Transcutaneous PCO₂ for Exercise Gas Exchange Efficiency in Chronic Obstructive Pulmonary Disease

Abstract

Gas exchange inefficiency and dynamic hyperinflation contributes to exercise limitation in chronic obstructive pulmonary disease (COPD). It is also characterized by an elevated fraction of physiological dead space (V_D/V_T). Noninvasive methods for accurate V_D/V_T assessment during exercise in patients are lacking. The current study sought to compare transcutaneous PCO₂ (TcPCO₂) with the gold standard—arterial PCO₂ (PaCO₂)—and other available methods (end tidal CO₂ and the Jones equation) for estimating V_D/V_T during incremental exercise in COPD. Ten COPD patients completed a symptom limited incremental cycle exercise. TcPCO₂ was measured by a heated electrode on the ear-lobe. Radial artery blood was collected at rest, during unloaded cycling (UL) and every minute during exercise and recovery. Ventilation and gas exchange were measured breath-by-breath. Bland-Altman analysis examined agreement of PCO₂ and V_D/V_T calculated using PaCO₂, TcPCO₂, end-tidal PCO₂ (P_ETCO₂) and estimated PaCO₂ by the Jones equation (PaCO₂-Jones). Lin’s Concordance Correlation Coefficient (CCC) was assessed. 114 measurements were obtained from the 10 COPD subjects. The bias between TcPCO₂ and PaCO₂ was 0.86 mmHg with upper and lower limit of agreement ranging −2.28 mmHg to 3.99 mmHg. Correlation between TcPCO₂ and PaCO₂ during rest and exercise was r²=0.907 (p < 0.001; CCC = 0.941) and V_D/V_T using TcPCO₂ vs. PaCO₂ was r²=0.958 (p < 0.0001; CCC = 0.967). Correlation between PaCO₂-Jones and P_ETCO₂ vs. PaCO₂ were r²=0.755, 0.755, (p < 0.001; CCC = 0.832, 0.718) and for V_D/V_T calculation (r²=0.793, 0.610; p < 0.0001; CCC = 0.760, 0.448), respectively. The results support the accuracy of TcPCO₂ to reflect PaCO₂ and calculate V_D/V_T during rest and exercise, but not in recovery, in COPD patients, enabling improved accuracy of noninvasive assessment of gas exchange inefficiency during incremental exercise testing.

Keywords:

• Arterial blood gas
• transcutaneous carbon dioxide partial pressure
• exercise testing
• dead space
• end-tidal PCO₂

Introduction

Shortness of breath related to dynamic hyperinflation limits the ability to exercise in many COPD patients [1, 2] while leg fatigue provides an additional impediment to exercise in some [3]. Dyspnea in COPD is exacerbated by increased “wasted” ventilation related to high physiological dead space [4]. Bohr [5], and subsequently Enghoff [6], evaluated the physiological dead space proportion (V_D/V_T) using alveolar (P_ACO₂) and then arterial PCO₂ (PaCO₂), conceptually calculating the fraction of the exhaled breath that does not participate in CO₂ exchange with the blood. Obtaining arterial blood samples, however, adds burden to the patient and the laboratory staff, is generally expensive and usually is not obtained with the necessary data density to assess the time course of the subject’s dead-space ventilation during exercise testing. Jones, et al. [7] introduced an equation to estimate PaCO₂ based on the end-tidal PCO₂ (P_ETCO₂) and tidal volume (V_T). However, as stated in their paper, this estimate is only valid in healthy subjects and differs from the PaCO₂ where abnormal ventilation to perfusion ratios (V̇_A/Q̇) occur, i.e., in most patients with lung disease [7–9]. Nevertheless, many manufacturers of cardiopulmonary exercise testing equipment utilize this equation (here termed, PaCO₂-Jones [7] = 5.5 + 0.9 * P_ETCO₂ − 0.0021* V_T) to provide a so-called “noninvasive estimate” of V_D/V_T in all tested subjects, irrespective of the presence of disease.

A noninvasive method to estimate PaCO₂ can be obtained using a heated buffered pH electrode to measure transcutaneous PCO₂ (TcPCO₂) [10, 11]. This system measures PCO₂ evaporated from the capillary bed of the heated skin and provides a close correlate of PaCO₂. Prior studies to compare TcPCO₂ with PaCO₂ during exercise in patients have often failed to obtain simultaneous measurements, have required at least one PaCO₂ measurement, or have not calculated V_D/V_T [11–14]. Recently, Fernandes et al. compared TcPCO₂ with PaCO₂, but the validation was only during rest and was not extended into exercise [15]. The most rigorous prior exercise study in 14 patients (including COPD, ischemic heart disease and hyperventilation syndrome) found narrow limits of agreement between V_D/V_T calculated from TcPCO₂ and PaCO₂ (mean bias: 0; CI −0.02 to +0.02), but used an “in vivo calibration” method where the TcPCO₂ values were calibrated using a resting PaCO₂ measurement prior to the exercise test [16].

The goal of this study was to determine the precision and accuracy of noninvasive TcPCO₂ measurement as a surrogate for PaCO₂ and calculation of V_D/V_T changes during an incremental exercise test in COPD patients. We hypothesized that TcPCO₂ would be acceptable surrogate for PaCO₂, and superior to P_aCO₂-Jones in calculating V_D/V_T for COPD patients during incremental exercise. In this way, we sought to establish that arterial blood sampling may not be required in clinical exercise testing to accurately determine V_D/V_T in diseased lungs. An additional goal of this study was to determine the accuracy and precision of V_D/V_T estimated using P_ETCO₂ in the same participants.

Materials and methods

The study was approved by the Institutional Review Board of the Lundquist Institute for Biomedical Innovation at Harbor-UCLA Medical Center. The details of all experimental procedures, including risks, were explained in detail to the participants prior to obtaining written informed consent.

Study subjects

Ten patients with COPD were invited to participate in this study. Male or female participants between 40 and 90 years of age, current or ex-smoker, no exacerbations within the past 3 months, able to perform acceptable pulmonary function tests and symptom limited cycle ergometry testing, with a post-bronchodilator FEV₁/FVC < 70% qualified for the study (for additional inclusion and exclusion criteria see Supplementary material Appendix).

Study procedures

After a medical history and physical exam was performed, all participants underwent pulmonary function testing [17] including spirometry, lung volumes by body plethysmography and single-breath carbon monoxide diffusing capacity (D_LCO) (Vmax Encore 29, Vyaire Medical, Yorba Linda, CA).

During the second visit a catheter was introduced percutaneously into the radial artery. Following catheter placement, a symptom limited ramp-incremental exercise test [18] was performed on an electromagnetically braked cycle ergometer (Excalibur Sport, Lode, Gröningen, The Netherlands). Tests were administered with 3 min of rest and 3 min of unloaded cycling (UL), followed by an increase in power output at 5 W/min or 10 W/min (for patients with FEV₁≤ 1 L or > 1 L, respectively). Each test continued until the limit of tolerance (see Supplementary material Appendix).

Data collection

An arterial catheter was introduced into the radial artery and blood samples were obtained while the participant breathed through the mouthpiece at rest, during UL, every minute during exercise and for 5 min into recovery [19, 20], samples were collected over ∼5 s. Samples were iced and analyzed within an hour for pH, PaCO₂, PaO₂ and hemoglobin concentration at 37˚C in a certified clinical arterial blood-gas laboratory (ABL820, Radiometer, Copenhagen, Denmark).

Subjects breathed through a mouthpiece with a nose clip in place. Ventilation and gas exchange were obtained breath-by-breath (Vmax Encore 29, Vyaire Medical, Yorba Linda, CA), and the data were subsequently exported in 10-second average bins. The TcPCO₂ was recorded using a noninvasive heated probe (44˚C) attached to the earlobe using double-sided tape and a spring clip. Breath-by-breath values of TcPCO₂ were sampled by the metabolic cart using analog-to-digital conversion and collected together with gas exchange data by the Vmax system. After TcPCO₂ probe placement, at least 5 min was allowed for equilibration prior to data collection (for details see Supplementary material Appendix).

V_D/V_T was calculated using the mass balance equation as follows [21], $$\frac{V_D}{V_T}=1-\frac{863}{\mathit{PaC}{O}_2\text{ * }\frac{{\dot{V}}_E}{\dot{V}C{O}_2}}-\frac{V_{DS}}{V_T}$$ where V˙_E/V˙CO₂ (ventilatory equivalent for CO₂, minute ventilation V˙_E (BTPS), CO₂ output V˙CO₂ (STPD)) was calculated from 10-second bin values matching the blood gas collection. The dead space of the mouthpiece and saliva trap together was 110 ml and this value was used to correct V_D/V_T values for the system dead space (V_Ds). The constant 863 provides the necessary conversions between BTPS and STPD and fractional CO₂ concentration to partial pressure. V_D/V_T was calculated by substituting the PaCO₂ in the equation with each of four alternative values: 1) PaCO₂ from the arterial blood gas (ABG); 2) TcPCO₂; 3) PaCO₂-Jones; and 4) P_ETCO₂.

Statistical analysis

Based on an anticipated standard deviation of ±5 mmHg, we estimated that 73 samples would be necessary to detect at least 2 mmHg difference between methods at an α level of 0.05 and power (1-β) of 0.95 (Medcalc Statistical Software version 19.1.3, Ostend, Belgium; https://www.medcalc.org; 2019). We therefore aimed to collect at least 8 samples from each participant during rest and exercise. The actual sample count was 114 (∼11/subject), resulting an attained power (1-β) of > 0.99 with an effect size of 0.92 (Cohen’s dz).

The Shapiro-Wilk test was used to test for normality. Linear regression was used for calculating Pearson`s correlation coefficients (r²), Bland-Altman analysis was used to describe the agreement between methods using the four expressions of PCO₂ and V_D/V_T measurements [22]. The reliability among the four expressions of PCO₂ and V_D/V_T was assessed using Lin’s Concordance Correlation Coefficient (CCC) [23]. The agreement was categorized as poor < 0.9, moderate 0.90 ∼ 0.95, substantial 0.95 ∼ 0.99 and perfect >0.99 [24]. p < 0.05 was considered statistically significant. The statistical analyses were performed with SPSS version 25.0 (SPSS, Chicago, III., USA), SigmaPlot Ver.13 (Jandel Scientific, San Jose, CA) and the CCC for Repeated Measures package [25] of R Ver.3.5.2 (Vienna, Austria).

Results

Subject characteristics

Demographic and physiological characteristics of the study population are presented in Table 1. Resting pulmonary function shows that the population had moderate to severe expiratory flow limitation, mild hyperinflation and moderately reduced D_LCO. The PaCO₂ values in the 114 rest and exercise arterial blood samples ranged from 26.7 mmHg to 51.0 mmHg. None of the participants reported immediate or remote complications of the catheterization procedures, and there were no other adverse events during the study.

Table 1. Demographics and resting pulmonary function of study population.

	Mean (SD)	% predicted
Age (yrs)*	61.6 (8.0)
M/F (N)	9/1
Race (Black/Caucasian)	4/6
Height (cm)	170.9 (6.2)
Weight (kg)	76.0 (17.0)
BMI (kg/m^2)	25.9 (5.2)
FEV1 (L)	1.81 (0.37)	62.10 (18.54)**
FVC (L)	3.52 (0.61)	89.90 (17.67)
FEV1/FVC	0.53 (0.12)
RV (L)	3.02 (0.93)	133.30 (39.35)***
FRC (L)	3.88 (1.16)	115.70 (31.33)
TLC (L)	6.79 (1.55)	110.00 (21.03)
RV/TLC	0.44 (0.08)
DLCO (ml/mmHg/min)	11.92 (4.92)	44.80 (17.45)

Definition of abbreviations: FEV₁, forced expiratory volume in one second; FVC, forced vital capacity; RV, residual volume; FRC, functional residual capacity; TLC, total lung capacity; D_LCO, diffusing capacity for carbon monoxide; BMI, body mass index.

* Age range 53 to 70 years.

** Reference values for spirometry from NHANES [26].

*** References values from the ECSC [27, 28].

Response of a representative subject

Data from a representative COPD patient’s response to incremental exercise featuring the time course of the four measures of PCO₂ and the four calculations of V_D/V_T is shown in Figure 1. During rest, UL and exercise, there was a remarkable agreement between PaCO₂ and TcPCO₂, and between V_D/V_TABG (V_D/V_T calculated using PaCO₂) and V_D/V_TTc (V_D/V_T calculated using TcPCO₂). During recovery, PaCO₂ fell rapidly, while TcPCO₂ remained elevated for several minutes (for more details see in the Supplementary material Appendix).

Figure 1. V˙̇_E/V˙CO₂, PCO₂, V_D/V_T during a ramp incremental exercise test using 4 different methods of PCO₂ assessment in a representative COPD patient. Ventilatory equivalents (left panel), PCO₂ (middle panel) and V_D/V_T (right panel). The four different expressions of PCO₂ and V_D/V_T are shown. Closed circles: PaCO₂, arterial partial pressure of carbon dioxide from blood gas; V_D/V_TABG, V_D/V_T calculated using PaCO₂. Open circles: V˙_E/V˙O₂, ventilatory equivalent for oxygen; TcPCO₂, transcutaneous partial pressure of carbon dioxide; V_D/V_TTc, V_D/V_T calculated using TcPCO₂. Closed triangles: V˙_E/V˙CO₂, ventilatory equivalent for carbon dioxide; PaCO₂-Jones, estimated partial pressure of carbon dioxide using the Jones equation [7]; V_D/V_T-Jones, V_D/V_T calculated usingPaCO₂-Jones. Open triangles: P_ETCO₂, end-tidal partial pressure of carbon dioxide; V_D/V_TETCO₂, V_D/V_T calculated using P_ETCO₂. The representative subject with FEV₁=40%predicted and FEV₁/FVC = 33% (GOLD 3 COPD). Vertical lines showed the change in exercise phase (from left to right): start of unloaded cycling; start of incremental exercise; lactate threshold and start of recovery.

Group responses

Exercise duration was variable among participants, therefore the total number of arterial samples varied between 7 and 14 per patient during rest, UL and exercise. In total, 10 subjects provided 114 measurements of PCO₂ and V_D/V_T indices during rest, UL and exercise.

There was no statistically significant difference between TcPCO₂ and PaCO₂ at rest, during UL, at the LT and at peak exercise (Table 2, Figure 2). Both TcPCO₂ and PaCO₂ were consistently greater than either PaCO₂-Jones or P_ETCO₂, with P_ETCO₂ being consistently the lowest of the four PCO₂ values (Figure 2). As a result, V_D/V_TTc was not different from V_D/V_TABG, but both were greater than V_D/V_T-Jones (V_D/V_T calculated using PaCO₂-Jones) at rest, with V_D/V_TETCO₂ (V_D/V_T calculated using P_ETCO₂) being significantly lower than the gold standard (V_D/V_TABG) at all stages of test (Table 2, Figure 2, Supplementary material Table S1).

Figure 2. Mean (±SE) V˙_E/V˙CO₂, PCO₂, V_D/V_T during a ramp incremental exercise test using 4 different methods of PCO₂ assessment in 10 COPD patients. Ventilatory equivalents (left panel), PCO₂ (middle panel) and V_D/V_T (right panel).UL, Unloaded cycling. LT, lactate threshold. The four different expressions of PCO₂ and V_D/V_T are shown. Closed circles: PaCO₂, arterial partial pressure of carbon dioxide from blood gas; V_D/V_TABG, V_D/V_T calculated using PaCO₂. Open circles: TcPCO2, transcutaneous partial pressure of carbon dioxide; V_D/V_TTc, V_D/V_T calculated using TcPCO₂. Closed triangles: PaCO₂-Jones, estimated partial pressure of carbon dioxide using the Jones equation [7]; V_D/V_T-Jones, V_D/V_T calculated using PaCO₂-Jones. Open triangles: P_ETCO₂, end-tidal partial pressure of carbon dioxide; V_D/V_TETCO₂, V_D/V_T calculated using P_ETCO₂. *p < 0.05, **p < 0.01, ***p < 0.001 vs. PaCO₂ or V_D/V_TABG. #p < 0.05, ##p < 0.01, ###p < 0.001 vs. LT (for PCO₂) or peak (for V_D/V_T).

Table 2. Ventilatory, gas exchange and blood gases at rest and during incremental exercise in COPD patients.

	Rest	UL	LT	Peak
Work Rate (W)	–	0 ± 0	40 ± 16	78 ± 24
Heart Rate (beats/min)	75 ± 17	90 ± 12	104 ± 13	132 ± 13
V̇O2 (L/min)	0.33 ± 0.10	0.79 ± 0.35	0.97 ± 0.27	1.57 ± 0.55
V̇E/V̇CO2	52.9 ± 16.3^###	41.6 ± 15.0	36.4 ± 14.1	38.5 ± 17.4
pH	7.41 ± 0.03	7.41 ± 0.03	7.37 ± 0.05	7.35 ± 0.08
PaO2 (mmHg)	88.8 ± 26.0	78.6 ± 15.1	78.7 ± 16.3	76.5 ± 18.5
PaCO2 (mmHg)	38.4 ± 5.0	38.8 ± 5.5	39.2 ± 6.7	38.0 ± 7.2
TcPCO2 (mmHg)	39.1 ± 4.1	39.8 ± 4.6	41.0 ± 5.1	40.0 ± 6.3
PaCO2-Jones (mmHg)	34.2 ± 4.9***^##	35.8 ± 6.6**	37.9 ± 6.9	36.4 ± 8.3*
PETCO2 (mmHg)	31.9 ± 5.4***^##	33.7 ± 7.3***	35.8 ± 7.7***	34.3 ± 9.3***
VD/VTABG	0.39 ± 0.11^#	0.32 ± 0.13	0.32 ± 0.10	0.27 ± 0.14
VD/VTTc	0.40 ± 0.12^##	0.34 ± 0.13	0.31 ± 0.11	0.31 ± 0.12
VD/VT-Jones	0.34 ± 0.12**^#	0.28 ± 0.07	0.27 ± 0.06	0.25 ± 0.07
VD/VTETCO2	0.30 ± 0.13***^#	0.24 ± 0.06***	0.21 ± 0.10***	0.20 ± 0.05***

Definition of abbreviations: UL, unloaded cycling; LT, lactate threshold; V˙O₂, oxygen uptake; V˙_E/V˙CO₂, ventilatory equivalent for carbon dioxide; pH: potential of hydrogen; PaO₂, arterial partial pressure of oxygen; PaCO₂, arterial partial pressure of carbon dioxide; ABG, arterial blood gas; TcPCO₂, transcutaneous partial pressure of carbon dioxide; P_ETCO₂, end-tidal partial pressure of carbon dioxide; PaCO₂-Jones, estimated PaCO₂ using the Jones equation [7]; V_D/V_TABG, V_D/V_T calculated using PaCO₂; V_D/V_TTc, V_D/V_T calculated using TcPCO₂; V_D/V_T-Jones, V_D/V_T calculated using PaCO₂-Jones; V_D/V_TETCO₂, V_D/V_T calculated using P_ETCO₂.

Data are mean ± SD.

*p < 0.05, **p < 0.01, ***p < 0.001 vs. PaCO₂ or V_D/V_TABG at the same exercise stage.

# p < 0.05, ## p < 0.01, ### p < 0.001 vs. lactate threshold (for PCO₂) or peak (for V_D/V_T).

Comparison of PCO₂ and V_D/V_T using transcutaneous measurements vs. ABG

Analyses were performed that included all data points during rest, UL, at the LT and peak exercise. Data in the recovery phase (see Supplementary material Appendix) were not included in the validation because TcPCO₂ did not track well the rapid changes observed in PaCO₂ during the first 4 min of recovery (Figure 2, Supplementary material Figures S1 and S2 and Table S1).

Figure 3(A,B) presents the comparison of TcPCO₂ with PaCO₂. Across the range of PaCO₂ values observed (between 27.6 and 51.0 mmHg), there was a strong positive correlation between TcPCO₂ and PaCO₂ (r² = 0.907). There was a significant positive intercept (Y₀ = 6.56 mmHg) and a slope slightly less than one (a = 0.86) (p < 0.001 for both) (Figure 3A). Bland-Altman analysis revealed a small positive bias (0.86 ± 1.60) mmHg with lower and upper limits of agreement between −2.28 and 3.99 mmHg (Figure 3B). On average, at low PaCO₂ (<35 mmHg), the transcutaneous device slightly overestimated PaCO₂. Lin’s concordance correlation analysis (CCC) confirmed that the agreement between PaCO₂ and TcPCO₂ was moderate (CCC = 0.941; 95% confidence interval = 0.838–0.979)) (Table 3).

Figure 3. Regression and Bland-Altman analysis of PCO₂ and V_D/V_T using ABG and transcutaneous measurements during rest and incremental exercise in 10 COPD patients. A. Scatter plot and linear regression of TcPCO₂ vs. PaCO₂. B. Bland-Altman plot of agreement between TcPCO₂ and PaCO₂. C. Scatter plot and linear regression of V_D/V_TTc vs. V_D/V_TABG. D. Bland-Altman plot of agreement between V_D/V_TTc and V_D/V_TABG.

Table 3. Lin's Concordance Correlation Coefficients (CCC) and Bland–Altman analyses among the four different expressions of PaCO₂ and V_D/V_T during rest and incremental exercise in COPD patients.

	CCC (95% CI)	Bland-Altman limits of agreement analyses
		Lower limit of agreement (95%CI)	Mean difference (95%CI)	Upper limit of agreement (95%CI)
TcPCO2 vs. PaCO2	0.941 (0.838 ∼ 0.979)	−2.276 (-2.790∼-1.763)	0.857 (0.560 ∼ 1.153)	3.990 (3.477 ∼ 4.504)
PaCO2-Jones vs. PaCO2	0.831 (0.588 ∼ 0.937)	−8.367 (-9.341∼-7.392)	−2.420 (-2.982 ∼ 1.857)	3.527 (2.553 ∼ 4.502)
PETCO2 vs. PaCO2	0.718 (0.411 ∼ 0.879)	−11.164 (-12.272∼-10.057)	−4.404 (-5.044∼-3.765)	2.356 (1.248 ∼ 3.464)
VD/VT Tc vs. VD/VTABG	0.966 (0.914 ∼ 0.987)	−0.034 (-0.041∼-0.027)	0.012 (0.007 ∼ 0.016)	0.057 (0.050 ∼ 0.065)
VD/VT-Jones vs. VD/VTABG	0.760 (0.487 ∼ 0.897)	−0.135 (-0.151∼-0.119)	−0.036 (-0.046∼-0.027)	0.063 (0.047 ∼ 0.079)
VD/VTETCO2 vs. VD/VTABG	0.448 (0.131 ∼ 0.681)	−0.204 (-0.225∼-0.182)	−0.075 (-0.087∼-0.063)	0.054 (0.033 ∼ 0.075)

Figure 3(C,D) presents the comparison of V_D/V_T using TcPCO₂ or PaCO₂ for calculation. The linear regression of V_D/V_TTc and V_D/V_TABG shows a strong correlation (r²=0.958). The intercept was zero (Y₀ = 0.00) and the slope of the regression is nearly on the line of identity (a = 1.05, p < 0.001) (Figure 3C). Bland-Altman analysis showed a minimal positive bias (0.01 ± 0.02), and a narrow limit of agreement (-0.03 to 0.06) (Figure 3D). CCC analysis confirmed the substantial agreement between V_D/V_TTc and V_D/V_TABG (CCC = 0.967 (95% confidence interval = 0.914–0.987)) (Table 3).

Comparison of PCO₂ and V_D/V_T using jones equation vs. ABG

The linear regression of PaCO₂-Jones with PaCO₂ had a negative intercept (-3.62 mmHg) with a positive slope close to the line of identity (1.03, p < 0.001) (Figure 4A). There was a negative bias (-2.42 ± 3.03) mmHg with wide limits of agreement of −8.37 to 3.53 mmHg and the regression coefficient was r²=0.755 (p < 0.001) (Figure 4B). CCC was poor (CCC = 0.832 (95% confidence interval = 0.588–0.937)) (Table 3).

Figure 4. Regression and Bland-Altman analysis of PCO₂ and V_D/V_T using ABG and Jones equation during rest and incremental exercise in 10 COPD patients. A. Scatter plot and linear regression of PaCO₂-Jones vs. PaCO₂. B. Bland-Altman plot of agreement between PaCO₂-Jones and PaCO₂. C. Scatter plot and linear regression of agreement between V_D/V_T-Jones and V_D/V_TABG. D. Bland-Altman plot of agreement between V_D/V_T-Jones and V_D/V_TABG.

In the V_D/V_T comparisons (Figure 4C,D), there was a positive intercept (0.08) and slope (0.63) of the regression line with an r²=0.793. Figure 4C shows that the only region where the V_D/V_T-Jones agreed with the V_D/V_TABG is where V_D/V_TABG was ∼0.25; in other words where dead-space was in the normal range. The Bland-Altman analysis confirmed a negative bias (-0.04 ± 0.05) with limits of agreement between −0.14 and 0.06 (Figure 4D). CCC displayed poor agreement (CCC = 0.760 (95% confidence interval = 0.487–0.897)) (Table 3).

Comparison of PCO₂ and V_D/V_T using end-tidal PCO₂ vs. ABG

A significant deviation from the line of identity with a good correlation (r²=0.755, p < 0.001) between P_ETCO₂ and PaCO₂ with a negative intercept (Y₀ = -10.12, p < 0.001) and a slope greater than one (a = 1.15, p < 0.001) (Figure 5A). Bland-Altman analysis showed a systematic bias (-4.40 ± 3.45) mmHg, with limits of agreement between −11.16 and 2.36 mmHg. These suggested that PaCO₂ was 4.40 ± 3.45 mmHg greater than P_ETCO₂ (p < 0.001) (Figure 5B). CCC showed poor agreement between P_ETCO₂ and PaCO₂ (CCC = 0.718 (95% confidence interval = 0.411 ∼ 0.879)) (Table 3).

Figure 5. Regression and Bland-Altman analysis of PCO₂ and V_D/V_T using ABG and end -tidal CO₂ during rest and incremental exercise in 10 COPD patients. A. Scatter plot and linear regression of P_ETCO₂ vs. PaCO₂. B. Bland-Altman plot of agreement between P_ETCO₂ and PaCO₂. C. Scatter plot and linear regression of V_D/V_TETCO₂ vs. V_D/V_TABG. D. Bland-Altman plot of agreement between V_D/V_TETCO₂ and V_D/V_TABG.

In the V_D/V_T comparisons (Figure 5C,D), there was a positive intercept (0.08) and a very shallow slope (0.50) with r²=0.610 (p < 0.001). This suggested a very narrow region could be used for V_D/V_T estimation (V_D/V_T around 0.2 ∼ 0.3, Figure 5C). Bland-Altman analysis showed a negative bias (-0.07 ± 0.07) with limits of agreement between −0.20 and 0.06 (Figure 5D). CCC showed poor agreement between V_D/V_TETCO₂ and V_D/V_TABG (CCC = 0.448 (95% confidence interval = 0.131 ∼ 0.681)) (Table 3).

Discussion

Since the transcutaneous PCO₂ measurement was approved for clinical use, the comparisons of TcPCO₂ with other methods had been examined [16, 29, 30]. The gold standard method to determine efficiency of gas exchange (V_D/V_T) during an exercise test requires direct sampling of the arterial blood gases to determine PaCO₂ and application to the Bohr equation with Enghoff modification [6]. However, arterial blood gases can be accompanied by patient discomfort, transient hyperventilation, poor patient acceptance, and complications such as hematoma, bleeding, or infection [31].

A noninvasive method to accurately and precisely determine PaCO₂ would be of great benefit to cardiopulmonary exercise testing for clinical assessment or research to better understand exercise induced changes in V_D/V_T and gas exchange in efficiency. Jones et al. [7] formulated an equation to estimate PaCO₂ from measurements of P_ETCO₂ and V_T, and validated it in healthy participants. However, the authors cautioned that the equation would be invalid under circumstances of V̇_A/Q̇ abnormality, which is an underlying cause of gas exchange inefficiency in many patients referred for clinical exercise testing. Theoretically, central shunt (patent or opening foramen ovale during exercise) can influence the difference between PaCO₂ and P_ETCO₂. Therefore, we sought to determine whether a noninvasive TcPCO₂ measurement would provide sufficient accuracy and precision to understand exercise V_D/V_T in patients diagnosed with COPD in the exercise physiology laboratory. It should be noted that V˙_E/V˙CO₂ is often referred to as “ventilatory efficiency”, and is related to V_D/V_T in clinical exercise testing [9]. However, V˙_E/V˙CO₂ is not an isolated measure of pulmonary abnormalities, as it also depends on the instantaneous value of PaCO₂. V_D/V_T provides a purer measure of lung gas exchange abnormality, isolated from differences in ventilatory control that act to influence PaCO₂.

Our results indicate moderate concordance between TcPCO₂ and P_aCO₂ over a wide range of PaCO₂ (from mid 20 s to low 50 s mmHg), and across all exercise testing phases except for recovery. It also showed substantial concordance for the resulting V_D/V_T calculated using V̇_E/V̇CO₂ from pulmonary gas exchange measurements and TcPCO₂. We confirmed that concordance of TcPCO₂ with PaCO₂ to measure V_D/V_T during exercise in COPD was stronger than either the Jones equation or using P_ETCO₂ alone.

Clark et al. suggested that in order to be able to follow relatively fast changes in PaCO₂ (and PaO₂) the noninvasive blood gas monitoring analyzer should have a response time (10–90%) about 20 s [32] or an order of magnitude faster than changes anticipated to be measured. The response time of the transcutaneous device with the current Severinghaus CO₂ electrode is between 60 and 90 s, depending on the temperature of the sensor. Carter demonstrated that the TcPCO₂ response time was decreased by heating the sensor [16], which was the rationale behind our selection of probe temperature of 44°C. Heating the sensor helps to ensure constant response time and has the added benefit of local vasodilation, which results in a capillary PCO₂ that is closer to the arterial PCO₂. This appears to be adequate during rest and incremental exercise testing. However, our data shows clearly that when PaCO₂ changes rapidly, as it does in recovery when a relative hyperventilation is operative, PaCO₂ changes were not followed well by the transcutaneous electrode. The TcPCO₂ values were significantly different from PaCO₂ from the 2^nd to 4^th minute of recovery (Table 2 and Supplementary material Table S1 and Figures S1 and S2). This implies that the response time of the TcPCO₂ electrode is slow relative the rate of change of the signal (PaCO₂), and/or the perfusion or diffusion altered dramatically in recovery. For these reasons we did not include samples from recovery in our analysis and do not recommend that transcutaneous measurements be used to assess V_D/V_T in recovery phase of exercise. We also caution against interpretation of V_D/V_T in subjects where PaCO₂ is expected to change abruptly during exercise, such as exercise-induced shunts in pulmonary hypertension (opening of a patent foramen ovale) or in subjects with dramatic and frank hyperventilation in exercise such as in mitochondrial myopathy. In these conditions other features of the cardiopulmonary exercise test data can be used to interpret the validity of V_D/V_T based on TcPCO₂ measurements.

Another goal of our study was to compare the accuracy and precision of V_D/V_T during exercise in COPD subjects determined using TcPCO₂ with the gold standard of direct PaCO₂ measurement. We found substantial concordance of V_D/V_T from TcPCO₂ and PaCO₂ over a range of V_D/V_T of ∼20% to ∼50%. Our analysis illustrated that, just as the values of TcPCO₂ and PaCO₂ were very similar, the V_D/V_T calculated from these two variables were within acceptable agreement during rest and exercise.

The general trend of V_D/V_T changes as exercise intensity increases is characteristic and somewhat different from PaCO₂ changes. PaCO₂ starts to decrease shortly after the lactate threshold (LT) (due to relative hyperventilation), but this does not result in appreciable change in V_D/V_T. Our data confirm that when using either the Jones equation or P_ETCO₂ alone the V_D/V_T concordance was poor in COPD patients. These two estimates of V_D/V_T are approximately accurate only when V_D/V_T is normal, otherwise they greatly underestimate the true value.

We considered the effects of temperature on both the TcPCO₂ probe and the blood gas analyzer values. The transcutaneous CO₂ device probe makes measurements at a temperature of 44 °C but corrects the measured TcPCO₂ values to 37 °C. Similarly, the blood gas analyzer conducts measurements at 37 °C, so no temperature correction is required to compare the two modalities for their PaCO₂ measurements. Also, the change in body temperature during a brief exercise study is small (< = 1 °C)y where temperature during exercise was studied during a 30 min exercise session, the average change in body temperature of the subject was very small (on average only 1 °C), and the resultant PaCO₂ changes were very small (≤1.6 mmHg) [33]. All of our exercise studies had a duration than 15–20 min in total duration.

Limitations

This is a single center study, and the results may not be fully generalizable to other centers. Also, as all participants in this study have COPD, it would be optimal to test the device in other pathological states that affect gas exchange efficiency, for example subjects with pulmonary hypertension, interstitial lung disease, or heart failure, where assessing the adequacy of pulmonary gas exchange might be important for diagnosis, assessing therapeutic interventions and outcomes. This necessity of further validation seems to be particularly relevant in conditions where peripheral microcirculatory abnormalities might affect the skin or the perfusion under the probe, e.g., in advanced stages of heart failure or scleroderma associated pulmonary hypertension.

Another limitation is that we have not corrected for the response kinetics of the transcutaneous electrode, which can influence the agreement of PCO₂ and V_D/V_T values, especially when PaCO₂ is rapidly changing. Our aim was to determine the validity the TcPCO₂ measurement without the added laboratory and participant burden for an “in vivo calibration” by ABG or prior knowledge of the response kinetics of the particular transcutaneous device – thus validating the approach for more widespread application. Our finding that V_D/V_TTc results showed substantial agreement with V_D/V_T at landmark physiological events during rest and exercise (LT, peak exercise) in COPD patients with a wide range of V_D/V_T supports this contention. In the participants studied, changes in PaCO₂ during incremental exercise were slow enough that the response time of the transcutaneous electrode did not impair accuracy of V_D/V_T measurement except during recovery.

Finally, we acknowledge that V_D/V_T assessment does not provide a full picture of gas exchange abnormalities in that it does not quantitatively describe the presence and magnitude of shunt and ventilation/perfusion imbalance.

Conclusion

We showed in patients with a range of COPD severity, that a commercially available TcPCO₂ measurement could provide V_D/V_T estimates that share substantial concordance with the gold standard measured by arterial blood gas analysis. V_D/V_T from TcPCO₂ proved in closer agreement than other commonly used alternatives for noninvasive V_D/V_T estimation (such as, the Jones equation or substitution with P_ETCO₂). In vivo calibration with a resting arterial blood gas measurement or correction for measurement response kinetics was not needed to maintain substantial concordance of V_D/V_T estimates during rest and exercise. However, concordance was lost during recovery where PaCO₂ changes rapidly. From the practical point of view, utilizing the noninvasive TcPCO₂ device during incremental cardiopulmonary exercise testing provides the current best noninvasive alternative to measuring PaCO₂ from the arterial blood samples for V_D/V_T assessment.

Author’s contributions

William W. Stringer contributed to the study conception, data collection, data analysis, abstract and manuscript preparation, and is the guarantor of the manuscript. Janos Porszasz, Richard Casaburi, Harry B. Rossiter contributed to study conception. Robert Cao contributed to subject recruitment. Janos Porszasz, Robert Cao, Robert Calmelat, Susan Corey contributed to data collection. Min Cao, Arin Orogian, Fang Lin, Janos Porszasz contributed to data analysis. Arin Orogian, Janos Porszasz contributed to abstract preparation. Min Cao, Janos Porszasz, Harry B. Rossiter, Richard Casaburi contributed to manuscript preparation. All authors approved the final submitted version.

Role of the sponsors

The sponsor had no role in the design of the study, the collection and analysis of the data, or the preparation of the manuscript.

Acknowledgements

We are grateful to Xiuqing Guo and Jingyi Tan, for their help in mathematical and statistical analysis during the data preparation.

Conflicts of interest

None of the authors report any conflict of interest

Additional information

Funding

This study was supported by a grant from the Pulmonary Education and Research Foundation.