain, maternal smoking and sensitivity adjustments (model 3 (c) and (f)) include full adjustment and EU-SILC. The Sensitivity adjustment excludes the Raine Study

Fig. 3

Association between T1DM and T2DM and offspring’s ADHD symptoms at 4–6 year of age. Forest plot showing Regression Coefficient (β) and random effect (RE) for ADHD percentile in children (4–6 years)) exposed to (1) T1DM and (2) T2DM versus those not exposed to T1DM and T2DM, respectively. Model 1 (a) and (d) include crude estimates, model 2 (b) and (e) full adjustment for child sex, maternal age, plurality and parity, BMI, pregnancy weight gain, maternal smoking and sensitivity adjustments (model 3 (c) and (f)) include full adjustment and EU-SILC. The Sensitivity adjustment excludes the Raine Study