ECG parameters in episodes of neonatal hyper- and hypoglycemia: an observational study.

Introduction: Abnormal glucose concentrations are common in the neonatal period and may entail long-term consequences. They are currently detected invasively through spot-measurements, contribute to blood loss and are painful depending on the sampling route. We hypothesized that hypo- and hyperglycemia may be associated with changes in neonatal electrocardiogram (ECG), such as PR interval, QRS duration, and ST segment durations. Methods: We analyzed ECG parameters and categorized blood glucose measurements [hypo-, normo-, hyperglycemia (locally defined as >180 mg/dL)] using generalized linear mixed-effects models. Potential ECG predictors of blood glucose abnormalities were first identified in univariable models and then entered into a multivariable model to assess associations, adjusting for postmenstrual age and time between two consecutive measurements. The direction and magnitude of observed associations were assessed in leave-one-subject-out sensitivity analyses. Results: 959 paired measurements (32 hypoglycemia, 714 normoglycemia, 213 hyperglycemia) from 15 neonates were analyzed. In the multivariable model, PR interval was significantly associated with hypo- and hyperglycemia. QRS duration was associated with hyperglycemia, whereas ST segment duration was related to hypoglycemia. All associations remained significant after adjustment for postmenstrual age and time between two consecutive measurements. Within the constraints of the limited sample size, the leave-one-subject-out sensitivity analysis suggested that the direction and magnitude of effect sizes were largely consistent across iterations. However, statistical significance was sensitive to single-subject exclusions, particularly for the PR interval in hypoglycemia. Intraclass correlation coefficients were 0.65 for hyperglycemia and 0.41 for hypoglycemia, indicating relevant intra-individual clustering. Discussion: The findings of this hypothesis-generating study in 15 individual infants suggest that cardiac electrophysiology may provide information about blood glucose abnormalities in the neonatal period. Especially findings on hypoglycemia remain exploratory due to the limited sample size. ECG parameters may serve as candidate features for non-invasive prediction of hypo- and hyperglycemia in critically ill neonates.