INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

A NEW AIR-DISPLACEMENT plethysmograph (BOD POD body composition system; Life Measurement Instruments, Concord, CA) for measurement of body composition has recently been described (6). This system utilizes the inverse relationship between pressure (P) and volume (V) (Boyle's law: P₁V₁ = P₂V₂) to determine body volume (Vb). Once Vb is determined, the principles of densitometry are used to determine body composition from body density (Db = mass/volume) (2, 4). A previous study in our laboratory indicated that body composition estimates by the BOD POD are not significantly different from those determined by hydrostatic weighing (HW) (16). Thus a chief advantage of the BOD POD is that it represents a densitometric method that is based on air displacement rather than on water immersion; therefore, it is simpler and more rapid than HW and potentially has wider clinical application.

Measurement of Db by HW typically requires determination of residual lung volume (VR) to correct for air remaining in the lungs after maximal exhalation (13). Any air remaining in the body (e.g., in the lungs or gastrointestinal tract) will have the effect of increasing buoyancy, leading to an overestimate of Vb, an underestimate of Db, and thus an overestimate of the percentage of body fat (%BF). Although VR can be measured accurately by numerous techniques, such as O₂ dilution, N₂ washout, and He dilution, several investigators have examined whether VR can be predicted without compromising the accuracy of body composition measurements by HW (8, 11, 14, 17, 22). Whereas group means can sometimes be predicted with accuracy (17, 22), VR is usually systematically over- or underestimated, and resulting errors have been observed of up to ~4.0% BF in normal, healthy subjects (8, 11, 14, 17). In addition, individual errors in the calculation of %BF resulting from the use of predicted VR (VR_pred) can be unacceptably high (22).

Similarly, measurement of Db by air displacement requires determination of the quantity of air in the lungs during normal tidal breathing, or the average thoracic gas volume (Vtg). As described by Dempster and Aitkens (6), Vb determined by Boyle's law is underestimated by 40% of the Vtg because the air in the lungs, because of its isothermal nature, is 40% more compressible than the surrounding air (adiabatic conditions). Failure to correct for this phenomenon will result in an overestimate of Db and, consequently, an underestimate of %BF. Thus an ancillary measurement of Vtg by the BOD POD is incorporated into the testing procedure (15). Alternatively, the BOD POD also offers the use of a predicted Vtg (Vtg_pred). In certain situations, such as when screening a large number of subjects or repeatedly testing the same subject to evaluate changes in body composition, the use of Vtg_pred would be time saving if it offered a reasonable estimate of Vtg.

Therefore, the purpose of this analysis was threefold. We sought to 1) compare measured Vtg (Vtg_meas) and Vtg_pred (Vtg_meas vs. Vtg_pred); 2) determine the effect of using Vtg_pred on the estimation of %BF; and 3) compare and contrast the use of Vtg_pred with air displacement with the use of VR_pred in conjunction with HW.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Air-displacement plethysmography. Body composition data were studied from the 50 subjects who had most recently been tested in our laboratory by air displacement with Vtg_meas. Subjects were healthy and not on medication known to affect lung volumes. One subject smoked cigarettes but had lung volume measurements that were well within the range of those for the remaining subjects. The methods used are described in detail elsewhere (6, 16). Briefly, after voiding the bladder, each subject was weighed to the nearest gram while wearing a swimsuit. Height was measured to the nearest centimeter. The BOD POD was calibrated according to the manufacturer's instructions, and raw body volume (Vb_raw) was determined. Finally, Vtg was measured in the BOD POD by using a technique, common to standard pulmonary plethysmography, called the "panting maneuver" (7). While wearing a noseclip, the subject breathed through a tube; after two to three normal breaths, the airway was occluded for 3 s at midexhalation. During this time, the subject was instructed to gently puff against the occlusion by alternately contracting and relaxing the diaphragm. This technique is analogous to the gentle repeated exhalations one might use to fog one's glasses before cleaning them.

(1)

(2)

HW. Twenty-four of the subjects who had been tested by air displacement had also undergone HW on the same day. The subjects were weighed underwater, according to standard procedures described in detail by McCrory et al. (16). VR was measured on land by using the O₂-dilution technique (21). Db from HW (Db_HW )was calculated as

(3)

where M_a is body mass in air, M_w is body mass in water, and D_w is the temperature-corrected water density (4). %BF was calculated from Db_HW by using the Siri formula (20).

Lung volume prediction. Because Vtg is measured at the midpoint of exhalation, Vtg_pred was calculated as

(4)

where FRC is functional residual capacity, and VT is the tidal volume estimated during normal breathing. FRC and VR were predicted from age and height according to the following formulas developed by Crapo et al. (5)

(5)

(6)

(7)

where predicted FRC (FRC_pred) and VR_pred are in liters, Ht is height in centimeters, and A is age in years. These prediction equations were developed from healthy subjects, ages 17-91 yr, by using single-breath He dilution to measure FRC and VR. %BF measured by air displacement and by HW were recalculated by using Vtg_pred and VR_pred, respectively.

Statistical analysis. Data were analyzed by using SAS-PC, version 6.04 (19). Group means were compared by either one-sample (paired) or two-sample t-tests where appropriate. Associations between variables were assessed by calculating Pearson correlation coefficients. To determine how well predicted variables reflected measured variables, the measured variable was regressed on the predicted variable with the use of linear-regression analysis. The residuals from this regression were then plotted against the predicted variable and tested for skewness, curvilinearity, and heteroskedasticity. All tests were two-tailed, and the level of significance was set at P = 0.05.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Physical characteristics of the subjects are shown in Table 1. On average, the men weighed more and were taller than the women (P < 0.01). Because of the height difference, the men had a larger Vtg_meas, Vtg_pred, and VR_meas than did the women (P < 0.01). The gender difference in VR_pred approached but did not reach statistical significance (P = 0.06). Vtg_pred was not significantly different from Vtg_meas for either gender; VR_pred was significantly higher than VR_meas in the women but not in the men (P < 0.03).

Mean values for the predicted and measured lung volumes and for %BF calculated by using predicted and measured lung volumes are shown in Table 2. There was no significant difference between Vtg_pred and Vtg_meas, nor in %BF measured by air displacement calculated by using Vtg_pred vs. Vtg_meas. In contrast, VR was overpredicted by 187 ml or 14% (P = 0.0004). This had the effect of significantly underestimating %BF by HW when calculated by using VR_pred (P = 0.0004). Results from the linear regression analyses of measured on predicted variables are also shown in Table 2. R² values for Vtg and VR were 0.63 and 0.52, indicating moderate agreement between predicted and measured lung volumes. Whereas the standard error of the estimate (SEE) for VR was less than one-half that for Vtg (0.22 vs. 0.44 liters), they are similar when expressed as a percentage of the mean corresponding predicted lung volume (±0.14 and ±0.12%, respectively). R² values for regressions of %BF calculated by using the measured lung volume vs. that calculated using the predicted lung volume were high for both methods (0.98 and 0.96 for air displacement and HW, respectively), and corresponding SEEs were ±1.36 and ± 1.67% BF, respectively.

Figure 1, A and B, illustrates the residuals from the linear regressions of Vtg_meas on Vtg_pred and of VR_meas on VR_pred, respectively, plotted against the corresponding predicted lung volume. For both Vtg and VR, there appear to be an approximately equal number of positive and negative residuals, indicating that the prediction errors are fairly evenly distributed around zero; i.e., there is little skewness. Tests for curvilinearity and heteroskedasticity were not significant.

View larger version (10K): [in this window] [in a new window]   Fig. 1.   Residuals from linear regressions of measured on predicted thoracic gas volume (Vtgmeas on Vtgpred, respectively; A) and of measured on predicted lung residual volume (VRmeas on VRpred, respectively; B) plotted against corresponding predicted lung volume.

Residuals from the linear regressions of %BF calculated by using measured lung volume on that calculated by using predicted lung volume are shown in relation to %BF calculated by using the predicted lung volume in Fig. 2, A and B. For both air displacement and HW, there appears to be a similar range of prediction errors (within ±2.0 to ±3.0% BF), with the exception of one data point for HW with a residual of nearly 4.0% BF.

View larger version (12K): [in this window] [in a new window]   Fig. 2.   Residuals from linear regressions of (A) air-displacement measurement of %body fat (%BF) calculated by using Vtgmeas on %BF calculated by using Vtgpred and of (B) hydrostatic-weighing measurement of %BF calculated by using VRmeas on %BF calculated by using VRpred, plotted against corresponding predicted lung volume.

For individual subjects, calculating differences in %BF by using measured and predicted lung volumes yielded the following results. For air displacement, estimates of %BF by using Vtg_pred were within ±1% BF calculated by using Vtg_meas for 58% of subjects and ±2% BF calculated by using Vtg_meas for 82% of subjects. However, for HW, estimates of %BF by using VR_pred were within ±1% BF calculated by using VR_meas for 25% of subjects and ±2% BF by using VR_meas for 46% of subjects.

The error in estimation of %BF by using predicted lung volume in relation to the difference between predicted and measured lung volumes is illustrated in Fig. 3. When Vtg is underpredicted, %BF is underestimated; when Vtg is overpredicted, %BF is overestimated. In contrast, when VR is underpredicted, %BF is overestimated; when VR is overpredicted, %BF is underestimated. Thus a given error in lung volume prediction by air displacement and HW affects the calculation of %BF in opposite directions, a function of Eqs. 2 and 3 (see MATERIALS AND METHODS). In addition, it can be seen from Fig. 3 that a given error in the prediction of Vtg will have less than one-half the effect on measurement of %BF by air displacement than will the same error in the prediction of VR on measurement of %BF by HW; this is also a function of Eqs. 2 and 3.

View larger version (24K): [in this window] [in a new window]   Fig. 3.   Difference in %BF calculated by using predicted vs. measured lung volumes as function of difference between predicted and measured lung volumes for air-displacement plethysmography () and hydrostatic weighing (). For any given difference between predicted and measured lung volume, the magnitude of the effect on %BF calculated by air displacement is only 40% of that calculated by hydrostatic weighing.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

The principal findings of this study show that, on average, Vtg_pred did not differ significantly from Vtg_meas, and there was no difference in body composition estimates measured by air displacement by using Vtg_pred rather than Vtg_meas. In contrast, VR was systematically overpredicted, leading to a group mean underestimate of 1.4% BF from HW. Further analysis revealed that, although the outer ranges of individual estimates of %BF by using the predicted vs. measured lung volume for air displacement and HW were similar (±3-4% BF), for air displacement the majority of subjects (82%) fell within ±2% BF of that when using Vtg_meas, but for HW, only about one-half of the subjects (46%) fell within ±2% BF of that when using VR_meas. Finally, it is notable that a given absolute error in Vtg_pred will lead to an error in the estimation of %BF that is less than one-half that of using VR_pred. This is attributable to the fact that only 40% of the Vtg enters into the calculation of Vb by air displacement (Eq. 2), whereas 100% of the VR enters into the calculation of %BF by HW (Eq. 3).

In this study, VR was overpredicted by 187 ml on average, and the overprediction of VR occurred to a greater extent in women compared with men. In a variety of studies conducted in healthy subjects, VR_pred has been reported to be, on average, 1.23 liters less than VR_meas (11), 495 ml greater than VR_meas (8), and approximately equal to VR_meas (8, 11, 17, 22). There appears to be no consistency as to whether VR is over- or underpredicted in women vs. men. The lack of congruence among these studies with regard to the accuracy of VR_pred may be related to intersubject variability but also to the particular equation used to predict VR. There are several prediction equations for VR (e.g., Refs. 1, 3, 5, 9, 10, 12, 18, 22), each developed by using different reference techniques (e.g., O₂ dilution and He dilution). The agreement of VR_pred and VR_meas likely also depends on the method used to measure VR. Forsyth et al. (8) reported that average VR_meas range in women was from 1.130 to 1.437 liters and in men was from 1.236 to 1.759 liters, depending on the method used to measure VR. The equations by Crapo et al. (5) were used in the present study to predict VR (Eqs. 6 and 7) because the manufacturer of the BOD POD incorporates the estimate of FRC by those authors in their prediction of Vtg (Eq. 5). To our knowledge, there are no published studies on the prediction of VR by using the Crapo et al. (5) equations in conjunction with HW, so a direct comparison with other studies is not possible. In the present study, VR was predicted by equations developed by using He dilution, yet VR was measured by O₂ dilution. These differing methods may be responsible to some degree for the differences between VR_pred and VR_meas.

Despite the lack of agreement among studies on the accuracy of VR prediction, most investigators have found that erroneous estimates of %BF for individual subjects can arise when using VR_pred. Wilmore (22) reported that using VR_pred had no effect on the estimation of %BF for the group, but individual estimates deviated quite substantially from that calculated by using VR_meas, in some cases deviating by >4% BF. Similarly, Hackney and Deutsch (11) calculated that, on average, with a standardized man of 70 kg and 15% BF as the subject, calculating %BF by using the VR_pred values obtained in their study resulted in average overestimates of 16.0% BF by using the Wilmore equation (22) and of 11.6% BF by using the Boren equation (3).

Keys and Brozek (13) have calculated that an error of ±100 ml in VR results in an error of ±0.8% BF by HW for a person weighing 70 kg with 20% BF. Because only 40% of the Vtg is used in the calculation of Vb measured by air displacement, an error of ±100 ml in Vtg would result in an error of only ±0.3% BF (assuming Vb_raw = 64.0 l, Vtg = 4.0 l, and SAA = 0.9 l). It would take an error of ~10% in Vtg determination for a Vtg of 4.0 liters (±400 ml) to produce an error of ±0.8% BF by air displacement in the same 70-kg, 20%-BF person. We have observed that, on average, the between-day coefficient of variation of Vtg_meas is ±3.5% (n = 22; unpublished observations). For a Vtg of 4.0 liters, this represents an error of ±140 ml and a resulting error of about ±0.4% BF.

Conclusion. Our findings support the use of Vtg_pred in conjunction with air-displacement plethysmography for group mean comparisons and for purposes such as screening in young to middle-aged individuals. This contrasts with the use of VR_pred in conjunction with HW, which leads to significant errors in the estimation of %BF. Furthermore, although the use of Vtg_pred may be valuable in some circumstances, obtaining Vtg_meas is relatively simple in most cases. Therefore we recommend that the use of Vtg_meas remain a part of standard experimental and clinical practice. The results of this study do not necessarily apply to other groups, such as pediatric or elderly subjects, or to those with pulmonary dysfunction. Additional research is needed to determine the validity of Vtg_pred in these groups.