Comparison of methods to assess change in children's body composition

Abstract

Background: Little is known about how simpler and more available methods to measure change in body fatness compare with criterion methods such as dual-energy X-ray absorptiometry (DXA) in children.

Objective: Our objective was to determine the ability of air-displacement plethysmography (ADP) and formulas based on triceps skinfold thickness (TSF) and bioelectrical impedance analysis (BIA) to estimate changes in body fat over time in children.

Design: Eighty-six nonoverweight and overweight boys (n = 34) and girls (n = 52) with an average age of 11.0 +/- 2.4 y underwent ADP, TSF measurement, BIA, and DXA to estimate body fatness at baseline and 1 +/- 0.3 y later. Recent equations were used to estimate percentage body fat by TSF measurement (Dezenberg equation) and by BIA (Suprasongsin and Lewy equations). Percentage body fat estimates by ADP, TSF measurement, and BIA were compared with those by DXA.

Results: All methods were highly correlated with DXA (P < 0.001). No mean bias for estimates of percentage body fat change was found for ADP (Siri equation) compared with DXA for all subjects examined together, and agreement between body fat estimation by ADP and DXA did not vary with race or sex. Magnitude bias was present for ADP relative to DXA (P < 0.01). Estimates of change in percentage body fat were systematically overestimated by BIA equations (1.37 +/- 6.98%; P < 0.001). TSF accounted for only 13% of the variance in percentage body fat change.

Conclusion: Compared with DXA, there appears to be no noninvasive and simple method to measure changes in children's percentage body fat accurately and precisely, but ADP performed better than did TSF or BIA. ADP could prove useful for measuring changes in adiposity in children.

Figures

FIGURE 1

Bland-Altman plots for magnitude bias in the estimation of percentage body fat (%BF) by dual-energy X-ray absorptiometry (DXA) at baseline by (A) air-displacement plethysmography (ADP) with use of the Siri equation (P = 0.02, R2 = 0.059, SEM = 0.027), (C) ADP with use of the Lohman adjustment of the Siri equation (NS, R2 = 0.001, SEM = 0.003), (E) skinfold-thickness measurements with use of the Dezenberg et al equation (P < 0.001, R2 = 0.38, SEM = 0.069), and (G) bioelectrical impedance analysis with use of the Lewy et al and Suprasongsin et al equations (P < 0.001, R2 = 0.14, SEM = 0.041), as well as at follow-up 1 y later by (B) ADP with use of the Siri equation (NS, R2 = 0.026, SEM = 0.018), (D) ADP with use of the Lohman adjustment of the Siri equation (NS, R2 = 1.3 × 10−5, SEM = 0.0004), (F) skinfold thicknesses with use of the Dezenberg et al equation (P < 0.001, R2 = 0.53, SEM = 0.081), and (H) bioelectrical impedance analysis with use of the Lewy et al and Suprasongsin et al equations (P < 0.01, R2 = 0.082, SEM = 0.032). The dashed lines indicate the mean differences, which was significantly different from zero for panels A, E, F, G, and H (described in text); the shaded area represents the limits of agreement (x̄ difference ± 2 SD); △, subjects scanned with use of the Hologic QDR2000; ○, subjects scanned with the Hologic QDR4500A.

FIGURE 2

Correlation plots of estimates of change in percentage body fat (%BF) by dual-energy X-ray absorptiometry (DXA) with estimates of change in %BF by (A) air-displacement plethysmography (ADP) with use of the Siri model (P < 0.001, R2 = 0.59, SEM = 0.084), (B) ADP with use of the Lohman age-specific model (P < 0.001, R2 = 0.57, SEM = 0.082), (C) bioelectrical impedance analysis (BIA) with use of the Lewy et al and Suprasongsin et al equations (P < 0.001, R2 = 0.44, SEM = 0.074), and (D) skinfold-thickness measurements with use of the Dezenberg et al equation (P < 0.001, R2 = 0.13, SEM = 0.040).

FIGURE 3

Bland-Altman plots for magnitude bias in the estimation of change in percentage body fat (%BF) by dual-energy X-ray absorptiometry (DXA) and (A) air-displacement plethysmography (ADP) with use of the Siri equation (P < 0.01, R2 = 0.12, SEM = 0.037), (B) ADP with use of the Lohman adjustment of the Siri equation (P < 0.01, R2 = 0.16, SEM = 0.043), (C) bioelectrical impedance with use of the Lewy et al and Suprasongsin et al equations (P < 0.001, R2 = 0.57, SEM = 0.084), and (D) skinfold-thickness measurements with use of the Dezenberg et al equation (P = NS, R2 = 0.002, SEM = 0.005). The dashed lines indicate the mean differences; the shaded area represents the limits of agreement (x̄ difference ± 2 SD).