Journal of Human Nutrition and Dietetics
RESEARCH PAPER A comparison of body composition measurement techniques S. E. Hillier, L. Beck A. Petropoulou & M. E. Clegg Functional Food Centre, Oxford Brookes University, Oxford, UK
Keywords air displacement plethysmography, bioelectrical impedance, body fat percentage, reproducibility, skinfold thickness. Correspondence M. E. Clegg, Functional Food Centre, Faculty of Health and Life Sciences, Oxford Brookes University, Oxford OX3 0BP, UK. Tel.: +44 1865 484365 Fax: +44 1865 483618 E-mail: [email protected] How to cite this article Hillier S.E., Beck L., Petropoulou A. & Clegg M.E (2014) A comparison of body composition measurement techniques. J Hum Nutr Diet. doi:10.1111/jhn.12197
Abstract Background: An understanding of the reproducibility of body composition measurements is essential for effective intervention studies. Air displacement plethysmography (ADP) and bioelectrical impedance (BIA) are two easy-touse measures of body composition. The present study aimed to assess the reproducibility of ADP and BIA and compare them with each other, as well as with skinfold measurement. Methods: Forty-one participants were tested on two occasions following an overnight fast. On test day 1, participants’ height, weight and % body fat (BF) were measured using ADP and BIA. Measurements were replicated to assess the within-day reproducibility. On test day 2, participants were again tested using ADP and BIA and had skinfold measurements taken. Three skinfold equations for BF calculation were applied. Comparisons of withinand between-day reproducibility and between measurement techniques were completed using Pearson correlations and Bland–Altman analysis. Results: Both Pearson correlation and Bland–Altman analysis showed good within- and between-day relationships and agreement for BF from ADP and BIA measurements. The two methods had a high correlation between them; however, the mean difference between the two was 3.1% (4.1%). From the skinfold equations used, the best agreement with ADP had a mean difference of 0.3% (0.8%) and, with BIA, had mean differences of 1.9% (4.2%). Conclusions: The data indicate that ADP and BIA cannot be used interchangeably, although both measurements had good within- and betweenday agreement.
Introduction An accurate assessment of body composition is necessary to correctly identify the health risks associated with excessively low or high body fat. It is also required to determine the potential changes of an intervention over time, such as weight loss, weight gain or exercise. To effectively evaluate body composition, the methods are not only required to be precise and consistent, but also simple to use, practical and accommodating to a wide range of potential individuals. As a result, determining the reproducibility of different body composition methods is essential. A number of different tools and methodologies have been developed to measure body composition in both a ª 2014 The British Dietetic Association Ltd.
clinical and field setting. The most popular clinical techniques include hydro densitometry (HW), air displacement plethysmography (ADP) and dual-energy X-ray absorptiometry. The most popular field techniques include bioelectrical impedance analysis (BIA), skinfolds (SF) and anthropometry (Wagner & Heyward, 1999). The previous ‘gold standard’ method for determining body composition and volume was HW. However, for the last 20 years, ADP and BIA, two measures of body composition, have become increasingly popular as a result of their ease of use and greater practical advantages (Fields et al., 2002). The BOD POD body composition system was developed to reduce the inconsistent validity and reliability reported of HW methodology (Dempster & Aitkens, 1995). To determine its long-term use in any given 1
Body composition measurement
population, the validity and reproducibility needs to be established. Previous research has investigated the reliability of ADP within-day and reports that ADP retains excellent precision but recommends repeating measurements to allow for erroneous volume measurements within one procedure (Collins & McCarthy, 2003). Other studies have examined the reliability between different test days (Levenhagen et al., 1999; Miyatake et al., 1999; Demerath et al., 2002). In addition, investigations comparing ADP with other established techniques, both in adults (Biaggi et al., 1999; Wagner & Heyward, 1999; Ball & Altena, 2004; Frisard et al., 2005; Kilduff et al., 2007) and children (Fields & Goran, 2000; Lockner et al., 2000; Ball & Altena, 2004), report good agreement between the methods. However, some studies have reported biased results in favour of the BOD POD (Collins et al., 1999; Demerath et al., 2002; Radley et al., 2003). The validity of all body composition methods is imperative; however, in many situations, the reliability is equally important. This is particularly evident when investigating changes in body composition over time, during and following an intervention. The present study therefore aimed to assess the reproducibility of percentage body fat measurements using ADP and BIA both within and between days. The study also aimed to compare ADP and BIA measurements with each other, as well as with skinfold measurements, to establish the best agreement between methods. Materials and methods Forty-one healthy participants, 19 men, aged 23 (19–29) years, BMI 21.4 (19–27) kg/m2, and 22 women, aged 23 (19–29) years, BMI 22.9 (19–29) kg/m2, were recruited to the study. Participants comprised staff and students from Oxford Brookes University and were considered healthy following the completion of a health questionnaire. Participants were excluded if they reported a BMI ≥30 kg/m2 or were aged ≤18 years or ≥50 years. Ethical approval was obtained from the University Research Ethics Committee at Oxford Brookes University in accordance with the guidelines laid down in the Declaration of Helsinki. Written informed consent was obtained from each participant before the study began. The study was conducted over two separate testing days (≤7 days). Participants reported to the laboratory following a 10–12 h rest. Height, weight and % body fat (BF) were measured using ADP (BOD POD; COSMED, Rome, Italy) and BIA (BC-418 MA Segmental Body Composition Analyser Class III; Tanita, Yiewsley, UK). Measurements were replicated to assess the within-day reproducibility of both ADP and BIA. On the second test 2
S. E. Hillier et al.
day, ADP and BIA measurements were repeated, with additional skinfold measurements taken. Stature was measured using a wall-mounted stadiometer (Seca, Birmingham, UK) to the nearest 0.1 cm with the head positioned in the ‘Frankfurt plane’. Body mass was measured to the nearest 0.1 kg using calibrated plethysmographic weighing scales. The participants were dressed in a skin-tight swimsuit and a swimming cap for all measurements. Skinfold thickness at eight sites was measured using standardised Harpenden skinfold callipers. The sites included subscapular, tricep, bicep, iliac crest, supraspinale, abdominal, front thigh and medial calf. Each measurement was taken twice unless the two readings differed by ≥10%, in which case a third was taken. Skinfolds were then averaged to represent skinfold thickness. Measurements were taken by one of two trained researchers, and the same researcher was used for all skinfold measurements on a single subject. Percentage body fat was calculated using three prediction equations (Durnin & Womersley, 1974; Peterson et al., 2003), including both males (Jackson & Pollock, 1978) and females (Jackson et al., 1980). Bioelectrical impedance analysis was completed in accordance with the manufacturer’s instructions. The participants were dressed in a skin-tight swimsuit and a swimming cap for all measurements. Participant height, age, sex and level of activity were entered into the machine. Height was entered to the nearest 1 cm. Level of activity was entered in accordance with the manufacturer’s instructions. If the person participated in ≥10 h intense exercise per week, they were categorised as ‘Athletic’; 150 mL, a third test was conducted. Body density (Db) was calculated by the BOD POD system software using the equation Db = weight/Vb. Predicted thoracic gas ª 2014 The British Dietetic Association Ltd.
S. E. Hillier et al.
Body composition measurement
volume was used within the calculations. Db is then converted into an estimate of body composition using the Siri equation. All calculations were completed in their entirety using the automated system software. Measures of the repeatability of both ADP and BIA were obtained from all participants, both within-day and between-days. Test–retest reliability was measured during test day 1 for both ADP and BIA. Between test reliability was calculated by comparing test day 1 measurements with test day 2 (≤7 days) for both ADP and BIA. A comparison of three skinfold equations was conducted to determine the best agreement with ADP and BIA. Data analysis was performed using SPSS, version 19.0 (IBM Corp., Armonk, NY, USA). Data are expressed as the mean (SD) unless otherwise stated. Pearson’s correlation coefficient was performed to assess the strength of relationships for both within- and between-day measurements of ADP and BIA. Bland–Altman analysis (Bland & Altman, 1986) was implemented to determine levels of agreement of ADP and BIA within themselves both within- and between-days and between each other and skinfolds on day 2. The differences between the means were calculated. For comparison between the methods, differences were calculated as ADP/BIA minus skinfold measurement. For comparison between the skinfold equations, differences are indicated in the results. Confidence limits of agreement were set at 95%. P < 0.05 was considered statistically significant. Results Within-day reproducibility Pearson correlation showed a significant relationship for within- day measurements of % BF using ADP and BIA measurement equipment (R = 0.977 and 0.999, respectively; P < 0.001 for both0 (Table 1). Bland–Altman analysis also indicated agreement within-day for ADP and BIA measurements of % BF. For ADP, the within-day mean difference was 0.3% (1.7%) and, for BIA, the difference was 0.1% (0.4%) within-day. There were similar
Table 1 Within-day and between-day correlations (R) and mean difference in percentage of body fat (% BF) for air displacement plethysmography (ADP) and bioelectrical impedance (BIA) Within-day
ADP BIA
correlations for males and females for the ADP (males: R = 0.959; females: R = 0.947; P < 0.001 for both) but with slightly larger mean differences for the males [males: 0.5% BF (1.6% BF); females: 0.1% BF (1.9% BF)]. In the BIA test, there were also similar correlations for males and females (males: R = 0.995; females: R = 0.997; P < 0.001 for both) and very similar mean differences [males: 0.1% BF (0.5% BF); females: 0.1% BF (0.4% BF)]. Between-day reproducibility The between-day correlations were significant for both ADP and BIA (R = 0.965 and 0.977, respectively; P < 0.001 for both). For ADP between-day differences in % BF were 0.1% (2.1%) and, for BIA, they were 0.5% (1.8%), indicating larger between-day differences for BIA. Between-day correlations were greater for females than for males for the ADP (males: R = 0.891; females: R = 0.959; P < 0.001 for both) but the mean differences were similar [males: (0.2% BF (2.6% BF); females: 0.0% BF (1.6% BF)]. In the BIA test, there were also similar between-day correlations for males and females (males: R = 0.946; females: R = 0.917; P < 0.001 for both) and similar mean differences [males: 0.3% BF (1.6% BF); females: 0.7% BF (2.0% BF)]. Comparison between methods There were highly significant correlations between all methods used to measure % BF – BIA, ADP and skinfold thickness using three sets of equations (P < 0.001). The mean difference between BIA and ADP was 3.1% (4.1%) (R = 0.886) (Fig. 1) with the ADP giving readings 3.1% lower than BIA. The best skinfold equation agreement with ADP had a mean difference of 0.3% (2.8%) (R = 0.930) (Jackson & Pollock, 1978; Jackson et al., 1980) compared to 5.4% (3.6%) (R = 0.889) (Durnin & Womersley, 1974) and 6.1% (3.5%) (R = 0.908) (Peterson et al., 2003). The BIA had mean differences with the skinfold measurements of 1.9% (4.2%) (R = 0.902) (Durnin & Womersley, 1974), 3.8% (4.0%) (R = 0.902) (Jackson & Pollock, 1978; Jackson et al., 1980) and 2.7% (3.8%) (R = 0.918) (Peterson et al., 2003).
Between-day
Pearson correlation (R)
Mean difference (% BF)
Pearson correlation (R)
Mean difference (% BF)
0.977* 0.999*
0.3 (1.7) 0.1 (0.4)
0.965* 0.977*
0.1 (2.1) 0.5 (1.8)
*P < 0.001: mean difference data are presented as the mean (SD).
ª 2014 The British Dietetic Association Ltd.
Comparisons between skinfold equations Between the skinfolds the best agreement was seen between Durnin & Womersley (1974) and Peterson et al. (2003) with a mean difference of 0.8% (1.8%) (former minus latter). The difference between the other skinfold measurements was 5.7% (1.7%) between Durnin & 3
S. E. Hillier et al.
Body composition measurement
Difference ADP -BIA (% Body fat)
10 8 6 4 2 0 –2
0
10
20
30
40
50
–4 –6
Figure 1 Bland–Altman plot comparing air displacement plethysmography (ADP) and bioelectrical impedance (BIA) measurements within day.
–8 –10 –12
Average ADP + BIA (% Body fat)
Table 2 Percentage difference between air displacement plethysmography (ADP), bioelectrical impedance (BIA) and skinfold measurements of percentage of body fat for each sex BIA Female ADP BIA Skinfold* Skinfold†,‡
Male
3.1 (4.0) – – –
Skinfold†‡
Skinfold* Female
1.6 (4.6) – – –
Male
4.4 (3.3) 0.5 (3.7) – –
Female
6.5 (3.6) 4.8 (2.7) – –
0.7 (2.8) 5.7 (4.0) 5.1 (1.2) –
Skinfold§ Male
Female
0.2 (2.7) 1.6 (2.8) 6.4 (2.0) –
5.7 0.7 1.3 6.4
(3.6) (3.6) (1.8) (1.8)
Male 6.7 4.9 0.1 6.5
(3.5) (2.6) (1.8) (1.5)
All values are column minus row and the mean (SD). *Durnin & Womersley (1974). † Jackson & Pollock (1978). ‡ Jackson et al. (1980). § Peterson et al. (2003).
Womersley (1974) and Jackson and Pollock (Jackson & Pollock, 1978; Jackson et al., 1980) and 6.5% (1.6%) between Jackson and Pollock (Jackson & Pollock, 1978; Jackson et al., 1980) and Peterson et al.(2003) (former minus latter). The difference between the methods for each sex is shown in Table 2. Discussion The present study aimed assess the within- and betweenday reproducibility of ADP and BIA in healthy young individuals and to compare them with each other, as well as with skinfold measurements. The results demonstrated that ADP and BIA measurements have good within-day and between-day reproducibility. The two methods showed a positive correlation between them; however, a mean difference of 3.1% body fat between methods was reported. The results also demonstrate highly significant correlations between all of the methods used to measure body fat percentage (BIA, ADP and skinfold thickness using three sets of equations). The results of the present study are in agreement with those reported previously, which indicate good reliability 4
between different methods (Biaggi et al., 1999; Wagner & Heyward, 1999; Collins & McCarthy, 2003; Frisard et al., 2005; Kilduff et al., 2007). The present study also displays data from participants with a BMI
RESEARCH PAPER A comparison of body composition measurement techniques S. E. Hillier, L. Beck A. Petropoulou & M. E. Clegg Functional Food Centre, Oxford Brookes University, Oxford, UK
Keywords air displacement plethysmography, bioelectrical impedance, body fat percentage, reproducibility, skinfold thickness. Correspondence M. E. Clegg, Functional Food Centre, Faculty of Health and Life Sciences, Oxford Brookes University, Oxford OX3 0BP, UK. Tel.: +44 1865 484365 Fax: +44 1865 483618 E-mail: [email protected] How to cite this article Hillier S.E., Beck L., Petropoulou A. & Clegg M.E (2014) A comparison of body composition measurement techniques. J Hum Nutr Diet. doi:10.1111/jhn.12197
Abstract Background: An understanding of the reproducibility of body composition measurements is essential for effective intervention studies. Air displacement plethysmography (ADP) and bioelectrical impedance (BIA) are two easy-touse measures of body composition. The present study aimed to assess the reproducibility of ADP and BIA and compare them with each other, as well as with skinfold measurement. Methods: Forty-one participants were tested on two occasions following an overnight fast. On test day 1, participants’ height, weight and % body fat (BF) were measured using ADP and BIA. Measurements were replicated to assess the within-day reproducibility. On test day 2, participants were again tested using ADP and BIA and had skinfold measurements taken. Three skinfold equations for BF calculation were applied. Comparisons of withinand between-day reproducibility and between measurement techniques were completed using Pearson correlations and Bland–Altman analysis. Results: Both Pearson correlation and Bland–Altman analysis showed good within- and between-day relationships and agreement for BF from ADP and BIA measurements. The two methods had a high correlation between them; however, the mean difference between the two was 3.1% (4.1%). From the skinfold equations used, the best agreement with ADP had a mean difference of 0.3% (0.8%) and, with BIA, had mean differences of 1.9% (4.2%). Conclusions: The data indicate that ADP and BIA cannot be used interchangeably, although both measurements had good within- and betweenday agreement.
Introduction An accurate assessment of body composition is necessary to correctly identify the health risks associated with excessively low or high body fat. It is also required to determine the potential changes of an intervention over time, such as weight loss, weight gain or exercise. To effectively evaluate body composition, the methods are not only required to be precise and consistent, but also simple to use, practical and accommodating to a wide range of potential individuals. As a result, determining the reproducibility of different body composition methods is essential. A number of different tools and methodologies have been developed to measure body composition in both a ª 2014 The British Dietetic Association Ltd.
clinical and field setting. The most popular clinical techniques include hydro densitometry (HW), air displacement plethysmography (ADP) and dual-energy X-ray absorptiometry. The most popular field techniques include bioelectrical impedance analysis (BIA), skinfolds (SF) and anthropometry (Wagner & Heyward, 1999). The previous ‘gold standard’ method for determining body composition and volume was HW. However, for the last 20 years, ADP and BIA, two measures of body composition, have become increasingly popular as a result of their ease of use and greater practical advantages (Fields et al., 2002). The BOD POD body composition system was developed to reduce the inconsistent validity and reliability reported of HW methodology (Dempster & Aitkens, 1995). To determine its long-term use in any given 1
Body composition measurement
population, the validity and reproducibility needs to be established. Previous research has investigated the reliability of ADP within-day and reports that ADP retains excellent precision but recommends repeating measurements to allow for erroneous volume measurements within one procedure (Collins & McCarthy, 2003). Other studies have examined the reliability between different test days (Levenhagen et al., 1999; Miyatake et al., 1999; Demerath et al., 2002). In addition, investigations comparing ADP with other established techniques, both in adults (Biaggi et al., 1999; Wagner & Heyward, 1999; Ball & Altena, 2004; Frisard et al., 2005; Kilduff et al., 2007) and children (Fields & Goran, 2000; Lockner et al., 2000; Ball & Altena, 2004), report good agreement between the methods. However, some studies have reported biased results in favour of the BOD POD (Collins et al., 1999; Demerath et al., 2002; Radley et al., 2003). The validity of all body composition methods is imperative; however, in many situations, the reliability is equally important. This is particularly evident when investigating changes in body composition over time, during and following an intervention. The present study therefore aimed to assess the reproducibility of percentage body fat measurements using ADP and BIA both within and between days. The study also aimed to compare ADP and BIA measurements with each other, as well as with skinfold measurements, to establish the best agreement between methods. Materials and methods Forty-one healthy participants, 19 men, aged 23 (19–29) years, BMI 21.4 (19–27) kg/m2, and 22 women, aged 23 (19–29) years, BMI 22.9 (19–29) kg/m2, were recruited to the study. Participants comprised staff and students from Oxford Brookes University and were considered healthy following the completion of a health questionnaire. Participants were excluded if they reported a BMI ≥30 kg/m2 or were aged ≤18 years or ≥50 years. Ethical approval was obtained from the University Research Ethics Committee at Oxford Brookes University in accordance with the guidelines laid down in the Declaration of Helsinki. Written informed consent was obtained from each participant before the study began. The study was conducted over two separate testing days (≤7 days). Participants reported to the laboratory following a 10–12 h rest. Height, weight and % body fat (BF) were measured using ADP (BOD POD; COSMED, Rome, Italy) and BIA (BC-418 MA Segmental Body Composition Analyser Class III; Tanita, Yiewsley, UK). Measurements were replicated to assess the within-day reproducibility of both ADP and BIA. On the second test 2
S. E. Hillier et al.
day, ADP and BIA measurements were repeated, with additional skinfold measurements taken. Stature was measured using a wall-mounted stadiometer (Seca, Birmingham, UK) to the nearest 0.1 cm with the head positioned in the ‘Frankfurt plane’. Body mass was measured to the nearest 0.1 kg using calibrated plethysmographic weighing scales. The participants were dressed in a skin-tight swimsuit and a swimming cap for all measurements. Skinfold thickness at eight sites was measured using standardised Harpenden skinfold callipers. The sites included subscapular, tricep, bicep, iliac crest, supraspinale, abdominal, front thigh and medial calf. Each measurement was taken twice unless the two readings differed by ≥10%, in which case a third was taken. Skinfolds were then averaged to represent skinfold thickness. Measurements were taken by one of two trained researchers, and the same researcher was used for all skinfold measurements on a single subject. Percentage body fat was calculated using three prediction equations (Durnin & Womersley, 1974; Peterson et al., 2003), including both males (Jackson & Pollock, 1978) and females (Jackson et al., 1980). Bioelectrical impedance analysis was completed in accordance with the manufacturer’s instructions. The participants were dressed in a skin-tight swimsuit and a swimming cap for all measurements. Participant height, age, sex and level of activity were entered into the machine. Height was entered to the nearest 1 cm. Level of activity was entered in accordance with the manufacturer’s instructions. If the person participated in ≥10 h intense exercise per week, they were categorised as ‘Athletic’; 150 mL, a third test was conducted. Body density (Db) was calculated by the BOD POD system software using the equation Db = weight/Vb. Predicted thoracic gas ª 2014 The British Dietetic Association Ltd.
S. E. Hillier et al.
Body composition measurement
volume was used within the calculations. Db is then converted into an estimate of body composition using the Siri equation. All calculations were completed in their entirety using the automated system software. Measures of the repeatability of both ADP and BIA were obtained from all participants, both within-day and between-days. Test–retest reliability was measured during test day 1 for both ADP and BIA. Between test reliability was calculated by comparing test day 1 measurements with test day 2 (≤7 days) for both ADP and BIA. A comparison of three skinfold equations was conducted to determine the best agreement with ADP and BIA. Data analysis was performed using SPSS, version 19.0 (IBM Corp., Armonk, NY, USA). Data are expressed as the mean (SD) unless otherwise stated. Pearson’s correlation coefficient was performed to assess the strength of relationships for both within- and between-day measurements of ADP and BIA. Bland–Altman analysis (Bland & Altman, 1986) was implemented to determine levels of agreement of ADP and BIA within themselves both within- and between-days and between each other and skinfolds on day 2. The differences between the means were calculated. For comparison between the methods, differences were calculated as ADP/BIA minus skinfold measurement. For comparison between the skinfold equations, differences are indicated in the results. Confidence limits of agreement were set at 95%. P < 0.05 was considered statistically significant. Results Within-day reproducibility Pearson correlation showed a significant relationship for within- day measurements of % BF using ADP and BIA measurement equipment (R = 0.977 and 0.999, respectively; P < 0.001 for both0 (Table 1). Bland–Altman analysis also indicated agreement within-day for ADP and BIA measurements of % BF. For ADP, the within-day mean difference was 0.3% (1.7%) and, for BIA, the difference was 0.1% (0.4%) within-day. There were similar
Table 1 Within-day and between-day correlations (R) and mean difference in percentage of body fat (% BF) for air displacement plethysmography (ADP) and bioelectrical impedance (BIA) Within-day
ADP BIA
correlations for males and females for the ADP (males: R = 0.959; females: R = 0.947; P < 0.001 for both) but with slightly larger mean differences for the males [males: 0.5% BF (1.6% BF); females: 0.1% BF (1.9% BF)]. In the BIA test, there were also similar correlations for males and females (males: R = 0.995; females: R = 0.997; P < 0.001 for both) and very similar mean differences [males: 0.1% BF (0.5% BF); females: 0.1% BF (0.4% BF)]. Between-day reproducibility The between-day correlations were significant for both ADP and BIA (R = 0.965 and 0.977, respectively; P < 0.001 for both). For ADP between-day differences in % BF were 0.1% (2.1%) and, for BIA, they were 0.5% (1.8%), indicating larger between-day differences for BIA. Between-day correlations were greater for females than for males for the ADP (males: R = 0.891; females: R = 0.959; P < 0.001 for both) but the mean differences were similar [males: (0.2% BF (2.6% BF); females: 0.0% BF (1.6% BF)]. In the BIA test, there were also similar between-day correlations for males and females (males: R = 0.946; females: R = 0.917; P < 0.001 for both) and similar mean differences [males: 0.3% BF (1.6% BF); females: 0.7% BF (2.0% BF)]. Comparison between methods There were highly significant correlations between all methods used to measure % BF – BIA, ADP and skinfold thickness using three sets of equations (P < 0.001). The mean difference between BIA and ADP was 3.1% (4.1%) (R = 0.886) (Fig. 1) with the ADP giving readings 3.1% lower than BIA. The best skinfold equation agreement with ADP had a mean difference of 0.3% (2.8%) (R = 0.930) (Jackson & Pollock, 1978; Jackson et al., 1980) compared to 5.4% (3.6%) (R = 0.889) (Durnin & Womersley, 1974) and 6.1% (3.5%) (R = 0.908) (Peterson et al., 2003). The BIA had mean differences with the skinfold measurements of 1.9% (4.2%) (R = 0.902) (Durnin & Womersley, 1974), 3.8% (4.0%) (R = 0.902) (Jackson & Pollock, 1978; Jackson et al., 1980) and 2.7% (3.8%) (R = 0.918) (Peterson et al., 2003).
Between-day
Pearson correlation (R)
Mean difference (% BF)
Pearson correlation (R)
Mean difference (% BF)
0.977* 0.999*
0.3 (1.7) 0.1 (0.4)
0.965* 0.977*
0.1 (2.1) 0.5 (1.8)
*P < 0.001: mean difference data are presented as the mean (SD).
ª 2014 The British Dietetic Association Ltd.
Comparisons between skinfold equations Between the skinfolds the best agreement was seen between Durnin & Womersley (1974) and Peterson et al. (2003) with a mean difference of 0.8% (1.8%) (former minus latter). The difference between the other skinfold measurements was 5.7% (1.7%) between Durnin & 3
S. E. Hillier et al.
Body composition measurement
Difference ADP -BIA (% Body fat)
10 8 6 4 2 0 –2
0
10
20
30
40
50
–4 –6
Figure 1 Bland–Altman plot comparing air displacement plethysmography (ADP) and bioelectrical impedance (BIA) measurements within day.
–8 –10 –12
Average ADP + BIA (% Body fat)
Table 2 Percentage difference between air displacement plethysmography (ADP), bioelectrical impedance (BIA) and skinfold measurements of percentage of body fat for each sex BIA Female ADP BIA Skinfold* Skinfold†,‡
Male
3.1 (4.0) – – –
Skinfold†‡
Skinfold* Female
1.6 (4.6) – – –
Male
4.4 (3.3) 0.5 (3.7) – –
Female
6.5 (3.6) 4.8 (2.7) – –
0.7 (2.8) 5.7 (4.0) 5.1 (1.2) –
Skinfold§ Male
Female
0.2 (2.7) 1.6 (2.8) 6.4 (2.0) –
5.7 0.7 1.3 6.4
(3.6) (3.6) (1.8) (1.8)
Male 6.7 4.9 0.1 6.5
(3.5) (2.6) (1.8) (1.5)
All values are column minus row and the mean (SD). *Durnin & Womersley (1974). † Jackson & Pollock (1978). ‡ Jackson et al. (1980). § Peterson et al. (2003).
Womersley (1974) and Jackson and Pollock (Jackson & Pollock, 1978; Jackson et al., 1980) and 6.5% (1.6%) between Jackson and Pollock (Jackson & Pollock, 1978; Jackson et al., 1980) and Peterson et al.(2003) (former minus latter). The difference between the methods for each sex is shown in Table 2. Discussion The present study aimed assess the within- and betweenday reproducibility of ADP and BIA in healthy young individuals and to compare them with each other, as well as with skinfold measurements. The results demonstrated that ADP and BIA measurements have good within-day and between-day reproducibility. The two methods showed a positive correlation between them; however, a mean difference of 3.1% body fat between methods was reported. The results also demonstrate highly significant correlations between all of the methods used to measure body fat percentage (BIA, ADP and skinfold thickness using three sets of equations). The results of the present study are in agreement with those reported previously, which indicate good reliability 4
between different methods (Biaggi et al., 1999; Wagner & Heyward, 1999; Collins & McCarthy, 2003; Frisard et al., 2005; Kilduff et al., 2007). The present study also displays data from participants with a BMI
