Background

This is a corrected version of the previously published article [1]. Arterial blood gas monitoring is crucial for management of patients with respiratory failure [2]. The gold standard technique involves an arterial puncture which is invasive, time-consuming, and only gives results at one point in time [3, 4]. Moreover, the delay in waiting for the results of blood gas analysis does not allow for real-time adaptation of oxygen therapy or mechanical ventilation. Oxygen saturation by pulse oximetry (SpO2) is widely used as a surrogate of arterial oxygen saturation (SaO2) [5]. Similarly, end tidal CO2 (EtCO2) allows for an indirect, but reliable and continuous assessment of arterial pCO2 for mechanically ventilated patients. However, for non-ventilated patients, assessment of EtCO2 is more complex, less accurate, and often impossible. For these patients, the recently recommended [6, 7] transcutaneous monitoring of carbon dioxide (PtCO2) could represent an alternative for immediate and continuous assessment of pCO2. Numerous studies of both children [8, 9] and adults [10–12] have found a good correlation between PaCO2 and PtCO2. Yet in the specific setting of the emergency department (ED) resuscitation room (RR), PtCO2 has been poorly studied. The main objective of this study was to investigate the relationship between measures of PtCO2 and PaCO2 for patients admitted to the ED RR. The secondary objective was to determine the variables that may disrupt the link between PtCO2 and PaCO2.

Methods

Setting

We conducted this single-center prospective observational study from January to June 2014 in the ED of Nîmes University Hospital, France. This study was reviewed and approved by our Institutional Review Board (number: 13/06–02) and was declared to and approved by the national commission for data processing and civil liberties. All patients provided written informed consent. This study is in compliance with the Helsinki Declaration.

Study population

All adult patients admitted to the RR with dyspnoea during business hours (from 9:00 to 17:00, weekend excluded) were included in the study if arterial blood gas measurements were indicated. In our ED, patients are admitted to the RR if they are level 1 or level 2 according to the Canadian Triage and Acuity Scale (CTAS). Thus, patients with dyspnoea are admitted to the RR if they suffered from severe respiratory distress, asthma, or important dyspnoea. Definition of CTAS level 1 and level 2 for dyspnoea are specified in Appendix 1. Exclusion criteria were incorrect installation of the sensor, signal abnormality on the monitor, and backup error on the memory of the device.

Measurement

The PtCO2 measurement was performed by a Stow-Severinghaus sensor (tc Sensor 92 by Radiometer™, Copenhagen, Denmark). The sensor heats skin to a temperature of 44 °C resulting in a dilatation of the capillary bed that allows for diffusion of gases (CO2 and O2) [13]. On the sensor, carbon dioxide reacts with water to form carbonic acid which dissociates into H + and HCO\(_{3}^{-}\), thereby changing pH values. These pH changes are translated into PtCO2 value through the Henderson-Hasselbalch formula [14]. Medical and paramedical staff were trained in the operation and maintenance of the PtCO2 TOSCA monitor (Radiometer™, Copenhagen, Denmark) before the study commenced. For included patients, the PtCO2 sensor was attached to the ear lobe of the patient allowing for continuous measurement of PtCO2. After stabilization of the monitor to obtain a good signal, arterial blood gases and PtCO2 measurements were performed simultaneously. The medical team was blinded to the value of PtCO2 measured.

Outcomes

The primary outcome was concordance between the simultaneous PaCO2 and PtCO2 values. The sample size calculation was based on the anticipated variation in the differences between the measurements and the required precision. Using a previous study [15] for an estimate of the variation between the differences, a sample size of 50 patients gave a precision of ± 0.19 kPa as the limits of agreement. The secondary outcome was to determine the factors that interfere with this correlation. PtCO2 values were automatically saved every ten seconds by the monitor. Medical patient data were collected and entered into an electronic database after initial collection on paper case report forms (CRF). Blood pressure, heart rate, respiratory rate, blood oxygen saturation, Glasgow coma scale, temperature, time to completion of arterial blood gases, catecholamine use, and non-invasive ventilation or tracheal intubation were recorded by the attending physician. Characteristics of patients such as ED arrival modalities, hospital length of stay, and biological data were collected on the CRF.

Statistical analysis

Patient characteristics were described using qualitative (frequencies and percentages) or quantitative variables (means and standard deviations or median with interquartile ranges - depending on type of distribution) where appropriate. The concordance between PtCO2 and PaCO2 was evaluated by linear regression (correlation coefficients) and Bland-Altman analysis, which determined bias, precision, and agreement of PtCO2 and PaCO2, taking the automated analysis in the laboratory as the reference standard. The Pearson correlation coefficient was used to demonstrate the presence or absence of a relationship between PtCO2 and PaCO2. Relationships between measurement differences (|PaCO2- PtCO2 |) and patient characteristics were investigated by regression analysis. Variables related to the difference between PtCO2 and PaCO2 in the univariate analysis (defined by p<.1, forward selection) were further analyzed in a multivariate model (analysis of covariance). We included PaCO2 in this model but did not included pH or PtCO2 to avoid a collinear bias. Overall model fit was assessed using the Hosmer-Lemeshow test. All statistical tests were two sided. A p-value less than.05 was considered significant for all analyses.

Analyses were performed with the use of R 3.0.2 (R Core Team 2013, R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria). The authors had full access to and take full responsibility for the integrity of the data.

Results

Between January 2014 and June 2014, 102 patients were screened for eligibility. Ninety patients were included and analyzed with 104 PtCO2 values (Fig. 1). Table 1 shows the patient characteristics corresponding to the 104 measurements. After linear regression analysis of 104 couples of measurements, we found a significant correlation between PaCO2 and PtCO2 with R2 =.83 (p<.001) (Fig. 2). The linear regression equation between the two variables was PaCO2 = (0.81 × PtCO2) +10.86. The Bland-Altman analysis is shown in Fig. 3. The mean bias was −1.4 mm Hg (± 7.7) and the limits of agreement (bias ± 1.96 SD) between the two techniques were −16.4 mm Hg and 13.7 mm Hg. The Pearson’s correlation coefficient was.94 (95 % CIs [0.87, 0.94]; p<.001). For the group with PaC02 < 60 mm Hg, R2 =.70 (p<.001) and the mean bias was −3.5 mm Hg (± 5.0). For the group with PaC02 > 60 mm Hg, R2 =.57 (p<.001) and the mean bias was 4.1 mm Hg (± 10.2).

Fig. 1
figure 1

Flow diagram

Full size image
Fig. 2
figure 2

Linear regression between transcutaneous PtcCO2 and PaCO2. Regression line is the continuous line, the dotted lines show the 95 % confidence interval

Full size image
Fig. 3
figure 3

Bland-Altman representation of comparison analysis between PaCO2 and PtcCO2 vs means of paired measurements

Full size image
Table 1 Patients’ characteristics
Full size table

In the univariate analysis, the only factor associated with a difference between PaCO2 and PtCO2 was PaO2 (Table 2). In multivariate analysis with three explanatory variables (PaCO2, PaO2, temperature), we found the temperature and the PaO2 to be significantly associated with a large difference between PaCO2 and PtCO2 (Table 2). The higher the temperature of the patient, the greater the difference between PaCO2 and PtCO2 (Fig. 4). We developed this model on a data set of 93 measurements (11 observations were exluded due to missingness). This model had a non-significant Hosmer-Lemeshow chi-square goodness-of-fit statistic.

Fig. 4
figure 4

Linear regression between temperature and difference between PaCO2 and PtCO2 (|PaCO2-PtCO 2|). Regression line is the continuous line, the dotted lines show the 95 % confidence interval

Full size image
Table 2 Relationships between measurement differences (|PaCO2-PtCO2 |) and patient characteristics using univariate and multivariate analysis (ANCOVA)
Full size table

Discussion

To our knowledge, our study is the largest cohort of PtCO2 measurements conducted in the ED. The mean bias was −1.4 mm Hg (± 7.7) and the limits of agreement (bias ± 1.96 SD) between the two techniques were −16.4 mm Hg and 13.7 mm Hg. There was a significant correlation between PaCO2 and PtCO2 (R2 =.83; p<.001). Because most of our patients were non-intubated, our results highlight the feasibility and the potential benefit of measuring PtCO2 since EtCO2 cannot easily be monitored in non-intubated patients. The correlation coefficient in our study was comparable to what was shown in a previous intensive care study (R2 =.86; p<.01) [10]. However, other studies have found a stronger correlation (R2 coefficient, ranged between.91 and.99 [16–18]). We assume this difference is not a consequence of the use of various devices since most of the studies were completed with a Radiometer™device. This difference can be explained by the selection of our patients, as only those in the RR with acute respiratory failure were included. Indeed, the high and extreme PaCO2 values were reported as possibly interfering with the correlation between PtCO2 and PaCO2 [10, 19–21]. In a study by Delerme et al. [11], patients had a lower PaCO2 than in our study (39 mm Hg vs. 46, respectively). Secondly, this difference may result from the use of the device by many physicians. Calibration, sensor placement and latency to reach the plateau value of PtCO2 may differ from one physician to another. Because some operators used the monitor less frequently, this may have led to poorer reproducibility. However, it also reflects our center’s daily practice and this issue may occur with any change of device. Thirdly, we did not correct the arterial blood gases according to the patient’s temperature and this may explain a portion of the increased difference. Finally, our population was more likely to have significant dyspnoea and therefore agitation or diaphoresis leading to movement of the sensor may have led to inaccurate measurements. Indeed, in the Gancel et al. study [12] where the difference was lower, the exclusion criteria were very rigorous. Authors did not study patients with status epilepticus, confusion or agitation. According to the authors, these criteria may have led to the exclusion of some patients with severe hypercapnia.

Our study found that PtCO2 values were generally greater than PaCO2 values. Indeed, our linear regression equation is PaCO2 = (0.81 × PtCO2) + 10.86. This overestimation is in accordance with available literature [10, 22, 23] and may have implications for patients requiring non-invasive ventilation and with no arterial blood gas reference. Thus, the recommendations highlight the need to conduct an arterial blood gas analysis to support the correlation between PaCO2 and PtCO2 values [6]. This issue is important given that the Bland-Altman analysis reveals a poorer correlation for PtCO2 values above 60 mm Hg. The value of the mean bias reported in our study corresponds to those found in the literature (−1.4 to 4.6 mm Hg) [16, 24, 25]. The decrease of the correlation for high PaCO2 values has been previously reported. The accuracy of PtCO2 seems to be better for patients with PaCO2 values below 56 mm Hg [26]. One explanation for this poor correlation is that the clinical manifestations of hypercapnia (excessive sweating and vasodilatation) leads to a lower diffusion of carbon dioxide [26]. In our study, after multivariate analysis, temperature was associated with a poor correlation between PaCO2 and PtCO2 (OR = 3.01; 95 % CIs [1.16, 7.80]; p=.03). The issue that the temperature can influence the correlation has been raised by Rodriguez et al. [27]. Our linear regression analysis revealed that the higher the body temperature, the greater the difference between PaCO2 and PtCO2 values. This poor correlation can be explained by the fact that asthe patient’s temperature increased, the difference between the patient and the temperature sensor (44 °C) decreased, resulting in small changes in local perfusion and production of CO2. This hypothesis follows directly from the operating principle of the sensor [6]. It could also be hypothesized that a high body temperature promotes sweating and vasodilatation making the sensor’s measurement more inaccurate. Finally, a low blood pressure could also be a cause of a poor correlation between PaCO2 and PtCO2 [28]. Unfortunately, this hypothesis cannot be confirmed by our data because few patients had shock criteria. Similarly, the assumption that the pH may explain a poor correlation [21] cannot not be confirmed in our study with the multivariate analysis.

Limitations

Although several studies have found a poor correlation between PaCO2 and PtCO2 in patients with shock who are treated with catecholamines [10], we did not analyze this particular relationship. Indeed, we included few patients with hemodynamic instability requiring the administration of intravenous fluids or vasopressor support. It is therefore difficult to assess the impact of decreased circulation on the correlation between PaCO2 and PtCO2. Several studies have shown that the correlation is not affected by catecholamines but by dermal vasoconstriction secondary to a state of shock [23, 27].

Secondly, body mass index (BMI) was not measured in our study. Several studies reported conflicting conclusions regarding the influence of skin thickness, indirectly estimated by BMI, on the CO2 diffusion to the skin and therefore on the PaCO2 values [10, 23, 25, 26]. However, there is no correlation between BMI and the skin on the earlobe, where the sensor was fixed [29].

Finally, one subject that remains to be explored is the intra-individual correlation. Most of the patients had only one arterial blood gas measurement during their management in the RR, which was inadequate for obtaining intra-individual correlations between different PtCO2 and PaCO2 values. This analysis would be important to predict PaCO2 values from continuous measurement of PtCO2, especially for patients requiring several hours of monitoring [27, 30].

Conclusions

There is a significant correlation between PaCO2 and PtCO2 values for patients admitted to the ED for acute respiratory failure. This correlation is particularly accurate for values below 60 mm Hg. One limiting factor to routine use of PtCO2 measurements in the ED is the presence of hyperthermia.

Key messages

  • There is a significant correlation between PaCO2 and PtCO2 values for patients admitted to the emergency department for acute respiratory failure.

  • This correlation is comparable to that which has been shown in intensive care.

  • One limiting factor to the use of PtCO2 measurements in the ED is the presence of hyperthermia.

Appendix 1: Definition of Canadian triage and acuity scale (CTAS) level 1 and level 2 for dyspnoea

Patients with dyspnoea and CTAS Level 1:

Severe respiratory distress: serious intracranial events, pneumothorax, near death asthma (unable to speak, cyanosis, lethargic/confused, tachycardia/bradycardia, arterial oxygen saturation below 90 %), chronic obstructive pulmonary disease exacerbations, cardiac heart failure, anaphylaxis and severe metabolic disturbances (renal failure, diabetic keto-acidosis).

Patients with dyspnoea and CTAS Level 2:

Asthma: severe asthma defined with a combination of objectives measures and clinical factors which relate to the severity of symptoms, vital signs and history of previous severe episodes. If the forced expiratory volume in 1 second or peak expiratory flow rate are below 40 % predicted or previous best, the patient is considered severe.

Dyspnea: this is subjective and may correlate poorly with lung function or deficits in oxygen uptake and delivery. Depending on the age, previous history and physical assessment one may not be able to distinguish between asthma chronic obstructive pulmonary disease, cardiac heart failure, pulmonary embolism, pneumothorax, pneumonia, croup, epiglottitis, anaphylaxis or a combination of problems.