A near-infrared interactance model for the estimation of infant body composition

In low and middle income settings, there is a pressing need for simple, low-cost and robust ways of determining infant body composition. Measuring and monitoring body composition plays an important role in addressing the burden of child deaths under five, where depleted fat stores may put infants at risk of mortality, morbidity in later life and delayed cognitive development. In a subset of 41 infants aged between 2 and 8 months, we develop and evaluate a potential model for using near-infrared interactance (NIR) features for the estimation of infant fat mass and the deuterium dilution method as the reference method. Model evaluation demonstrates the potential of NIR to capture variability in body composition, with a correlation of R = 0.90 and Bland-Altman agreement (mean bias with 95% confidence intervals) of -8 (-553, 536) g with fat mass estimated by a four-compartment (4C) model. With continued development in the larger dataset, this model may enable simple and low-cast body composition assessment of infants in low and middle incomes settings where the burden of infant deaths is greatest.