Background
The partial pressure of blood gases can be estimated through measurement of dissolved gases that diffuse to the skin surface [
1,
2]. Measurement of transcutaneous PO
2 (PtcO
2) and PCO
2 (PtcCO
2) requires local heating of the skin, which dilates vessels and increases arterial blood supply to the skin capillary bed under the sensor, resulting in accelerated gas diffusion [
3,
4]. In clinical practice, this method is widely used to assess pulmonary gas exchange function in infants and children, and in adults with acute or chronic respiratory failure [
5‐
7]. It may also be applied to monitoring the condition of patients on mechanical ventilation and managing limb ischemia [
8‐
10].
Although previous studies have investigated the relationship between PtcCO
2 and PaCO
2 over time, the time courses of transcutaneous data for the estimation of arterial blood gas analysis (BGA) are not well characterized [
11‐
13]. Various factors may influence the time course of agreement including the response speed of the electromechanical gas measuring system, the speed of skin heating, and the time to equilibration of gases [
3,
13‐
15]. This information would allow physicians to choose a convenient (early) time-point for transcutaneous BGA for the estimation of arterial BGA and an optimal time-point for increased accuracy.
A previous study suggested that the correlation between arterial BGA and transcutaneous BGA data via sensors on the chest is stronger than that observed via sensors on the arm; however, this report involved only anesthetized adult patients [
16]. The most commonly recommended sensor location, according to the guidelines established by the American Association for Respiratory Care, is the upper chest followed by the lateral side of the abdomen, chest, buttock, inside of the upper thigh, forearm, the zygomatic bone, the ear lobe, cheek, or the forehead in neonates and small pediatric patients [
9]. In the beginning, we compared data obtained from sensors placed on the chest, forearm, earlobe, and forehead in spontaneously breathing adults. In the early stage of the study, we decided to use only a chest or forearm sensor (data are shown later).
Arterial blood samples are drawn with the patient being in a steady state [
17]. The usual clinical practice for arterial BGA in fully conscious patients involves a single arterial puncture performed after a waiting period of 20–30 min [
17,
18]. The procedure of arterial puncture may cause pain and hyperventilation, thereby altering subsequent arterial BGA data due to respiratory alkalosis [
17]. In mechanically ventilated patients, the stability after a change in F
IO
2 is reached between 10 and 30 min depending on the physiological and pathophysiological conditions of the patient [
19]. Therefore, in the present study, the arterial BGA data with one-time arterial puncture after a waiting (resting) period of 30 min in the supine position was defined as the gold standard blood gas data.
We evaluated the transcutaneous BGA data at 1-min intervals comparing the final goal of arterial BGA at 30 min. This novel approach will answer the following questions: “From which time point are the transcutaneous BGA data meaningful?” and “How accurately are the current transcutaneous BGA data predicting arterial BGA?” In addition, the results of the subgroup analyses which may help to understand transcutaneous BGA, are shown. Finally, we discuss the most important subgroup (i.e., severe hypercapnia with PaCO2 > 50 mmHg) and recommend a reasonable time-saving step for the accurate diagnosis of these patients.
Discussion
By comparing the agreement between minutely obtained transcutaneous BGA data and the final answer data of arterial BGA at 30 min, we obtained the following findings. Firstly, the sensors placed on the chest and forearm are equally preferred. Secondly, the method to predict PaCO2 at 30 min is to initially measure PtcCO2 at 4 min without bias, and observe PtcCO2 at 8 min or later considering a bias of 4–5 mmHg. Thirdly, although PtcCO2 is useful, it cannot completely replace the actual levels of PaCO2 due to occasional large PCO2 bias. Fourthly, the subgroup analyses showed that gender, younger age, PaCO2 levels, and PaO2 levels affected PO2 and/or PCO2 biases. Fifthly, a reasonable step to reach accurate diagnosis of PaCO2 > 50 mmHg using transcutaneous BGA data was recommended. Finally, we showed that the prediction of PaO2 by PtcO2 was unrealistic in Asian adults.
Previously, it was reported that the 1.96SD between venous PCO
2 and PaCO
2 was 15.0 mmHg [
23]. Venous PCO
2 is occasionally used as a surrogate with a bias of 5 mmHg. Our approach enabled to answer the question of “From which time point are the PtcCO
2 data meaningful?” The answer is “From 4 min.”, because the limits of agreement between PaCO
2 and PtcCO
2 at 4 min or later were ± 13.6 mmHg or narrower. Of note, they were narrower than the limit of agreement (± 15.0 mmHg) between PaCO
2 and venous PCO
2. By waiting longer, we can obtain more accurate PtcCO
2 data for the estimation of PaCO
2. Several studies have indicated that PtcCO
2 is more accurate than end-tidal PCO
2 as a surrogate measure of PaCO
2 [
24‐
28]. While 1.96SD data between end-tidal PCO
2 and PaCO
2 ranged from 6.9 to 14.4 mmHg, 1.96SD data between PtcCO
2 and PaCO
2 ranged from 4.6 to 10.4 mmHg. The PtcCO
2 data at 12–13 min or later were within the acceptable clinical range of agreement for PtcCO
2 (± 7.5 mmHg) recommended in the guideline of the American Association for Respiratory Care [
9].
As a whole, the prediction of PaCO
2 is possible. It involves initial measurement of PtcCO
2 at 4 min without bias, and observation of PtcCO
2 at 8 min or later considering a bias of 4–5 mmHg. In a steady state, PtcCO
2 is higher than PaCO
2 because the former is an epidermal parameter which does not exclusively reflect arterial blood, and CO
2 is produced by living epidermal cells [
7,
29,
30].
The 1.96SD between PtcO
2 and PaO
2 displayed a continual decline without an obvious plateau at 30 min. Even the minimal limit of agreement of ±24.0 mmHg at 30 min is not negligible in clinical practice. Therefore, PtcO
2 is not an appropriate substitute for PaO
2. Kesten et al. reported that the 90% response speed of PtcCO
2 in this system was approximately three times faster than that of PtcO
2 [
14]. The Krogh’s constants of diffusion for O
2 and CO
2 in water and aqueous tissues may be important to understand the difference between these gases [
15]. In water and aqueous tissues, the Krogh’s constant of diffusion for CO
2 has been reported to be 20–25 times higher than that for O
2.
A change in the baseline level with time is termed “drift” [
31]. The calibration was performed prior to measuring each subject according to the protocol provided by the manufacturer. The duration of the measurement was only 30–40 min per subject; therefore, the “drift” effect was considered negligible.
A few previous studies have investigated the relationship between PtcCO
2 and PaCO
2 over time [
11‐
13]. Fuke et al. compared PaCO
2 via an arterial catheter and PtcCO
2 over time (
n = 6), yielding evaluations of individual agreements [
11]. Excellent agreement over time was shown in three subjects. Both Cuvelier et al. [
12] and Storre et al. [
13] compared PaCO
2 via an arterial catheter and PtcCO
2 over time (
n = 12 and
n = 10, respectively), demonstrating parallel changes without any description of concordance over time. The present noninvasive study without arterial catheterization is in line with the actual clinical practice for spontaneously breathing patients, and provides data from a larger sample of subjects compared with previous studies [
11‐
13]. The subgroup analyses revealed that gender and younger age affected the biases. Further investigation is necessary to confirm this observation. The absolute values of biases (for both PCO
2 and PO
2) were larger in the PaCO
2 < 31 mmHg group than in the normal group. Arterial vasoconstriction by hyperventilation may be involved in this phenomenon [
23]. If PaCO
2 is low, the PCO
2 bias may increase and underestimation of hyperventilation might occur. However, Bendjelid et al. reported that the PaCO
2 level did not affect the PCO
2 bias (
n = 55, Caucasians 85%) [
32]. The absolute values of biases (for both PCO
2 and PO
2) were smaller in the hypoxemia group than in the normal group. Hypoxic vasodilation may be involved in this phenomenon [
33].
Another limitation of the study is that arterial BGA was performed only at 30 min. However, it is worth performing Bland–Altman analysis for the comparison of the single time point arterial BGA data with the minutely obtained transcutaneous data. All gas data were collected during a short period (30–40 min) in the steady state, which was validated by observations that SpO
2 data were almost constant in each subject from the sensor fixation to the arterial blood sampling procedure. The effect of changing body position (e.g., from sitting to supine position) on PaCO
2 has been reported to be smaller than that exerted on PaO
2 [
34,
35]. Bland and Altman compared data from two different peak flow meters which cannot be performed simultaneously [
20].
Correct diagnosis of severe hypercapnia with PaCO2 > 50 mmHg is important to avoid CO2 narcosis. This technology of TCM4 with a Severinghaus electrode is useful in identifying subjects with PaCO2 > 50 mmHg. By performing arterial BGA after detecting PtcCO2 ≥ 50 mmHg during an observation for 12 min, PaCO2 > 50 mmHg can be accurately measured (without exceptions at least in our 13 subjects). We recommend this reasonable step for the efficient use of PtcCO2 data.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.