Metabolomic alterations in cord blood improve the prediction of childhood-onset neurodevelopmental disorders

We aimed to identify cord blood metabolic measures associated with childhood-onset neurodevelopmental disorders (NDDs) and to examine whether they improve prediction beyond known early-life risk factors. In a prospective study of 858 children followed to a median age of 14.1 years (IQR 13.7-14.8), cord blood levels of 110 metabolic measures were assessed at birth and NDD diagnoses obtained from nationwide register. We identified 12 cord blood metabolic measures together explaining 5.0% of variance in any NDD, seven related to HDL synthesis, concentration and composition, one to concentration of lipoprotein particles, two to fatty acids and two to amino acids; their levels were 0.26-0.36(95%CI: -0.55, -0.06) standard deviations lower in the children who were later diagnosed with any NDD compared to the children with no mental disorders (p-value for overall model including all 12 metabolic measures <0.0001). They were also associated with early-life risk factors of NDDs, and improved the prediction of any NDD over a model that included only sociodemographic, prenatal and birth-related early-life risk factors(AUC = 0.74,95% CI 0.69-0.79; R2 = 16.6%; p < 0.0001 vs. AUC 0.71,95% CI, 0.66-0.76; R2 = 12.2%; p < 0.0001; Likelihood Ratio Test [LRT] p = 0.04). The improvement was more modest when the model additionally included maternal polygenic risk scores for ADHD and ASD (AUC = 0.75,95% CI 0.70-0.80; R2 = 18.3%; p < 0.0001 vs. AUC 0.72,95% CI, 0.67-0.78; R2 = 13.9%; p < 0.0001; LRT p = 0.08). Cord blood metabolic measures predicted NDDs diagnosed from birth through adolescence and improved prediction beyond early-life risk factors, with smaller incremental value when maternal PRSs for ADHD and ASD were included.