Clinical performance and practical implications of continuous glucose monitoring in very preterm infants: a cohort study.

Abstract

Objective: To explore the clinical performance of continuous glucose monitoring (CGM) in very preterm infants during the first week of life, using data from the multicentre Real time continuous glucose monitoring (REACT) randomised controlled trial (RCT) and to consider how CGM can best support neonatal glucose management.

Design: Prespecified secondary analysis of paired sensor glucose (SG) and blood glucose (BG) measurements from the REACT RCT.

Setting: 13 neonatal intensive care units (NICUs) in the UK, Spain and the Netherlands.

Patients: 155 very preterm infants (≤1200 g birth weight or <34 weeks' gestation) with CGM data collected during the first week of life.

Main outcome measures: Mean absolute relative difference (MARD), ISO 15197:2013 system agreement, Bland-Altman bias, error grid analysis and clinical contextualisation of CGM performance.

Results: Across 2223 SG-BG pairs, overall MARD was 11.9%, and 98%-99% of values fell within clinically acceptable error grid zones. SG tended to be less accurate at BG extremes, although error grid analyses show that clinical risk from discrepancies was low. CGM detected episodes of dysglycaemia not captured by intermittent sampling. Glucose variability did not materially influence accuracy.

Conclusions: CGM can provide clinically meaningful information in very preterm infants despite modest analytical inaccuracy when compared with single point-of-care values. Its greatest utility lies in continuous trend monitoring, early detection of silent dysglycaemia and reduced invasive sampling, rather than replacement of diagnostic blood testing. Clinicians should interpret SG values in context, confirm unexpected extremes with BG and focus on patterns rather than single values.