Body Composition at 6 months of Life: Comparison Of Air Displacement Plethysmography and Dual-Energy X-Ray Absorptiometry




Body composition assessment during infancy is important because it is a critical period for obesity risk development, thus valid tools are needed to accurately, precisely, and quickly determine both fat and fat-free mass. The purpose of this study was to compare body composition estimates using dual-energy x-ray absorptiometry (DXA) and air displacement plethysmography (ADP) at 6 months old. We assessed the agreement between whole body composition using DXA and ADP in 84 full-term average-for-gestational-age boys and girls using DXA (Lunar iDXA v11–30.062; Infant whole body analysis enCore 2007 software, GE, Fairfield, CT) and ADP (Infant Body Composition System v3.1.0, COSMED USA, Concord, CA). Although the correlations between DXA and ADP for %fat (r = 0.925), absolute fat mass (r = 0.969), and absolute fat-free mass (r = 0.945) were all significant, body composition estimates by DXA were greater for both %fat (31.1 ± 3.6% vs. 26.7 ± 4.7%; P < 0.001) and absolute fat mass (2,284 ± 449 vs. 1,921 ± 492 g; P < 0.001), and lower for fat-free mass (5,022 ± 532 vs. 5,188 ± 508 g; P < 0.001) vs. ADP. Inter-method differences in %fat decreased with increasing adiposity and differences in fat-free mass decreased with increasing infant age. Estimates of body composition determined by DXA and ADP at 6 months of age were highly correlated, but did differ significantly. Additional work is required to identify the technical basis for these rather large inter-method differences in infant body composition.


In recent years, a number of prenatal and early postnatal factors have been associated with lifetime risk for obesity. Consequently, there is an increasing interest in monitoring early life growth of children. Both fat and fat-free mass are increasing rapidly during infancy, but the majority of studies have relied upon serial measurement of weight, length, and head circumference to assess growth, with very little information regarding the composition of weight gain over time. Because absolute weight change has limited utility in the identification of infants at risk of later adverse health outcomes that stem from elevated adiposity, and/or reduced fat-free mass, body composition assessment techniques that are both accurate and precise are greatly needed in evaluating the quality of weight gain during this key period of life.

Recently, dual X-ray absorptiometry (DXA) (1) and air displacement plethysmography (ADP) (2) have been used to accurately and noninvasively measure changes in fat-free mass and fat mass in children (3). ADP i.e., Pea Pod (4) has also been developed to measure body composition in infants from birth to 6 months of age (1–8 kg) (5). The ADP in infants is accurate, quick, and noninvasive, thus providing new opportunities for studying infant body composition (5).

A comparison of these two methods is important because they are the most commonly used methods today for infant body composition assessment, and are less costly, faster, and less invasive than the criterion four-compartment model. Nonetheless, each has limitations that may persuade users to choose one over the other. Although DXA accurately measures whole and regional body composition, concerns remain regarding radiation exposure (though extremely low) and the necessity to keep the subject still during the scan, thus subject compliance is problematic. If DXA and ADP provide similar estimates of body composition, this allows for ADP to be used where radiation or participant movement is of particular concern. The main limitations of ADP include the lack of regional estimates of body composition and bone mass or density, and the inability to measure infants larger than ∼8 kg. The upper weight limit of ADP poses particular difficulty in longitudinal studies of infant body composition, as most children will exceed this limit by approximately 6 months of age. Therefore, a change in instrumentation from ADP to DXA may have to be considered.

The purpose of this study was to compare body composition using DXA to that using ADP at 6 months of life in order to examine whether the measurement of fat and fat-free mass are comparable between the two methods. To our knowledge, this is the first study to compare the two methods in the first months of life.


Eight-four healthy full-term boys (n = 37) and girls (n = 47) participated in the study with pertinent birth data by group as follows: (gestational age 39.2 ± 1.0 weeks; birth weight 3,442 ± 369 g; and birth recumbent length 51.1 ± 2.5 cm). Subjects were recruited from various growth studies being conducted at the University of Oklahoma Health Sciences. Inclusion into the study required a mother between the ages of 18–45 years at the time of delivery, a term pregnancy lasting ≥37 weeks, singleton birth, and a hospital stay for the infant of less than 3 days following delivery. Exclusion criteria were tobacco use or alcohol consumption (>1 drink per week) during pregnancy, pre-gestational, or gestational diabetes, and infants with presumed or known congenital birth defects.

All testing took place on the same day in the OU Children's Physicians Building located at the University of Oklahoma Health Sciences Center. The Institutional Review Board approved all procedures for Human Participants at the University of Oklahoma Health Sciences. Before testing, all mothers signed an informed consent and a Health Insurance Portability and Accountability Act authorization.

Assessment of body composition

ADP. Infant body composition was obtained by ADP using the Pea Pod Body Composition System (COSMED USA, Concord, CA). ADP testing and procedures have been described previously (5,6). Crown-to-heel length was obtained using a Seca 416 infantometer (Seca, Hamburg, Germany) in duplicate with both measures having to be within 0.5 cm. The infant was weighed nude using the scale from the ADP. After body weight was acquired, the infant was placed inside the chamber wearing only a wig cap where body volume was obtained. The determination of the body volume lasted approximately 3 min with body density converted to percent fat using gender specific equations developed by Fomon (7).

DXA. Whole body composition was determined by DXA using a LunariDXAv11–30.062 (Infant whole body analysis enCore 2007 software, GE, Fairfield, CT) scanner. The infant was placed supine on the scanning bed wearing only a disposable diaper while swaddled in a light cotton blanket provided by the laboratory. Typically, the infant remained awake during the procedure although the lights were off as an animated movie played on a portable DVD player outside of the scanning field. No infant was re-scanned. Body composition variables were calculated using standard procedures outlined in the procedure manual. The same person (David A Fields) positioned and analyzed the scans.


Means and s.d. were calculated for all descriptive and body composition variables. The bivariate relationships between ADP and DXA body composition variables (%fat, fat mass, and fat-free mass) were examined using Pearson correlation coefficient analyses.

Paired sample t-tests were used to examine differences in %fat, fat mass, and fat-free mass measured by ADP vs. DXA whereas scatter plots and 95% confidence intervals were performed on %fat, fat mass, and fat-free mass.

Potential bias between the DXA and ADP was examined using a Bland—Altman analysis (8). This test examines the difference in %fat between the techniques (i.e., ADP and DXA) as a function of the mean %fat by each of the techniques. A nonsignificant correlation indicates no bias across the range of fatness.

The alpha level was set at P ≤ 0.05. All statistics were calculated using SPSS software version 18.0 (SPSS, Chicago, IL).


Body composition estimates by DXA were significantly greater for both relative %fat (31.1 ± 3.6% vs. 26.7 ± 4.7%; P < 0.001) and absolute fat mass (2,284 ± 449 vs. 1,921 ± 492 grams; P < 0.001), with estimates for fat-free mass being significantly less (5,022 ± 532 vs. 5,188 ± 508 g; P < 0.001) vs. ADP estimates (Table 1). Scatter plots were generated for ADP on DXA for each body composition variables (i.e., fat-free mass, fat mass, and %fat) with 95% confidence intervals (Figure 1a—c). The intra-class correlation coefficient revealed significant (P > 0.001) correlations between DXA and ADP for %fat (r = 0.925), absolute fat mass (r = 0.969), and absolute fat-free mass (r = 0.945) (Table 2).

Table 1.  Growth and body composition variables at 6 months (n = 84)
inline image
Figure 1.

The scatter plot of air displacement plethysmography on dual-energy x-ray absorptiometry with 95% confidence intervals for body composition variables of interest with (a) %fat, (b) fat mass, and (c) fat-free mass (n = 84).

Table 2.  Correlation matrix for body composition variables at 6 months (n = 84)
inline image

The Bland—Altman analysis revealed a significant positive correlation (r = 0.558; P < 0.001) for the difference in %fat by ADP and DXA with the mean of %fat by the two methods (Figure 2). ADP % fat was lower than DXA % fat, for all but three individuals.

Figure 2.

A Bland—Altman analysis between air displacement plethysmography (ADP) and dual-energy x-ray absorptiometry (DXA) in the sample of 84 subjects. The middle dashed line represents the mean difference between %fat from ADP and %fat from DXA whereas the upper and lower dashed line represents ±2 s.d. from the mean. The solid line is the regression line. Bias between the techniques was observed, as indicated by a significant P value (r = 0.558; P < 0.000).

In an attempt to better understand and elucidate potential reasons for the difference in %fat calculated by each method, a residual plot examined the associations of body weight and length at 6 months with the difference in %fat. A significant correlation (r = 0.383; P < 0.001) existed between body weight at 6 months and the difference in %fat between ADP and DXA. However, this relationship between body weight at 6 months and the difference in %fat between ADP and DXA only existed in those infants weighing < 7 kg (n = 34; r = 0.355; P < 0.05) as compared with those weighing > 7 kg (n = 50; r = 0.220; P = 0.124). No association existed between body length at 6 months and the difference in %fat between ADP and DXA (r = 0.096; P = 0.384).


Our major finding is that although measures of body composition by DXA and ADP were highly correlated among infants at 6 months of age, there were significant differences in absolute and relative amounts of fat and fat-free mass. Specifically, ADP estimates of %fat were significantly lower than DXA estimates of %fat with few exceptions. This finding is unique, since to our knowledge this is the first study that directly compares body composition data from DXA and ADP in infancy.

DXA is commonly used in the assessment of body composition in children, but is less common in children less than 2 years old. Only a handful of studies have been performed in the first year of life (3,9,10). The paucity of data in infants using DXA is primarily because of radiation exposure, albeit fractionally small, and because of poor subject compliance during testing. A valid test is dependent upon the subject remaining relatively throughout the 5-min test, which is difficult to achieve in children less than 2 years of age.

Results of this study are consistent with studies of older children comparing body composition measured by ADP and DXA. Among both normal weight and overweight children and adolescents, ADP estimates of total fat mass and %fat have been lower than those measured by DXA (11,12,13). In addition, results are consistent with DXA validation studies performed on piglets in which body fat was overestimated by DXA in comparison with the chemical analysis of whole carcass (14,15,16). Validation studies comparing infant %fat measured with ADP to that measured with stable isotope (6) or with the four-compartment model (17) showed no differences in %fat. Together, these findings suggest that ADP may estimate body composition more accurately than DXA, but more work is needed to confirm this.

There are at least three other possible explanations for the discordance between DXA and ADP: infant disposition, testing attire, and differences in the underlying assumptions and principles of DXA and ADP. First, the infant disposition during testing is less critical for ADP vs. DXA. It is important to note that although the subject must remain comparatively still during a DXA scan, movement can be tolerated during the ADP body volume estimate. Although it is possible that greater movement during ADP may have contributed to the discrepancy in body composition estimates, unpublished data from this laboratory suggests that movement during ADP does not greatly influence the body volume measurement per se; volume estimates are made continuously over the 2 min exam with data collected at 2400 Hz, thus reducing the influence of movement at any one of them.

The second potential contributor to the discrepancy in body composition estimates is different attire was permitted during the testing protocols. During ADP, the infant is nude except for a wig cap that compresses scalp hair, which has been shown to negatively affect estimates of body volume (18). During DXA, all infants wore a disposable diaper while swaddled with a light receiving blanket provided by the laboratory. Consequently, although the use of these standardized procedures minimized inter-individual variation during DXA, we cannot rule out the possibility that the greater amount of clothing permitted during DXA may have negatively influenced estimates in body composition.

Third, body composition calculated by DXA and ADP are based upon different principles and assumptions which may have influenced the results. A likely candidate in explaining partially the discrepancy between methods is how DXA deals with pixels with all three body constituents (i.e., bone, fat, and lean). In a typical DXA scan approximately 40–45% of the pixels contain all three constituents, thus are excluded from analysis (19). An underlying principle in DXA assumes that excluded pixels are not different from pixels analyzed to obtain body composition. We are unaware of literature dealing with this issue in infants, but it is reasonable to assume in our study that the number of pixels excluded would be greater given the smaller area of the infant relative to the bone mass of the infant. Although Pieltain and colleagues (20) have shown that DXA has sufficient precision and sensitivity in the determination of body composition early in life, others have questioned its ability to be considered the gold standard (21).

Despite the limitations of DXA, a major benefit relative to ADP is that it gives regional estimates for fat, lean, and bone mineralization. This feature can provide invaluable information in disentangling the relationship between fat and lean mass and the role they play on future disease risk, namely diabetes, cardiovascular disease, and obesity. This is an important topic because trunk fat confers greater risk to future health than appendicular fat (22,23). A second benefit of DXA is that it can be used to measure composition of larger infants, whereas currently available ADP equipment is not equipped for infants larger than ∼8 kg. An understanding of the relevant strengths and weaknesses of different body composition methods in infants will help guide researchers to choose appropriate methods for their particular studies. For instance, longitudinal growth studies requiring numerous repeated estimates of body composition during infancy may likely benefit from the convenience and lack of radiation exposure of ADP, whereas studies investigating regional differences or bone mineralization would require DXA. For studies using both methods, results should be interpreted with caution given the discrepancy in absolute estimates of fat and lean mass.

In conclusion, estimates of %fat (31.1 ± 3.6 vs. 26.7 ± 4.7 %; P < 0.001), total fat mass (2,284 ± 449 vs. 1,921 ± 492 g; P < 0.001), and fat-free mass (5,022 ± 532 vs. 5,188 ± 508 g; P < 0.001) by DXA and ADP produced significantly disparate results, though the methods were highly correlated. Inter-method differences were particularly large in smaller and younger infants. Additional work is needed to identify the technical basis for these rather large inter-method differences. The ability to differentiate different tissues in the body (i.e., fat, fat-free mass, lean tissue, and bone) is important because it provides objective endpoints for individualized nutritional recommendations not only for premature infants but for infants who may be at increased risk of becoming obese later in life (e.g., offspring from obese or gestational diabetic mothers). Future research may consider using both methods to complement each other or switching to DXA when infants become too large for the Pea Pod. However, because of the large error, cross-calibration studies and a correction factor may be necessary if investigators want to use both methods in a single study.


We are thankful to mothers for allowing their infants to participate in this study. Further, we acknowledge Michelle Morrow, Brittney Criswell, Catherine Wolf, and Jeff Lo for their work in coordinating mother/infant visits and data collection/entry. Partial support was provided by Nestlé.


The author declared no conflict of interest.

See the online ICMJE Conflict of Interest Forms for this article.