Air Displacement Plethysmography: Validation in Overweight and Obese Subjects


  • The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. 1734 solely to indicate this fact.

Weight Control Unit, 1090 Amsterdam Avenue, 14th Floor, New York, NY 10025. E-mail:


Objective: Patients with moderate and severe obesity, because of their physical size, often cannot be evaluated with conventional body composition measurement systems. The BOD POD air displacement plethysmography (ADP) system can accommodate a large body volume and may provide an opportunity for measuring body density (Db) in obese subjects. Db can be used in two- or three-compartment body composition models for estimating total body fat in patients with severe obesity. The purpose of this study was to compare Db measured by ADP to Db measured by underwater weighing (UWW) in subjects ranging from normal weight to severely obese.

Research Methods and Procedures: Db was measured with UWW and BOD POD in 123 subjects (89 men and 34 women; age, 46.5 ± 16.9 years; BMI, 31.5 ± 7.3 kg/m2); 15, 70, and 10 subjects were overweight (25 ≤ BMI < 30 kg/m2), obese (30 ≤ BMI < 40 kg/m2), and severely obese (BMI ≥ 40 kg/m2), respectively.

Results: There was a strong correlation between Db(kilograms per liter) measured by UWW and ADP (r = 0.94, standard error of the estimate = 0.0073 kg/L, p < 0.001). Similarly, percent fat estimates from UWW and ADP using the two-compartment Siri equation were highly correlated (r = 0.94, standard error of the estimate = 3.58%, p < 0.001). Bland-Altman analysis showed no significant bias between Db measured by UWW and ADP. After controlling for Db measured by ADP, no additional between-subject variation in Db by UWW was accounted for by subject age, sex, or BMI.

Discussion: Body density, an important physical property used in human body composition models, can be accurately measured by ADP in overweight and obese subjects.


Phenotyping human subjects for adiposity is central to basic and clinical obesity research. A limitation of some current body composition methods, however, is an inability to accommodate subjects with a very large physical size (1). Methods such as DXA(2), magnetic resonance imaging (3), and computerized axial tomography (3) are usually unable to accommodate subjects who have a BMI of over ∼35 kg/m2. Thus, a need exists for an approach capable of providing body composition estimates on subjects with moderate and severe obesity.

One traditional method, underwater weighing(UWW),1 is capable of providing body density (Db) estimates on very large subjects (4). The UWW system at our center is able to obtain Db measurements on compliant patients weighing up to 250 kg. However, not all obese patients are able to participate in the UWW procedure, which requires climbing steps, descending into a tank of water, a breath-hold, and submersion followed by maximal exhalation. Most UWW systems are also not practical to install in clinical and research settings.

The recent introduction of air displacement plethysmography (ADP; LMI, Concord, CA) provides another means by which to measure Db in research and clinical settings(5). The ADP procedure, provided commercially as the BOD POD system, uses variation in pressure and volume, while the subject rests inside a sealed chamber, to estimate Db(6). The ADP method has been validated in normal weight adults (7), but studies of overweight and obese subjects are presently limited. In one study, Vescovi et al. (8) reported accurate measurement of Db by ADP for 29 “overweight” subjects as defined by percent fat compared with UWW as the reference. The maximum BMI evaluated in the study of Vescovi et al. was 37.2 kg/m2. Petroni et al.(9) recently showed the feasibility of using ADP to evaluate nine severely obese patients ranging in BMI from 36.4 to 58.8 kg/m2.

Db is an important physical property incorporated into several body composition models that provide estimates of two or more components (10). The availability of accurate Db estimates that are practical to acquire in subjects with moderate and severe obesity would facilitate phenotyping efforts. Accordingly, in this study, we compared Db measurements provided by ADP to those obtained by UWW in subjects varying widely in BMI, notably in a sample enriched in overweight, obese, and severely obese subjects.

Research Methods and Procedures

Protocol and Subjects

The subjects were adults over the age of 18 years who had Db measured by ADP and UWW on the same day. Subjects with a diagnosed illness were excluded from the cohort, and all subjects were participants in other unrelated research programs. The subjects were instructed to fast overnight before body composition studies the following morning.

The ADP measurement was carried out first, followed by UWW, because the BOD POD estimates are sensitive to subject moisture (11). The Body Composition Core Laboratory of the New York Obesity Research Center uses a 1 to 10 point grading system for compliance with the UWW procedure including maximal exhalation underwater (12). Subjects with high scores are those able to optimally comply with the UWW procedure. All subjects were graded by one trained technician, and only those subjects with UWW scores of 9 and 10 were included in this study. Standing height was measured without shoes to the nearest 10 mm with a wall-mounted Holtain stadiometer (Cross-well, Wales, United Kingdom). All participants signed an informed consent approved by the St. Luke's—Roosevelt Hospital Center Institutional Review Board.


Details regarding the physical concepts and operational principles of ADP are reported by Dempster and Aitkens (6). Subjects were clothed in a tight-fitting bathing suit and acrylic bathing cap. Subjects were weighed to the nearest 0.01 kg using the ADP system's electronic scale (Tanita Corp., Tokyo, Japan). The scale was calibrated daily using a 20-kg weight. Before subject evaluation, a two-point chamber calibration was performed using the empty chamber and a 50.218-liter calibration cylinder. Two trials were performed on each subject, and the measured subject volumes were averaged if they were within 150 mL or 0.2%, whichever was larger (7). If the measurements did not meet the reproducibility criteria, a third trial was performed, and the closest two values were averaged. If none of the three volume measurements were within 150 mL (or 0.2%) of each other, the system was recalibrated, and the test was repeated.

The subject's thoracic gas volume (VTG) was estimated using the BOD POD breathing circuit system (6). The subject was connected to the breathing circuit housed in the rear chamber through a disposable filtered tube, and the subject was instructed to breathe normally until the moment the system induced an airway occlusion. VTG was calculated during occlusion, where subjects were instructed to puff gently. Final body volume was computed based on the initial body volume corrected for VTG and a surface area artifact. All subjects had acceptable BOD POD lung volume measurements. The within-subject day-to-day coefficient of variation for Db measurement of four adults, age and weight range from 22 to 33 years and 52 to 95 kg, respectively, was 0.0062 kg/L.


Db was measured using a four-point force transducer-platform scale system (13) (Precision Biomedical System, University Park, PA). The subject's Db was measured while the subject was clothed in a bathing suit. Subjects were asked to maximally expel as much air as possible from their lungs during complete submersion. An underwater weight was recorded after 5 to 10 trials as the average of the highest three values reflecting the greatest exhalations (12). Residual lung volume was determined before UWW using the oxygen dilution method (14). The within-subject day-to-day coefficient of variation for Db measurement in our laboratory is 0.0035 kg/L. Patients were weighed to the nearest 0.1 kg using a calibrated digital scale (Weight-Tronix; Scale Electronics Development, New York, NY).

Statistical Methods

All statistical analyses were computed using SPSS 12.0 (SPSS, Chicago, IL). Analyses included the entire cohort as well as subgroups in four BMI categories: normal weight (<25 kg/m2); overweight (≥25 and <30 kg/m2); obese (≥30 and <40 kg/m2); and severely obese (≥40 kg/m2).

Simple linear regression analysis was used to compare Db estimates by UWW and ADP and corresponding percent fat estimates by UWW and ADP. Multiple linear regression analyses were performed to evaluate whether the relationship between Db by UWW and ADP was affected by age, sex, and BMI. Bland-Altman plots (15) were used to study potential bias in ADP Db measurements.

The means ± SD for Db measured by ADP and UWW were computed. Between-method Db differences for the total group and for each subgroup were examined using paired Student's t tests. One-way ANOVA was carried out to determine whether between-group Db differences existed among the BMI subgroups. Statistical significance was set at two-tailed p < 0.05.


The baseline characteristics of the subjects are presented in Table 1. There were 123 evaluated adults, 89 men and 34 women, with a BMI ranging from 17.5 to 58.4 kg/m2 and an age ranging from 18 to 73 years. Of the 123 subjects, 15 (12.2%) were overweight, 70 (56.9%) were obese, and 10 (8.1%) were severely obese.

Table 1. . Subject characteristics
 Total (n = 123)Normal (n = 28)Overweight (n = 15)Obese (n = 70)Severely obese (n = 10)
Age (years)46.516.918 to 7339.715.018 to 7335.410.421 to 5753.216.519 to 7134.79.423 to 45
Height (cm)1759147 to 19116810150 to 1841769159 to 1911787147 to 1871769159 to 182
Weight (kg)97.526.742.4 to to 77.884.010.270.4 to 104.4107.79.271.8 to 123.4146.727.7103.0 to 190.2
BMI (kg/m2) to 58.421.91.917.5 to 24.627.11.525.1 to 29.834.01.930.4 to 38.847.25.540.6 to 58.4

The mean Db values for UWW and ADP did not differ significantly for the whole subject group (Table 2), and the two measures were highly correlated (r = 0.94; Db UWW = 0.96 × Db ADP + 0.046; standard error of the estimate (SEE) = 0.0073 kg/L; p < 0.001; Figure 1). After entering Db measured by ADP, no significant additional between-subject variation in UWW-measured Db was explained by multiple regression models that added age, sex, and BMI as independent variables. No significant bias (p = 0.680) was detected in the Bland-Altman analysis (Figure 2).

Table 2. . Db (kilograms per liter) estimates by BMI group
BMI groupnBMI range (kg/m2)UWW Db mean (SD)ADP Db mean (SD)Db difference UWW − ADP mean (SD)*
  • *

    No significant between-group differences were present by ANOVA (p = 0.177).

Normal weight28<25.01.053 (0.019)1.051 (0.020)0.002 (0.008)
Overweight1525.0 to 29.91.047 (0.022)1.043 (0.022)0.004 (0.007)
Obese7030.0 to 39.91.024 (0.013)1.025 (0.014)−0.001(0.007)
Severely obese10>40.01.001 (0.013)1.000 (0.012)0.001 (0.007)
Total12317.5 to 58.41.032 (0.022)1.031 (0.022)0.001 (0.007)
Figure 1.

: Db measured by ADP vs. Db measured UWW for the total study group [Db UWW = 0.046 + 0.96 × (Db ADP); r = 0.94, SEE = 0.0073 kg/L; p < 0.001]. The intercept of the regression line was not significantly different from zero (p = 0.143), whereas the slope was significantly different from one (p < 0.001). The dashed line is the line of identity and solid line is the regression line.

Figure 2.

: Agreement between Db from UWW and ADP. The solid line represents the mean differences (ΔDb) between UWW and ADP (r = 0.038, p = 0.680). The dashed lines represent the mean difference (−0.001 kg/L) and confidence intervals (±1.96 SD).

There were no significant differences in Db measured by UWW and ADP for the subgroups of normal weight (difference, 0.002 ± 0.008 kg/L), overweight (0.004 ± 0.007 kg/L), obese (−0.001 ± 0.007 kg/L), and severely obese (0.001 ± 0.007) subjects. In addition, there were no systematic differences among the four subgroups (p = 0.177; Table 2).

The relationship between percent fat estimates from UWW and ADP is shown in Figure 3. The two percent fat estimates were highly correlated (r = 0.94; percent fat UWW = 0.95 × percent fat ADP + 1.22, SEE = 3.58%, p < 0.001). There was no significant difference between UWW and ADP in the group mean percent fat estimate. Bland-Altman analysis revealed no significant bias for the estimation of percent fat by ADP (p = 0.685) compared with UWW as the reference (Figure 4).

Figure 3.

: Percent fat estimates by ADP vs. percent fat estimates by UWW for the total study group [percent fat UWW = 1.22 + 0.95 × (percent fat ADP); r = 0.94, SEE = 3.58%; p < 0.001]. The intercept of the line was not significantly different from zero (p = 0.242), whereas the slope was significantly different from one (p < 0.001). Dashed line is the line of identity, and solid line is the regression line.

Figure 4.

: Agreement between percent fat from UWW and ADP. The solid line represents the mean differences (Δpercent fat) between UWW with ADP (r = 0.037, p = 0.685). The dashed lines represent the mean difference (−0.33%) and confidence intervals (±1.96 SD).


The phenotyping of obese subjects for adiposity is important to several current research areas. Although BMI is a useful measure of adiposity at the population level, BMI is inaccurate for characterizing adiposity in individual subjects (16). Traditional imaging methods such as DXA, magnetic resonance imaging, and computed tomography are often incapable of accommodating subjects with a very large body volume and weight (3), as is characteristic of patients with moderate and severe obesity. These concerns prompted us to examine ADP as a means of measuring Db in subjects with a wide range of BMIs. Our findings suggest that Db estimates by ADP do not differ significantly from those provided by the traditional UWW method, even in subjects with severe obesity who have BMIs of up to 58.4 kg/m2. Although a time-honored method, UWW is impractical to implement in most clinical and research settings, and some patients are incapable of cooperating sufficiently with the water submersion and maximal exhalation.

Our findings support and extend the ADP studies of earlier investigators. Two approaches have been used to evaluate ADP in these earlier reports. The first approach is to compare Db by ADP to Db by UWW, as in this study, and the second is to compare ADP-percent fat estimates from the two-compartment model to those provided by other methods such as DXA and the total body water two-compartment model. Studies in children (12,17,18), adults (5,8,12,19,20,21,22), elderly (23), adult athletes (24,25,26), Mexicans (27), Japanese (28), and African Americans (17,20) report good agreement using these two evaluation approaches, with an occasional small bias detected between ADP and other methods. Most of this earlier research, however, was carried out in non-obese subjects. In one of the few studies that included a large population, Vescovi et al. (8) evaluated 29 “overweight” subjects as defined by percent fat levels. No significant differences were detected in either Db or percent fat estimates compared with UWW for the “overweight” subjects who had a maximum BMI of 37.2 kg/m2. Our results in 95 overweight and obese subjects, as defined by BMI, similarly show that no between-method bias is present in measured Db up to a maximum BMI of 58.4 kg/m2.

ADP is a rapid and relatively simple method to apply in clinical and research settings. The dependence on the subject to properly perform the UWW procedure for accurate Db measurement serves as a disadvantage compared with the minimal subject participation required for ADP measurements. Entering and exiting the UWW system is often difficult and unsafe for patients with severe obesity. Some patients are apprehensive about the UWW procedure or are unable to perform the required maneuvers (1). Measurements of body volume by ADP are highly reproducible for the calibration cylinder (6), healthy adults varying widely in body weight(5,8,29), and athletes (30).

The comfort of a subject in the BOD POD relates to their physical dimensions and not only their weight. Height and distribution of fat are more likely to limit the ability to accurately measure body volume. Subjects in our study up to weights of 190 kg (418 lb) were successfully measured, and the BOD POD manufacturer reports measurements of subjects to weight of 500 lb. Nonetheless, it can be assumed that some patients who are very tall or have larger midregions may not fit into the BOD POD. In some instances, subjects do get claustrophobic while inside the BOD POD chamber; however, this reaction is not limited to obese subjects.

In conclusion, our findings strongly support the validity of BOD POD Db estimates in overweight and obese subjects. These convenient and practical ADP measurements should prove useful in phenotyping subjects for adiposity.


  • 1

    Nonstandard abbreviations: UWW, underwater weighing; Db, body density; ADP, air displacement plethysmography; VTG, thoracic gas volume; SEE, standard error of the estimate.