Reliability and Observer Dependence of Signs of Neonatal Hypoglycemia

To evaluate, using video documentation, the sensitivity, specificity, and interobserver reliability of visualizable signs of neonatal hypoglycemia at different glucose concentrations in neonates.

In a prospective cohort study of 145 neonates with and without risk factors for hypoglycemia, 430 videos were recorded before blood glucose measurements and analyzed by 10 blinded investigators of different professions. The primary outcome measures were sensitivity and specificity for clinical detection of hypoglycemia.

The overall sensitivity to detect low blood glucose (<55 mg/dL [<3.1 mmol/L]) based on signs was 30%, and the specificity was 82%. Significantly more investigators suspected hypoglycemia while viewing videos of infants with blood glucose levels of 46-54 mg/dL (2.6-3.0 mmol/L) and 30-45 mg/dL (1.7-2.5 mmol/L) compared with ≥55 mg/dL (≥3.1 mmol/L) (29 ± 3% and 31 ± 4% vs 18 ± 1%; P = .001; P = .007). After 48 hours of life, significantly more investigators suspected hypoglycemia in videos of infants with blood glucose levels of ≤45 mg/dL (≤2.5 mmol/L) compared with blood glucose levels of >45 mg/dL (>2.5 mmol/L) (28.9 ± 8.1% vs 10.9 ± 1.8%; P = .007). For blood glucose levels 30-45 mg/dL (1.7-2.5 mmol/L), sensitivity varied widely between investigators, ranging from 5% to 62%. Three hypoglycemic episodes <30 mg/dL (<1.7 mmol/L) were only partially recognized.

Clinical observation of signs is neither sensitive nor specific to detect neonatal hypoglycemia, and there are large interobserver differences. Thus, guidelines on neonatal hypoglycemia should reconsider whether distinguishing between asymptomatic and symptomatic hypoglycemia provides useful information for the management of neonatal hypoglycemia, because it may pose a risk for systematic under-recognition and undertreatment, leading to an increased risk for neurodevelopmental impairment.

German Clinical Trials Register: DRKS00021500 www.drks.de/drks_web/setLocale_EN.do.