Download Full Text (2.2 MB)


Background: Infants born prematurely are at increased risk for neurodevelopmental impairments during childhood and early neurodevelopment can be influenced by multiple factors (perinatal and environmental exposures, genetics). DNA methylation (DNAm) can also be influenced by these factors and be reflective of health and development, and thus may provide an integrated measure of early life risk for neurodevelopmental outcomes.

Objective: To test whether neonatal DNAm is predictive of cognitive, language, and/or motor performance at 24 months of age in children that were born very preterm.

Design/Methods: We studied 433 neonates born < 30 weeks postmenstrual age in the Neonatal Neurobehavioral Outcomes in Very Preterm Infants (NOVI) study. Neonatal DNAm was measured from buccal swabs at NICU discharge via the Illumina MethylationEPIC BeadArray. Cognitive, motor, and language performance were assessed by the Bayley Scales of Infant and Toddler Development-III (BSID-III) at 24 months adjusted age, with mild and moderate impairment defined as composite scores < 85 and < 70. We used an elastic net regression and leave-one-out cross-validation to limit overfitting and selection bias, to identify sets of DNAm sites that were predictive of composite Bayley scores. For the leave-one-out cross-validation, a model was trained on all but one sample to make a prediction on that held-out sample. This step is repeated until each sample was held out and predicted once. In addition, for a given hold-out sample, all siblings were removed from its training set. The performance of predicted composite scores were evaluated by their correlation with the measured Bayley scores and the area under the receiver operating characteristic (ROC) curve (AUC) based on dichotomized Bayley scores (< 85 and < 70).

Results: We identified sets of DNAm sites (including 33-306 sites) that predicted composite scores for the BSID-III at 24 months of age, which were positively (r>0.3) and significantly (p-value <1e-11) correlated with reported scores (Figure 1A-C). Additionally, neonatal DNAm at these sites was predictive of mild cognitive (AUC=0.68), language (AUC=0.71), and motor impairments (AUC=0.66) (Figure 1D-F), and yielded similar predictive statistics for moderate impairments.

Conclusion(s): DNAm patterns measured at NICU discharge were indicative of cognitive, language, and motor performance at two years of age, among children born very preterm.

Presented at the 2021 PAS Virtual Conference

Publication Date




When and Where Presented

Presented at the 2021 PAS Virtual Conference

Neonatal DNA methylation as a predictor of cognitive, language, and motor performance at 24 months adjusted age, among children born very preterm

Included in

Pediatrics Commons