Body composition among Sri Lankan infants by 18O dilution method and the validity of anthropometric equations to predict body fat against 18O dilution

BACKGROUND Body composition indicators provide a better guidance for growth and nutritional status of the infants. This study was designed to (1) measure the body composition of the Sri Lankan infants using a reference method, the (18)O dilution method; (2) calculate the body fat content of the infants using published skinfold prediction equations; and (3) evaluate the applicability of the skinfold equations to predict body fat among Sri Lankan infants against the (18)O dilution method. METHODS Twenty five healthy, exclusively breast-fed infants were randomly recruited at well-baby clinics, for this cross-sectional study. Body composition was measured using (18)O dilution. Infant body weight, length, skinfold thicknesses and mid upper-arm circumference were measured using standard procedures. The Bland and Atlman pair-wise comparison method was used to evaluate the agreement of body fat generated using the anthropometric prediction equations against the (18)O dilution values as the reference. RESULTS Mean (SD) body weight and length of the infants were 6.5 kg (0.9) and 64.7 cm (2.8) respectively. Mean total body water, fat free mass, fat mass and % fat mass as measured by (18)O dilution method were 58.8% (5.0), 4.6 kg (0.8), 1.9 (0.5) and 29.5% (6.1). Total body water and fat free mass were significantly higher in boys when compared to girls. With the exception of three prediction equations (Bandana et al., Goran et al. and Durnin and Wormsley), most of the other commonly used anthropometry-based prediction equations yielded a bias which was not constant but a function of the % fat mass. CONCLUSIONS Body composition of Sri Lankan infants is comparable to the normative data available from the industrialized countries. Most of the commonly used anthropometric prediction equations generated a bias which varies with the size of the body fat. Only three prediction equations (Bandana, Goran, Durnin & Wormsley) yield a constant bias. The Durnin & Wormsely equation showed the smallest bias when compared to the (18)O dilution values with the narrowest limits of agreement. Accuracy of some of the prediction equations is a function of gender.