Exercise testing is one of the best methods for evaluating the pulmonary and circulatory systems from an integrative standpoint; the information it provides cannot be inferred from resting functional assessment. During an exercise test, numerous variables are collected, allowing computation of different indices. Among these, measurement of the dead-space fraction (VD/VT) is indicative of wasted ventilation [1]. Its original computation, based on CO2 balance, was developed by Bohr and later modified by Enghoff, who substituted arterial PaCO2 for mean alveolar PACO2, the accurate measurement of which is difficult. This resulted in the following formulation: VD/VT=1−(863×VCO2)/(PaCO2×VE). This introduced the concept of physiological dead space, with an anatomical component, referring to the fraction of ventilation occurring only in the conducting airways, and an alveolar component, referring to the fraction of alveolar ventilation with absent or underperfused pulmonary circulation. This approach requires invasive sampling and is sensitive to other factors, such as ventilation–perfusion (V˙A/Q˙) heterogeneity and shunt [2,3]. Therefore, both ventilatory variables, such as VT, and mixed ventilatory-metabolic variables, such as VE/VCO2, are important to analyze.

Because interpretation of dead-space dynamics during exercise is an important part of the diagnostic process, accurate determination of PaCO2 is essential. Given the invasiveness of this measurement, the number of evaluations during the test is limited, usually to 2 or 3 snapshots at rest, peak exercise, and ventilatory threshold. However, the dynamics of dead space during exercise and recovery are not captured.

Several substitutes for PaCO2 have been used, such as PETCO2, the end-tidal partial pressure of CO2 in exhaled gas. However, this is not always accurate and may overestimate or underestimate PaCO2 during exercise in healthy subjects [4]. This discrepancy also occurs in patients with cardiovascular and pulmonary disease. Thus, accurate noninvasive measurement of PaCO2 would be very helpful in exercise testing. PtcCO2 enables measurement of PCO2 by means of a cutaneous, noninvasive probe. However, its reliability and accuracy compared with arterial blood sampling are subject to some limitations under specific conditions. Furthermore, cutaneous PaCO2 measurement exhibits a time delay relative to arterial changes (reviewed in [5]); therefore, these features require further assessment if this technique is to be used during exercise testing.

The present study was undertaken to evaluate the accuracy of PtcCO2 compared with PaCO2 measurement and to monitor the continuous dynamics of VD/VT during exercise and recovery in different settings, including healthy volunteers and patients.

Patients were eligible if they were aged 18 years or older and were recruited during routine diagnostic cardiopulmonary exercise testing. Healthy volunteers were also recruited for exercise testing or hyperventilation. Diagnostic assignment is detailed in the supplemental files.

Approval for this study was granted by the Ethics Committee of Versailles, France. Written informed consent was obtained from all patients and volunteers. The study was registered in the NCT database under number NCT03718780.

In all cases, PtcCO2 was measured with a transcutaneous probe (SenTec®, Therwil, Switzerland) clipped onto a vasodilated earlobe, with a probe temperature of 42°C; see supplemental files. Simultaneously, PaCO2 was determined by analysis of samples obtained by vasodilated earlobe puncture (arterialized blood). A few patients in intensive care were originally included for intra-arterial benchmarking; however, the low number of samples obtained (n=17) precluded meaningful analysis.

The accuracy of PtcCO2 measurement was evaluated by comparing paired results, arterialized blood vs transcutaneous measurements, obtained in patients or healthy volunteers during exercise testing; see supplemental files for details. In addition, healthy subjects performed voluntary 2-min hyperventilation at rest to determine the time delay between the induced arterial PaCO2 modification and cutaneous recording.

Continuous dead-space measurement was performed during routine exercise testing. After a PtcCO2 probe calibration/equilibration phase of 10–15min, an incremental cycling exercise test was performed with the following steps: rest, 3min of gentle warm-up, ramp-up with a stepwise workload increase every minute up to maximal exertion, and then 5min of recovery. Arterialized PaCO2 was sampled 4 times: at rest, at ventilatory threshold, at , and at the end of the recovery period. PtcCO2 was recorded throughout the test. Usual ventilatory and cardiac parameters were also recorded.

After completion of the exercise test, ventilatory Medisoft® data strings, generated at the various respiration-driven time points, and 1-second-interval Sentec® PtcCO2 data were processed. Dead space was computed using the Bohr–Enghoff equation, based on PECO2, the partial pressure of carbon dioxide in total mixed expired gas, and PaCO2, with corrections for mouth tubing dead space (VD instr) and PICO2, the inspired partial pressure of CO2. The following equation was used: VD/VT=([PaCO2−PECO2]/[PaCO2−PICO2])−(VDinstr/VT). Computation was performed at each time point at which ventilatory data had been recorded, using, for PaCO2 in the Bohr equation, either of the following: (a) the PtcCO2 readout at the given time point, with or without consideration of the observed lag time; (b) interpolated PaCO2, with interpolation performed using Expair® software and derived from the 4 earlobe samples; or (c) PETCO2 measured at the given time point. For exercise data analysis and to align with arterialized PaCO2 measurements, gas exchange was measured breath by breath and interpolated into 1- and 10-s average bins for all calculations, including VE/VCO2, PtcCO2, and PETCO2 measurements during exercise.

The following statistical assumptions were applied; see also supplemental files. Sixty subjects were required for validation. Comparisons of PCO2 methods were performed using Bland–Altman analysis and Passing–Bablok regression. Statistical analysis of physiological exercise variables involved groups with n≥10 and used a 2-factor mixed-effects model.

To monitor the dynamics of VD/VT during exercise, we compared transcutaneous PCO2 values with arterialized PaCO2, which closely matches arterial values [6]. Our results show that the Sentec® system provides continuous and precise PaCO2 readings. In our study, 72% of samples were within 3mm Hg of PaCO2, and 83% to 86% of samples were within 4mm Hg. Bias was minimal, ranging from −0.41 to +0.06mm Hg. This compares favorably with data from the literature. In one review [5], bias ranged from −6.1 to +3mm Hg, and a recent study [7] reported a bias of −0.7mm Hg, with limits of agreement between −7.8 and +6.4mm Hg. Earlobe probe placement enables more accurate readings [8]; in addition, preliminary earlobe vasodilation may favor cutaneous gas transfer.

Transcutaneous readout introduces a time bias because of CO2 diffusion through the skin. We showed that the total response time was approximately 80s, as measured during a resting hyperventilation test at rest. The literature on this topic is limited; see supplemental files. Other investigators have observed values of the same order of magnitude, namely a 2-min lag time [9,10], or higher values [11], notably with differences in equipment, clinical and technical settings, and probe positioning. Although we accounted for this lag time, the observed difference did not appear to meaningfully affect clinical interpretation, as shown in Fig. 6. This is because changes in exhaled CO2, namely PECO2, have a greater impact on VD/VT during exercise transitions than gradual PaCO2 shifts. However, we cannot exclude the possibility that exercise interferes with PtcCO2 readings, especially around peak exercise and early recovery, as described elsewhere [12].

We observed that PaCO2, after an initial slight upward trend, decreased during the second part of the exercise ramp-up and tended to overshoot during most of the recovery period. This is probably explained by persistent hyperventilation during most of recovery, as suggested by our ventilatory data during recovery; see Fig. 3A and supplemental Table 5. This is a well-known feature [13,14]. In addition, possible VA/Q abnormalities, observed for at least 20min after maximal exercise in healthy volunteers [15], with possible interstitial edema, may also be involved [16,17,18].

In contrast to PaCO2, dead space adjusted rapidly at the very start of exercise, during the first 3min of gentle pedaling, with a sizeable decrease ranging from 24% to 100% of the overall observed dead space decrease; see supplemental table 3. This early decrease at the beginning of very mild exercise has been described previously [19,20]. The range of maximal dead-space decrease was wide, spanning from 76% of the resting value in healthy volunteers to 16–25% of the resting value in patients with COPD or pulmonary fibrosis.

Mechanisms underlying the decrease in VD/VT include increased perfusion in the apices of the lung relative to ventilation, with improved ventilation–perfusion matching during light exercise, followed by deterioration during heavy exercise. In addition, VT increases to a much greater extent than the small increase in anatomical dead space resulting from larger airway diameter at higher ventilation [1,21,22]. This is not contradicted by the VD/VT stability observed in our resting hyperventilation maneuvers; indeed, besides the increase in VT, a tripling of respiratory rate was also noted. This suggests that during exercise, other factors, possibly hemodynamic, may intervene. In patients with abnormal ventilation–perfusion ratios, such as those with emphysema, cardiac disease [23], or pulmonary vascular disease [24], VD/VT is often elevated at rest and, despite a small decrease during the early phase of exercise probably related to the increase in VT, may fail to continue decreasing or may even increase as exercise progresses. Patients with COPD may also have early endothelial damage and, later, reduced pulmonary capillary volume [25].

After peak exercise, dead space began to rise again 1min later, despite the continued downward trend in PaCO2, reflecting incomplete reduction in ventilation with persistent hyperventilation. It follows from these disconnected trends that dead space reversal is driven mainly by modifications in the PECO2 component of the Bohr equation, with both VT (Fig. 3B) and VE/VCO2 (Fig. 3D) initiating a progressive return toward resting values. The speed and extent of recovery depend both on the magnitude of the changes induced by the exercise challenge, with deeper adaptive changes requiring longer readjustment, and on the adaptability of the pulmonary microvasculature according to etiology. Some patient groups, including COPD and deconditioning, almost recovered their resting dead space value at the 5-min recovery mark. In the other groups, as in healthy volunteers, dead space had not returned to pretest values. Our data show that many of the changes, especially ventilation, were not fully reversed by the end of 5min. The presumed persistence of postexercise V˙A/Q˙ abnormalities in healthy volunteers may also apply to patients. Additional mechanisms may also be involved in some groups. For example, groups with sizeable hyperventilation may have a persistent abnormal ventilatory pattern independent of the pulmonary V˙A/Q˙ ratio.

This study confirms that noninvasive PETCO2 is not an adequate substitute for PaCO2 measurement. Bias ranged from underestimation of PaCO2 to overestimation, particularly in healthy volunteers, with a gray zone for some patient groups. Conversely, PtcCO2 appears to be a good surrogate for PaCO2, enabling accurate monitoring of dead space dynamics. However, when comparing dead space data computed with PtcCO2 with those obtained using interpolated values from 4 discrete PaCO2 samples obtained during the test, namely at rest, ventilatory threshold, peak exercise, and the end of recovery, the added value appears marginal because the slow variation in PaCO2 over time makes interpolation reasonably accurate. However, this holds true only when 4 arterial PaCO2 values are sampled throughout the exercise test, including at the end of recovery, which is almost never the case. Nevertheless, continuous transcutaneous measurement can substitute for invasive and potentially painful arterial sampling.

Weaknesses of our study include the limited number of patients in some groups. Although the COPD, hyperventilation syndrome, heart failure, healthy volunteer, and asthma groups were sizeable, other groups were small, with n=3–4 each. This precludes any strong group-specific conclusion for these latter groups. However, no unexpected finding singled out any of these small groups, and, aside from nuances, the main findings were shared across all groups.

In summary, continuous PtcCO2 monitoring during exercise testing is feasible and accurate, unlike PETCO2, and enables dynamic dead-space computation. Transcutaneous lag time does not meaningfully affect the VD/VT pattern. In all tested patients and subjects, PaCO2 decreased significantly during exercise and beyond, most often without returning to baseline values, driven by persistent hyperventilation. In contrast, dead space began to adjust immediately at exercise onset, even at low intensity, and began readjusting early after exercise cessation, also without returning to pretest values. The magnitude of dead-space adjustments ranged from those observed in healthy volunteers at one end of the spectrum to those observed in patients with COPD at the other, without any particular differentiating feature among the various patient groups. Ventilatory, vascular, and metabolic adjustments induced by maximal exercise require more than 5min for full reversal.

Serge Kouzan: conceptualization, investigation, methodology, project administration, resources, supervision, validation, writing – original draft, and writing – review and editing. Léo Blervaque: conceptualization, data curation, formal analysis, methodology, resources, software, visualization, and writing – review and editing. Vincent Peigne: conceptualization, investigation, methodology, resources, validation, and writing – review and editing. Cécile Ricard: data curation, formal analysis, methodology, resources, software, validation, visualization, and writing – review and editing. Fabienne Prieur: data curation, investigation, project administration, resources, and writing – review and editing. Pierantonio Laveneziana: conceptualization, methodology, supervision, validation, visualization, and writing – review and editing.

The authors declare that no artificial intelligence program or tool was used in the writing of this article or in the production of tables, figures, legends, or references.

None declared. However, a specific biological device, a transcutaneous PtcCO2 measurement probe, was lent by a homecare company, Agiradom®, which did not participate in project management, data analysis, manuscript writing, approval of any version of the submitted manuscript, or the decision to publish.

PL reports personal fees and support for attending meetings and/or travel from Chiesi, Trudell International, and Elivie, outside the submitted work. SK, LB, VP, CR, and FB report no competing interests.