The results partially support our hypothesis that the new DUCHARME formula offers a better estimation of measured TGV for young adults than the prediction formula in the Bod Pod software (CRAPO) or the child-specific formula of Fields et al. (2004) (FIELDS). DUCHARME was the only formula that produced a group mean that was not significantly different from measured TGV. The regression line closely matched the line of identity. Additionally, DUCHARME resulted in fewer errors exceeding ± 2% BF compared to the other prediction formulas. However, the SEEs were similar across all three prediction equations. Further, a bias exists for all three equations such that TGV is overestimated in people with a small TGV and underestimated in people with a large TGV. The magnitude of error is somewhat blunted with DUCHARME compared to CRAPO, but the bias is still present. Thus, prediction errors can be large for people with small or large TGVs.

The prediction formula in the Bod Pod software is from Crapo et al. (1982) (CRAPO). Numerous investigators reported no significant mean difference between CRAPO and measured TGV (Collins and McCarthy 2003; McCrory et al. 1998; Otterstetter et al. 2015; Wagner 2015). However, consistent with the findings in the present study, other investigators reported lower values for CRAPO compared to measured TGV, leading to an underestimation of %BF when this prediction formula is used (Ducharme et al. 2021; Miller et al. 2023). Ducharme et al. (2021) noted a sex-specific difference might exist for CRAPO such that this formula underestimates men but is valid for women. Correspondingly, we found the mean difference between CRAPO and measured TGV for the women in the present study was not significant, and the underestimation was more pronounced in men. Close inspection of the studies that reported no significant mean difference between CRAPO and measured TGV (Collins and McCarthy 2003; McCrory et al. 1998; Otterstetter et al. 2015; Wagner 2015) revealed that all of them had a preponderance of women; whereas, the present investigation and those that found CRAPO underestimates measured TGV included samples that were either exclusively (Miller et al. 2023) or predominantly (Ducharme et al. 2021) men.

More important than the mean difference is the issue of bias in the CRAPO prediction. Consistent with the present investigation, multiple researchers have reported an increasing magnitude of error as CRAPO gets further away from a typical value (Ducharme et al. 2021; Minderico et al. 2008; Wagner 2015). Ducharme et al. (2021) noted that women with a measured TGV ≥ 3.5 L and men with a measured TGV ≥ 4.5 L would be underpredicted when using the CRAPO formula. We concur. Further, women with measured TGV ≤ 2.7 L and men with measured TGV ≤ 3.3 L are likely to be overpredicted by > 0.5 L when using the CRAPO formula.

The over- and underestimation bias evident in CRAPO (see Fig. 2a) led Ducharme et al. (2022) to create a new TGV prediction formula based on height, body mass, and sex rather than assumptions about tidal volume and functional residual capacity used by Crapo et al. (1982). During the validation of their equation, Ducharme et al. (2022) reported that a bias still existed for their formula, but the prediction errors were less than those of CRAPO. The findings from the present study are consistent with those of Ducharme et al. (2022); a bias still exists for DUCHARME, but the constant error and percentage of participants with %BF errors ≥ 2% BF is less when using the DUCHARME formula compared to CRAPO.

Regarding FIELDS, we were curious as to how a formula developed from a sample of 6- to 17-year old children would perform when applied to a sample of young adults. Not surprisingly, measured TGV was underpredicted by nearly 0.5 L, leading to an underestimation of %BF. However, the SEEs from the regression analysis and bias from the Bland and Altman (1999) plots were similar between FIELDS and DUCHARME; thus, an adjustment to the y-intercept (see Fig. 1c) or constant error might improve the suitability of this child-specific formula for adults. Fields et al. (2004) developed FIELDS because CRAPO overestimated measured TGV in their sample of children. Other investigators have also found FIELDS to be more suitable than CRAPO when predicting the TGV of children (Higgins et al. 2006; Holmes et al. 2011; Radley et al. 2007).

Both Fields et al. (2004) and Ducharme et al. (2022) highlighted a strong relationship between height and measured TGV in the derivation of their formulas. Consequently, height is the primary predictor variable in each of their TGV prediction formulas (see Table 1). Fields et al. (2004) reported Pearson correlations of 0.79 and 0.89 for the girls and boys, respectively, in their study of children. Similarly, Ducharme et al. (2022) reported a height:TGV relationship of 0.76 in their validation sample of 140 young adults. In the present study, the height:TGV relationship was not as strong at 0.67. This weaker relationship between height and measured TGV in the present study might account for the slightly larger prediction errors that we observed compared to what was reported by Ducharme et al. (2022). Ducharme et al. (2022) reported an SEE of 0.56 L for their prediction equation, whereas the SEE for DUCHARME in our study was 0.63 L.

As noted at the beginning of the results section, even after five trials and extensive coaching 13% of the recruited participants were unable to successfully perform the puffing maneuver required for measured TGV. Few researchers report the failure rate to successfully complete the measured TGV procedure. Of those that have acknowledged it, the failure rate in adults has ranged from 9 to 26% (Anderson 2007; Miller et al. 2023; Wagner 2015) with a failure rate of 31% reported in children (Lockner et al. 2000). The inability of some individuals to produce an acceptable “merit” score for TGV according to the Bod Pod software necessitates the use of a predicted value. This further highlights the importance of this research and the need for accurate TGV prediction formulas.

There are several strengths to this study. First, it is the only known study to compare simultaneously all known TGV prediction equations recommended for use with the Bod Pod. Second, our sample was similar in size to the original derivation studies of Crapo et al. (1982), Fields et al. (2004), and Ducharme et al. (2022), and was larger than most subsequent validity studies of these TGV prediction formulas. Further, our sample was diverse in height and body size leading to a wide range of lung volumes and %BF values. Nevertheless, there are several study limitations. First, measured TGV in the Bod Pod is an estimate based on plethysmography principles and relies on physiological assumptions (e.g., isothermal gas compression, airway compliance, and pressure transmission) (Criee et al. 2011; Wanger et al. 2005). Therefore, the observed agreement between DUCHARME and measured TGV may reflect convergence of estimation errors. We did not include an alternative comparison method, such as another body plethysmograph (e.g., Body Box) or gas dilution (e.g., helium dilution or nitrogen washout). However, previous research demonstrated strong reliability and validity of Bod Pod measured TGV compared to gas dilution techniques (Davis et al. 2007). Second, the findings from this study are limited to young adult athletes.

In summary, all of the TGV prediction equations have a bias such that individuals with a small TGV are overpredicted and those with a large TGV are underpredicted. Consequently, it is best to measure TGV when possible. A measured TGV is imperative if accuracy of a Bod Pod assessment is critical to a study or athlete evaluation. If using the Bod Pod to track longitudinal changes in body composition, a predicted TGV may be acceptable because the error will be constant from pre-test to post-test. If TGV must be predicted, the formula of Ducharme et al. (2022) (DUCHARME) is recommended for adults. Although a bias exists for DUCHARME, this formula blunts the magnitude of error compared to other prediction formulas, particularly for individuals at the extremes of the TGV continuum. Fewer individuals were misestimated by ≥ 2% BF when using DUCHARME compared to CRAPO or FIELDS in a sample of young adult athletes. Thus, we recommend that the Bod Pod manufacturer replace the CRAPO equation with the DUCHARME equation as the default prediction for TGV in their software.