Validation of selected body fatness methods against the deuterium technique in 6-8 year old black South Africans : the BC-IT Study

Abstract

Various scholars report overweight and obesity as major public health problems. Though an increase in obesity, especially among children, has been observed over decades, accurate body composition measurements remain a challenge. Despite the detected increase in obesity, methods to accurately predict body fat mass (FM), fat-free mass (FFM), and total body water (TBW) in children are limited. Body mass index (BMI), and body size perception is another measure of obesity. Therefore, complex measures such as bioelectrical impedance, deuterium dilution, under water weighing (UWW) and dual energy x-ray absorptiometry (DXA) are used for accurate predictions. The first aim of the thesis was to develop and validate the prediction equation for estimating FFM and TBW determined by available bioelectrical impedance analysis (BIA) prediction equations compared to deuterium dilution method (DDM). Secondly, the research aimed to assess the level of agreement between body image perception and actual body weight determined by body mass index (BMI) z-score and %BF measured by the (DDM). The third aim was to assess the level of agreement between percentage body fat (%BF) measured by BIA and skinfold thickness in South African children. A sub-study that followed a cross-sectional design on 299 children (predominately from the Tswana and Sotho ethnic groups) from the Body Composition using Isotopes Techniques study (BC-IT study) from the Dr JB Marks municipality (previously known as Tlokwe) of the North- West Province, South Africa. BMI z-scores were calculated relative to World Health Organization (WHO) reference data. The stable isotopes used the International Atomic Energy Agency (IAEA) protocol to analyse saliva samples using fourier transform infrared (FTIR) spectroscopy (FTIR 4500t spectrophotometer, Agilent), and BIA (Bodystat 1500®MDD) for TBW, FFM, % BF, and fat mass (FM). Body image perception was assessed by a simple questionnaire and silhouettes from Stunkard Figure rating scales (1983). Statistical analyses were performed using the statistical package for social sciences (SPSS v26.0). Paired t-test was used to determine the limits of agreement between %BF, FFM, and TBW obtained from BIA and the selected prediction equations, multiple regressions and Bland-Altman plots computed. Cohen kappa’s statistic (κ) determined the magnitude of agreement between perceived and actual body size. Kendall's tau-c assessed concordance among the children’s body image perception silhouette choice, BMI, and %BF determined by DDM. The level of significance was set at p≤0.05. The results from this thesis show that some published BIA equations for TBW and FFM validated in the study demonstrated under- or over-estimations. The newly developed South African equation had higher accuracy for both TBW and FFM prediction. When the newly developed and validated BIA prediction equations were plotted on Bland Altman plots, the results showed a good level of agreement between TBW and FFM against the criterion of the DDM.

When the level of agreement was assessed between the DDM and Silhouettes, the results indicated discrepancies between fatness based on body image perception compared to the BMI z-scores classification and DDM measurement. A sample of 202 children, (32.2%, 55.1%, 8.8%, and 2.4%), perceived their body size as underweight, normal, overweight, and obese, respectively. Based on the BMI z-score, (18.8%, 72.8%, 6.9%, and 1.5%) were classified as underweight, normal, overweight, and obese, respectively. Body fatness measurement through the DDM showed that (2.5%, 48.0%, 21.8%, and 29.7%) were underweight, normal weight, overweight, and obese, respectively. The level of agreement between body size perception, body fatness, and BMI z-score was poor (k=0.040, p=0.450). Levels of agreement differed between the BIA measurement and the three prediction equations. Significant mean differences (p<0.001) were found in the predicted %BF means determined by BIA, (Slaughter, Wickramasinghe and Dezenberg). The standard for evaluating prediction errors (SEE) was above 5. The magnitude of the regression lines varied from low to moderate (r2 = 0.46-0.61). The conclusion drawn from this study was that significant discrepancies between the various measurements compared to the criterion method were observed. Most of the prediction equations validated were not suitable for the sample of 6- to 8-year-old South African children. The newly developed equation demonstrates the potential for practical and accurate estimation of TBW and FFM for use in South African children aged 6- to 8-years. Still, prediction equations probably need to be population specific.