The total-breath method yields higher values of DLCO and TLC than the conventional method

Consecutive patients with the clinical diagnosis of COPD [8] were recruited from the pulmonary department at the Hirslanden Hospital Zurich, Switzerland. Healthy subjects, who had no history of respiratory disease, were recruited from hospital staff. Subjects with other types of lung disease than COPD were excluded. The study protocol (clinicaltrial.gov NCT04531293, first trial registration 28/08/2020) was approved by the local ethics committee (Kantonale Ethikkommission Zürich (Cantonal Ethics Committee), BASEC 2020–02139), and all participants gave their written informed consent. All experiments were performed in accordance with relevant guidelines and regulations, in particular the Declaration of Helsinki.

Spirometry was performed using the EasyOne Pro (ndd Medical Technologies, Switzerland) following international recommendations [9]. Forced expiratory volume in 1 s (FEV₁), forced vital capacity (FVC) and their ratio FEV₁/FVC were reported, with predicted values from GLI (Global Lung Function Initiative) [10]. Lung volume from body plethysmography (TLC_body) was estimated using the Master Screen Body (Vyaire Medical, Höchberg, Germany). Diffusing capacity for carbon monoxide (CO) was determined in a single-breath maneuver using the EasyOne Pro (ndd Medical Technologies, Switzerland) and the integrated rapid gas analyzer. The single-breath maneuver was performed twice, separated by at least 5 min [11], and expiration was terminated if a volume plateau was reached. Data was analyzed conventionally by averaging over a defined sample of expiratory volume (conventional sample of 500 mL, CS), or utilizing the total time course of in- and expiration (total-breath, TB) for helium concentration (see Fig. 1). The conventional analysis was available in the EasyOne Pro device (software version V3.7.2.2), which has been shown to yield reliable values in previous studies [12‐15]. The total-breath analysis was performed via a custom algorithm after export of the raw data, using the entire tracer gas (helium) concentration during inspiration and expiration (see Fig. 1) within a mass-balance equation. DLCO and TLC were calculated using the residual concentration of helium in the exhaled air as measured at the start of each test. VA was related to TLC by dead space computed as 2.2 mL/kg body weight for BMI < 30 kg/m², and by height² (in m) divided by 189.4 for BMI ≥ 30 kg/m² [4]. Predicted values were taken from the Global Lung Function Initiative reference equations [10, 16].

Continuous data were summarized as mean values and standard deviations (SD). Bland–Altman plots were used to assess the agreement between estimates of DLCO and TLC obtained by different methods. Pearson’s correlation coefficients and linear regression analysis were employed to quantify associations between a set of explanatory variables (sex, age, height, BMI, Z-Scores of FEV₁ or FEV₁/FVC) and the differences between the values of TLC or DLCO determined by the two different methods, whereby FEV₁ or FEV₁/FVC were tested separately due to their collinearity. A similar approach was followed with plethysmographic TLC as dependent variable. Adjusted R² values and residual standard deviations (RSD) were used to quantify the accuracy of predictions. Groups were compared using the unpaired t-test, or analysis of variance (ANOVA) with post-hoc comparisons according to Duncan in case of more than two independent groups. Outcome values estimated in the same subject were compared using the paired t-test.

The concordance between variables was quantified by their normalized root mean square difference (NSD), defined as standard deviation of the individual differences between two variables divided by their mean value [17]. This is a measure of the residual error of estimation of one variable from another variable after removing their potential systematic difference (bias). The sample size was based on findings by Horstman et al. [5]. Following the aim to detect a difference of the magnitude of 50% of its standard deviation in three pairwise comparisons at a total alpha of 0.05 and a power of 0.95, at least 69 subjects were needed; due to the probably higher variability particularly in patients with severe COPD we aimed at a final size of n = 120. Statistical significance was assumed for a type I error of p < 0.05. All statistical analyses were performed using the package SPSS (Version 26, IBM, Armonk, NY, USA).

Data from 23 healthy subjects and 95 patients with COPD were available for analysis (Table 1). Subjects with airflow obstruction included all four GOLD grades; with more subjects with GOLD 1 and 2 compared to 3 and 4 (Table 1).

The time between the duplicate assessments was at minimum 312 s, with a median (quartiles) of 342 (330; 363) seconds. The repeatability of DLCO from the duplicate measurements was high for both the TB and CS method, mean values (95%CI) of differences between 1^st and 2^nd assessments being -0.002 (-0.069; 0.065) and -0.008 (-0.070; -0.055) mmol*min^-1*kPa^-1, respectively. Similar repeatability was observed for TLC, mean values (95%CI) of differences being -0.010 (-0.049; 0.029) and -0.018 (-0.050, 0.014) L for the TB and the CS method, respectively. Therefore, the mean values of the duplicate assessments were used in all subsequent analyses.

The correlation between the values of DLCO_TB and DLCO_CS was high (r = 0.994; p < 0.001). To elucidate their relationship in detail, Fig. 2A shows a Bland–Altman plot of the difference of DLCO_TB and DLCO_CS versus their mean value. On average (± SD), DLCO_TB was higher by 0.469 ± 0.267 (95%CI: 0.420; 0.517; p < 0.001) mmol*min^-1*kPa^-1 than DLCO_CS. The difference was systematically offset and not a function of the magnitude of DLCO, as it was not correlated with the mean value. The NSD between both values was 4.3%.

Accounting for sex, age, height and BMI as covariates, the difference between DLCO_TB and DLCO_CS increased as the FEV₁/FVC Z-Score decreased (adjusted R² = 0.216, Fig. 3A, p < 0.001). A similar result was observed for the ratio of DLCO_TB to DLCO_CS (adjusted R² = 0.423, Fig. 3B). Both associations were weaker in terms of adjusted R², when using the Z-Score of FEV₁, indicating the adequacy of FEV₁/FVC as associated index of airway obstruction.

Over the whole range of values, the correlation between TLC_TB and TLC_CS was high (r = 0.953; p < 0.001). Figure 2B shows the results of Bland–Altman plots of the difference of TLC_TB and TLC_CS versus their mean value. Mean (± SD) TLC_TB was higher by 0.495 ± 0.371 L than TLC_CS (95%CI: 0.427; 0.562 L; p < 0.001). Their difference did not significantly correlate with their mean value, and the NSD was 6.7%.

Using sex, age, height and BMI as covariates, again the Z-Score of FEV₁/FVC was found to be significantly associated with the difference between TLC_TB and TLC_CS (p < 0.001, adjusted R² 0.417) or their ratio (p < 0.001, adjusted R² 0.429) and being superior to the Z-Score of FEV₁.

Both TLC_CS and TLC_TB were correlated with TLC_body (r = 0.709 and 0.824, respectively; p < 0.001 each). The mean (± SD) difference (95%CI) between TLC_body and TLC_CS was 1.558 ± 0.940 (95%CI: 1.387; 1.739) L and between TLC_body and TLC_TB 1.064 ± 0.740 (95%CI: 0.929; 1.199) L; for the differences in single groups see Table 1. The corresponding NSD values were 15.4% for TLC_CS and 11.7% for TLC_TB. Thus, TLC_TB was closer to TLC_body than TLC_CS.

We then elucidated the relationship between TLC values from gas dilution and plethysmography using regression analysis, assuming that their relationship depended on airway obstruction [18] choosing the Z-Scores of FEV₁ or FEV₁/FVC that should always be available in practice, in addition to sex, age, height, BMI as covariates. The dependent variables were the ratios or differences of TLC_CS and TLC_TB versus TLC_body. In terms of adjusted R², the Z-Score of FEV₁ turned out to be slightly superior to that of FEV₁/FVC for all of these outcomes. Figure 4A shows the ratios TLC_CS/TLC_body and TLC_TB/TLC_body plotted against the Z-Score of FEV₁, and Fig. 4B the respective differences.

Using linear regression analysis with TLC_TB as the only predictor, TLC_body could be estimated with a residual error of 0.720 L, which was reduced to 0.621 L by inclusion of the Z-Score of FEV₁ (adjusted R² 0.676 and 0.756, respectively). Using TLC_CS as predictor, the residual errors were 0.895 and 0.745 L, respectively (adjusted R² 0.499, 0.651). The NSD for the estimation of TLC_body from TLC_TB and FEV₁ Z-Score was 9.1%, for the estimation from TLC_CS and FEV₁ Z-Score 10.8%.

The number of patients of GOLD grades 3 and 4 was smaller than that of grades 1 and 2. A higher number of grade 3 and 4 patients might have allowed a more detailed investigation of the factors that explain the difference between plethysmographic TLC and TLC assessed by the TB method, as well as the difference between DLCO values determined by conventional and TB method. On the other hand, this question would probably have benefitted more from additional information such as CT scans in these patients, which were not possible in this study. Such information might be suitable to explain part of the differences between the methods compared in this work. In addition, two of the healthy control subjects showed a Z-Score of less than -1.645 for FEV₁/FVC. This was a percentage expected by chance, and we left these subjects in the group as we aimed to achieve a continuum of airway obstruction; their omission did not qualitatively change the results. We also considered the fact that the control group was much younger than the COPD group, as an advantage instead of a limitation, since we aimed to include a group for the comparison of the methods, in which optimal lung ventilation and thus minimal differences between parameters could be assumed. From a clinical point of view, the inclusion of patients with restrictive lung disorders or with asthma would have been of interest. Although these patients probably would show similar results as the healthy control subjects, unexpected findings might occur that could be informative for phenotyping. We do not know, whether the TB method is available in other devices than that used in the present study; if that is the case, such devices should be compared to warrant a standard. Moreover, the use of methane instead of helium as tracer gas might result in lower values of TLC particularly in patients with severe COPD and therefore increase the discrepancy from plethysmographic TLC. In addition, it might be helpful to assess TLC_body using equipment from different manufacturers, as there is no obligatory standardization regarding these devices.