Abstract
BACKGROUND: Body composition is an important nutritional status indicator among preterm infants, but is challenging to measure in the neonatal intensive care unit (NICU). Bioelectrical impedance analysis (BIA) is highly feasible, but accuracy depends on using a valid prediction equation. We aimed to evaluate the accuracy of available prediction equations, relative to air displacement plethysmography (ADP) as the criterion method, within a cohort of hospitalized preterm infants. METHODS: Preterm infants (23-35 weeks' gestation) underwent concurrent BIA and ADP up to three times between birth and term equivalent age. Using 11 published prediction equations, we estimated infant body composition from BIA data. To determine validity, we used Bland-Altman plots to compare BIA-derived fat-free mass (FFM) with FFM measured concurrently by ADP. RESULTS: One hundred and fifteen infants, with a median 30 6/7 weeks' gestation, underwent 150 instances of concurrent BIA and ADP at ages 1 to 135 (median: 40) days. Agreement between BIA and ADP-derived FFM varied widely depending on the BIA equation used, with mean bias (±95% confidence limits) ranging from 0.14 (±0.24) to 1.34 (±1.51) kg. The Dung, Lingwood, and Tint equations demonstrated the greatest agreement with ADP, with a mean bias of 0.14-0.32 kg and narrow limits of agreement (±0.23-0.28 kg). All equations demonstrated some bias. CONCLUSION: BIA is a feasible tool for measuring body composition among preterm infants. Existing published equations demonstrate reasonable agreement with ADP but require a correction factor to adjust for bias. A novel prediction equation specific to preterm infants might offer improved agreement and reduced bias.