Article Text

Can transcutaneous carbon dioxide pressure be a surrogate of blood gas samples for spontaneously breathing emergency patients? The ERNESTO experience
  1. Nicolas Peschanski1,2,3,
  2. Léa Garcia1,
  3. Emilie Delasalle2,
  4. Lynda Mzabi1,
  5. Edwin Rouff1,
  6. Sandrine Dautheville1,
  7. Fayrouz Renai1,
  8. Yann Kieffer1,
  9. Guillaume Lefevre4,
  10. Yonathan Freund5,6,
  11. Patrick Ray1,6
  1. 1Department of Emergency Medicine and Surgery, Centre Hospitalo-Universitaire Tenon Saint Antoine, Assistance-Publique Hôpitaux de Paris (AP-HP), Paris, France
  2. 2Department of Emergency Medicine, Centre Hospitalo-Universitaire Rouen, Rouen, France
  3. 3Institut National de la Sante et de la Recherche Médicale U1096, Université de Rouen, Rouen, France
  4. 4Department of Biochemistry, Centre Hospitalo-Universitaire Tenon Saint Antoine, Paris, France
  5. 5Department of Emergency Medicine and Surgery, Groupe Hospitalo-Universitaire Pitié-Salpêtrière, Paris, France
  6. 6DHU Fighting against Ageing and Stress (FAST), Paris Sorbonne Université, Université Paris-06, Paris, France
  1. Correspondence to Dr Nicolas Peschanski, CHU de Rouen, Service des Urgences Adultes, Pavillon Dévé 2, Hôpital Charles Nicolle, 1 rue de Germont 76031, Rouen, Cedex 76000, France; n.peschanski{at}


Background It is known that the arterial carbon dioxide pressure (PaCO2) is useful for emergency physicians to assess the severity of dyspnoeic spontaneously breathing patients. Transcutaneous carbon dioxide pressure (PtcCO2) measurements could be a non-invasive alternative to PaCO2 measurements obtained by blood gas samples, as suggested in previous studies. This study evaluates the reliability of a new device in the emergency department (ED).

Methods We prospectively included patients presenting to the ED with respiratory distress who were breathing spontaneously or under non-invasive ventilation. We simultaneously performed arterial blood gas measurements and measurement of PtcCO2 using a sensor placed either on the forearm or the side of the chest and connected to the TCM4 CombiM device. The agreement between PaCO2 and PtcCO2 was assessed using the Bland–Altman method.

Results Sixty-seven spontaneously breathing patients were prospectively included (mean age 70 years, 52% men) and 64 first measurements of PtcCO2 (out of 67) were analysed out of the 97 performed. Nineteen patients (28%) had pneumonia, 19 (28%) had acute heart failure and 19 (28%) had an exacerbation of chronic obstructive pulmonary disease. Mean PaCO2 was 49 mm Hg (range 22–103). The mean difference between PaCO2 and PtcCO2 was 9 mm Hg (range −47 to +54) with 95% limits of agreement of −21.8 mm Hg and 39.7 mm Hg. Only 36.3% of the measurement differences were within 5 mm Hg.

Conclusions Our results show that PtcCO2 measured by the TCM4 device could not replace PaCO2 obtained by arterial blood gas analysis.

  • respiratory
  • ventilation
  • COPD
  • non invasive
  • qualitative research

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Key messages

  • What is already known on this subject?

  • Transcutaneous carbon dioxide pressure measurements could be a non-invasive alternative to partial pressure of carbon dioxide measurements obtained by blood gas sampling, as suggested in previous studies.

  • What might this study add?

  • Our study evaluates the reliability of a non-invasive device in the ED.

  • It suggests that this cutting-edge technology is not yet able to replace the arterial blood gas measurement, which remains the gold standard of oxygen and carbon dioxide partial pressure measurements in a breathless patient in the ED.


Arterial blood gas (ABG) sampling is important for evaluating patients in an emergency department (ED) setting. It is an invasive procedure that can be uncomfortable for patients and has an inherent significant (albeit uncommon) risk of complications. Non-invasive measures such as venous blood gas, pulse oximetry and end-tidal carbon dioxide measurements are therefore often used.1–6 However, each of these may have shortcomings. Pulse oximetry may be inaccurate in shock and other low flow states.7 Venous pH has good correlation and close approximation with arterial pH.2 Venous and end-tidal partial pressure of carbon dioxide (PCO2) correlate well with arterial PCO2 (PaCO2) when the PaCO2 is normal, but when PaCO2 is elevated in ventilatory failure, the venous and end-tidal PCO2 readings may fail to track PaCO2 closely and yield misleading results.2–6

An alternative approach using transcutaneous PCO2 (PtcCO2) to measure PCO2 has been proposed and studied with promising results.8–10 Admittedly, this technique may be less useful for an initial emergency estimate of PaCO2 than it is for following PCO2 continuously over time in order to evaluate the response to or failure of initial treatment. Our hypothesis was that a non-invasive device fitted to the spontaneously breathing dyspnoeic patient could be a surrogate for the gold standard of ABG measurement for PCO2 monitoring. Our aim was to evaluate the reliability of a new device to measure the PCO2 in an emergency setting.


Study design

This was an observational and prospective study. As medical management was unchanged, waived consent was authorised. The study was approved by the ethics committee (Comité de Protection des Personnes, CHU Pitié-Salpétrière, Paris, France).

Study setting

The study was conducted in two different EDs: one in Tenon Hospital (Paris) and the other in Charles Nicolle Hospital, Rouen (Normandy). Both were urban adult EDs and teaching hospitals with 40 000 and 80 000 annual new patient attendances, respectively.

Inclusion and exclusion criteria

Patients were included if they presented with two of the following criteria for respiratory distress:

  • Respiratory rate (RR) ≥30/min

  • Blood oxygen saturation (SpO2) ≤90% without oxygen

  • Clinical signs: seesaw respiration, intercostal indrawing, pronounced use of accessory muscles, cyanosis, sweating

  • Mottling

  • Systolic arterial blood pressure (BP) ≤90 mm Hg.

The exclusion criteria were: pregnancy, age <18 years or unable to wait 15 min to have blood gas sample (even with oxygen).

Study protocol, measurements and data collection

Trained nurses (or physicians) set the device on the patient and waited 15 min, with or without oxygen, to perform blood gas samples simultaneously with the reading of PtcCO2 on the TCM4 screen.

The TCM4 CombiM (Radiometer, Villeurbanne, France) is an easy to use device. All nurses and doctors received practical training at the Radiometer Company. It uses a heated Clark and Stow-Severinghaus PtcCO2 and PtcO2 electrode. The electrode is placed on a non-hairy part of the skin (chest or forearm) with a patch and two drops of electrolyte contact gel. It is heated at 44°C so that the blood is arterialised. The PtcCO2 and PtcO2 measurements are read after 15 min (the time at which the curves on the monitor reach their plateau). All repeated PtcCO2 measures are done ‘fresh’, with a new sensor used on patients who contributed more than one-paired measurements. We limited the length of time the device was left on the patients to 1 h and changed the electrode if needed during the 4 h of monitoring.

ABG samples were sent to the biochemical laboratory and analysed using the Radiometer ABL 705 (Radiometer) or the Roche Cobas (RocheDiagnostic, Meylan, France). The results were usually available on our server in 1.5 h.

ED senior physicians filled a standard record sheet with all the patient information including age, sex, reason for admission to the ED, medical history, temperature, systolic and diastolic arterial BP, heart rate, RR, SpO2 and oxygen delivery.

Final diagnosis

The final diagnosis was based on the summary of the medical chart during the hospital stay. Medical history, clinical presentation, biomarker findings, CT scanning and/or echocardiography results when available and outcome (discharge or admission, admission to intensive care unit, mortality during the first 48 h) were recorded.

End points

The primary outcome was the percentage of outliers. Outliers were defined as PaCO2/PtcCO2 pairs whose values differed by more than ±5 mm Hg. We considered that a difference between PtcCO2 and PaCO2 of 5 mm Hg or less would make PtcCO2 measurement clinically useful to assess PaCO2. Previous results suggested that there would be less than 5% outliers,11 and thus the sample size calculations suggested the need for 60 patients. According to these previous results and assuming that no outliers would be observed, we calculated that at least 59 patients should be included to obtain an upper limit of the 95% CI  of less than 5% so that the device would be reliable. The secondary end points were bias and precision.

Data analysis

Descriptive data are presented as mean±SD (min−max) and as number and proportion. Mean difference, proportions and correlations are reported as point estimates bounded by 95% CIs. The agreement between PtcCO2 and PaCO2 was evaluated by Bland–Altman plot analysis which determines bias, precision and agreement of PtcCO2 and PaCO2.12 Statistical analysis was performed using NCSS V.6.0 software (Statistical Solutions, Cork, Ireland).


From February 2012 to April 2013, 67 patients were included in the study, 53 in Tenon hospital and 14 in Rouen. The main characteristics of the patients are shown in table 1.

Table 1

Main characteristics of patients (N=67)

The final diagnosis was mainly heart failure, pneumonia or chronic obstructive pulmonary disease (COPD) exacerbation (28%). Most of the patients were admitted to an observation unit and 14 (21%) were transferred to the ICU. Ten patients had non-invasive ventilation and one was intubated. Four patients died in the first 48 h.

Sixty-seven measurements of PtcCO2 and PaCO2 were performed and 64 were analysed. In one case we were unable to obtain the PtcCO2 measurement because of sweat and in two cases the ABG results were not available because of technical failure or interpretation issues (venous blood).

No side effects were observed during the use of the TCM4 CombiM and no patient complained of burning or pain.

The mean difference between PaCO2 and PtcCO2 was +8.4 mm Hg (range −47 to +52) with 95% limits of agreement of −19.5 and 36.3 mm Hg. The 95% CI of the limits of agreement were −25.1 to −15.6 and 32.4 to 41.9, respectively.

With a 5 mm Hg limit for outliers, only 33 measurements (36.3%) were classified as acceptable. The Bland–Altman plot is shown in figure 1.

Figure 1

Concordance between arterial (PaCO2) and transcutaneous (PtcCO2) carbon dioxide pressure.


Our study demonstrates poor agreement between ABG PaCO2 and PtcCO2 using the TCM4 CombiM. Although the agreement between blood gas PaCO2 and PtcCO2 remains similar regardless of the average PaCO2 values, the majority of the measurements were outside the limits of acceptable agreement. Otherwise, the variability is consistent across the Bland–Altman plot as the scatter around the bias line remains similar regardless of the average PaCO2 values. Furthermore, the delay in getting the initial reading because warming of the skin is required for adequate arterialisation of flow is a possible inconvenience factor. Thus, the TCM4 CombiM cannot be considered as a tool for the evaluation of PCO2.

These results are disappointing compared with previous studies performed in the ED (including ours).9–11 Using a different device, the TOSCA 500 (also from Radiometer), Delerme et al found a bias of 1 mm Hg and limits of agreement of −3.4 and 5.6 mm Hg.11 All measurement differences were within 5 mm Hg and 64% were within 2 mm Hg. Our study population was similar, with slightly more hypercapnic patients. Previous studies have suggested that correlation between arterial and non-invasive PCO2 is worse in hypercapnic patients. Using a Radiometer TINA TCM3 device in patients with stable COPD, Cuvelier et al13 found that PtcCO2 was less reliable in patients with PaCO2 >56 mm Hg. The same limitation (ie, hypercapnia) was found by Storre et al14 who used a SenTec Digital monitor in patients admitted for initiation of non-invasive positive pressure ventilation and by Bobbia et al15 using a Stow-Severinghaus sensor (tc Sensor 92; Radiometer, Copenhagen, Denmark). Most of the patients had a PaCO2 >50 mm Hg and the bias was 4.6 mm Hg with limits of agreement of −3.9 and 13.2 mm Hg. Using the TOSCA 500, Gancel et al10 found a bias of 0.1 mm Hg and limits of agreement of −6 and +6.2 mm Hg. McVicar et al,9 using the same device in a similar setting, found a bias of 0.15 mm Hg and limits of agreement of −6.6 and 6.9 mm Hg. In both cases the study included more patients with COPD (57% and 55%, respectively) and more patients with a PaCO2 ≥50 mm Hg.9 ,10

We suggest that the different devices used could explain the discrepancy between the studies. Indeed, in an ED setting, using a Radiometer TCM4 device, Kelly and Klim16 also found poor agreement between PtcCO2 and PaCO2 in patients whose median PaCO2 was 60 mm Hg. In their study the bias was 6.1 mm Hg and limits of agreement were −10.1 and 22.3 mm Hg, which is closer to our results. They found that all patients with a difference between PtcCO2 and PaCO2 of >10 mm Hg had a PaCO2 of 60 mm Hg or more; 20% of the patients included in their study had a difference between PtcCO2 and PaCO2 of more than 10 mm Hg. In our study 37% of the patients had a difference between PtcCO2 and PaCO2 of more than 10 mm Hg. However, Urbano et al17 found similar results using three different devices (Tosca 500, TINA 3 and Sen Tec) in critically ill children.

Hemodynamic instability has been discussed in the literature as a potential cause of poor reliability.18 ,19 However, we could not test this hypothesis because none of our patients had a systolic BP lower than 90 mm Hg.

Limitations of the study

The size of our sample is small and therefore caution is needed in the interpretation of our results. A further limitation of the PtcCO2 measurement was the device itself. We had a few problems with changing the membrane of the TCM4 CombiM every 2 weeks so only the main investigator was allowed to change it. This issue would probably be alleviated with time and familiarity with the device. Because this device needs 15 min to provide an accurate measurement of PtcCO2, some physicians were reluctant to use it. Even if the device is easy to use, some nurses were reluctant to use it in an urgent situation. In a few cases we could not use it, mostly in sweating patients. However, patients seemed really enthusiastic with the idea of a non-invasive device, especially those with COPD, and no patient expressed discomfort with it. Another limitation is the study design itself. It was only a two-centre study with inclusion authorised only according to the judgement of the physicians in charge (only 67 patients in 1 year).


PtcCO2 measurement using the TCM4 CombiM is not a reliable method to evaluate PaCO2 in dyspnoeic patients who present to the ED. Further studies should evaluate other devices to assess PtcCO2.


Supplementary materials

  • Abstract in French

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Twitter Follow Nicolas Peschanski at @DocNikko

  • Contributors NP and LG contributed equally to the article redaction. NP, LG, ED, LM, ER, SD, FR, YK and PR contribute equally to patient selection, data collection and analysis. GL validated the biological results and collected biochemical data from the biochemistry department. NP and YF reviewed all the data and statistical analysis. NP and PR directed the study in the two emergency departments, respectively.

  • Competing interests NP received a personal fee from Radiometer SAS as a conference moderator during a symposium set in June 2014.

  • Ethics approval Ethics approval was obtained from Paris-06 Research and Ethics Committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Data for this article are available on request from the lead author. There are no additional unpublished data from the study.