Abstract
BACKGROUND/OBJECTIVES: Fat-mass (FM) assessment since birth using valid methodologies is crucial since excessive adiposity represents a risk factor for adverse metabolic outcomes. AIM: To develop infant FM prediction equations using anthropometry and validate them against air-displacement plethysmography (ADP). SUBJECTS/METHODS: Clinical, anthropometric (weight, length, body-mass index -BMI-, circumferences, and skinfolds), and FM (ADP) data were collected from healthy-term infants at 1 (n = 133), 3 (n = 105), and 6 (n = 101) months enrolled in the OBESO perinatal cohort (Mexico City). FM prediction models were developed in 3 steps: 1) Variable Selection (LASSO regression), 2) Model behavior evaluation (12-fold cross-validation, using Theil-Sen regressions), and 3) Final model evaluation (Bland-Altman plots, Deming regression). RESULTS: Relevant variables in the FM prediction models included BMI, circumferences (waist, thigh, and calf), and skinfolds (waist, triceps, subscapular, thigh, and calf). The R(2) of each model was 1 M: 0.54, 3 M: 0.69, 6 M: 0.63. Predicted FM showed high correlation values (r ≥ 0.73, p < 0.001) with FM measured with ADP. There were no significant differences between predicted vs measured FM (1 M: 0.62 vs 0.6; 3 M: 1.2 vs 1.35; 6 M: 1.65 vs 1.76 kg; p > 0.05). Bias were: 1 M -0.021 (95%CI: -0.050 to 0.008), 3 M: 0.014 (95%CI: 0.090-0.195), 6 M: 0.108 (95%CI: 0.046-0.169). CONCLUSION: Anthropometry-based prediction equations are inexpensive and represent a more accessible method to estimate body composition. The proposed equations are useful for evaluating FM in Mexican infants.