Figure 1.
Conceptual illustration for deriving the ‘best’ BIA equation.
Calibration equations aim to describe the relationship between fat-free mass (FFM) and height2/impedance (HT2/Z) or impedance (Z). Three types of calibration equations were investigated: (i) a generic equation with no ethnic terms; (ii) a simple ethnic-specific equation (same slope, different intercept); (iii) a complex ethnic-specific equation with ethnic-impedance interaction terms (different slopes, different intercepts). Model can have very similar average residuals but the distribution of residuals may vary across ethnic groups (iv). To avoid ethnic bias, models with even residual distribution (v) are preferred.
Table 1.
Background characteristics of the study populations.
Figure 2.
Ethnic differences in anthropometric measurements in boys.
Dots represent mean values, error bars represent standard error of the mean. Black African/Caribbean boys generally had higher mean values of all circumferences and standing height, but lower sitting/standing height ratio, at each age group as compared to White and South Asian children. An increasing trend for ethnic differences in body circumferences is evident, especially after 9 years of age. Despite its heterogeneity, the ‘Other’ group tend to track White and South Asian group more closely than Black African/Caribbean group. Similar pattern was observed in girls (Figure S3). Due to relatively small sample size >10 years old (n = 12) the estimates may be biased therefore the trend lines from 10 to 11 years old were replaced by dotted lines.
Figure 3.
Hattori's body composition chart according to sex.
Regression line for each ethnic group is shown with individual data points removed. Regression line is shown from fat-free mass index (FFMI) 10 to 18 due to fewer individuals at both ends of the regression line. Error bars show standard error of the mean. There were no significant ethnic differences in fat-mass index (FMI) for either boys or girls. In both boys and girls, FFMI was significantly higher in Black African/Caribbean children than in White (boys: p<0.001, girls: p<0.001), South Asian (boys: p<0.001, girls: p<0.001) and ‘Other’ (boys: p<0.01, girls: p<0.05) children.
Table 2.
Models' selection criteria.
Figure 4.
Residuals distribution for each ethnic group for model A2.
Residuals from model A2 were regressed, for each ethnic group, on height2/impedance as indicated by the bold regression line. The shaded area represents a 95% confidence interval for the mean of residuals as a function of height2/impedance. Correlation coeffients (95% confidence interval) are also shown. No mean, and no slope, differed significantly from zero.