Anthropometric prediction models of body composition in 3 to 24month old infants: a multicenter international study.

Background: Accurate assessment of body composition during infancy is an important marker of early growth. This study aimed to develop anthropometric models to predict body composition in 3-24-month-old infants from diverse socioeconomic settings and ethnic groups.
Methods: An observational, longitudinal, prospective, multi-country study of infants from 3 to 24 months with body composition assessed at three monthly intervals using deuterium dilution (DD) and anthropometry. Linear mixed modelling was utilized to generate sex-specific fat mass (FM) and fat-free mass (FFM) prediction equations, using length(m), weight-for-length (kg/m), triceps and subscapular skinfolds and South Asian ethnicity as variables. The study sample consisted of 1896 (942 measurements from 310 girls) training data sets, 941 (441 measurements from 154 girls) validation data sets of 3-24 months from Brazil, Pakistan, South Africa and Sri Lanka. The external validation group (test) comprised 349 measurements from 250 (185 from 124 girls) infants 3-6 months of age from South Africa, Australia and India.
Results: Sex-specific equations for three age categories (3-9 months; 10-18 months; 19-24 months) were developed, validated on same population and externally validated. Root mean squared error (RMSE) was similar between training, validation and test data for assessment of FM and FFM in boys and in girls. RMSPE and mean absolute percentage error (MAPE) were higher in validation compared to test data for predicting FM, however, in the assessment of FFM, both measures were lower in validation data. RMSE for test data from South Africa (M/F-0.46/0.45 kg) showed good agreement with validation data for assessment of FFM compared to Australia (M/F-0.51/0.33 kg) and India(M/F-0.77/0.80 kg).
Conclusions: Anthropometry-based FFM prediction equations provide acceptable results. Assessments based on equations developed on similar populations are more applicable than those developed from a different population.
(© 2024. The Author(s).)