The gold standard for measuring carbon dioxide levels in the blood is currently via arterial blood gas sampling. This can be a difficult procedure and is often uncomfortable for the patient, particularly when repeated samples are required to monitor treatment. The insertion of an arterial line can facilitate multiple blood gas samples and is feasible within the emergency department (ED), but has its own associated complications such as thrombosis, aneurysm formation and distal ischaemia. Transcutaneous carbon dioxide monitors using a sensor taped to the chest wall were first developed in the neonatal ICU setting1 to reduce blood sampling, but technical problems have limited their utility in adult patients. Recent developments in technology have produced more reliable and practical monitors with less complex calibration requirements. These offer a potential non-invasive means of determining arterial carbon dioxide tension (Paco₂) and have the advantage over all methods of arterial access of producing a continuous readout. Several studies have looked at the effectiveness of such devices in relatively stable patients in sleep laboratories, outpatients and ICUs,2^–7 but the potential benefits to more acute subjects in the ED have not yet been examined. This is the first study to assess the accuracy and feasibility of one such monitor, the TOSCA 500, (Linde Medical Sensors AG, Basel, Switzerland) (fig 1) in patients in the ED.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

TOSCA 500 monitor.

METHODS

Study design

This was a prospective observational cohort study consisting of a convenience sample of ED patients recruited over a 2-month period.

Study setting and population

The study was conducted in an urban ED at the Royal Liverpool University Hospital. The ED is predominantly an adult unit with an annual new patient attendance of 86 000. All patients attending the ED who, in the opinion of their treating physician, warranted arterial blood gas sampling as part of the diagnostic investigation were eligible for inclusion. The only exclusion criteria were an inability to give informed consent owing to reduced consciousness or mental capacity.

Study protocol, measurements and data collection

Eligible patients were identified by their treating physician and then enrolled by one of the senior ED doctors who had undergone hands-on training from the manufacturer in the use of the monitor. After informed consent, a standardised data collection sheet recorded the patient’s age, sex, temperature, pulse rate, respiratory rate, blood pressure, oxygen saturation level and the indication for arterial blood gas sampling. Any abnormality of the ears was also recorded.

A single-use ear clip was attached to one of the patients’ ears and the sensor from the TOSCA monitor, with a drop of contact gel applied, was clipped in place. The monitor’s “Quickstart” mode was selected; this aims to arterialise the blood in the pinna by heating it to a temperature of 42°C over 5 min. The monitor was allowed to stabilise over 5 min and an arterial blood gas sample was then obtained from the radial artery and analysed immediately. The TcPco₂ reading on the monitor at the moment of withdrawal of the needle was recorded.

Primary outcome

The primary outcome variable was concordance between the simultaneous arterial and transcutaneous carbon dioxide measurements.

Data analysis

Descriptive data are presented as means with ranges, medians with interquartile ranges (IQR) and as frequencies. Mean differences are presented as point estimates with 95% confidence intervals. The Pearson correlation coefficient was used to demonstrate the presence or absence of a relationship between the paired arterial and transcutaneous carbon dioxide measurements. A Bland-Altman8 scatterplot was constructed by plotting the differences between the two measurements against the mean of these measurements for a given individual. This allowed for the calculation of bias (mean difference) and limits of agreement (±2SD). Relationships between measurement differences and patient characteristics were investigated by scatterplot and by regression analysis.

Sample size calculation was based on the anticipated variation in the differences between the measurements and the required precision. Using Heuss et al9 for an estimate of the variation between the differences, a sample size of 50 patients gives a precision of ±0.19 kPa in the limits of agreement.

Statistical analysis was conducted using SPSS Version 13 (SPSS Inc, Chicago, Illinois, USA).

RESULTS

Fifty-one patients were enrolled in the study with paired measurements for arterial and transcutaneous carbon dioxide tension recorded for each. Technical problems with the monitor led to the exclusion of measurements for two patients. Patient characteristics are shown in table 1. The median age was 66 years (IQR 53–75). An exacerbation of chronic obstructive pulmonary disease (COPD) was the presenting problem in 55% of patients.

The mean difference (bias) between measurements was 0.02 kPa. Pearson’s correlation coefficient for the measurements was 0.94 (p<0.001). This shows that a relationship exists between the measurements but does not necessarily mean that they agree. The Bland-Altman matrix is given in fig 2. The 95% limits of agreement of the arterial and transcutaneous carbon dioxide measurements were ±0.9 kPa. No relationship could be identified between the difference in measurements and other patient variables (age, sex, temperature, pulse rate, respiratory rate, blood pressure, oxygen saturation level or medical condition).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Bland-Altman plot of absolute differences between arterial and transcutaneous measurements of carbon dioxide tension versus means of paired measurements. The limits of agreement are demonstrated by the unbroken lines around the mean difference of 0.02 kPa.

DISCUSSION

Transcutaneous measurement of carbon dioxide tension makes use of the fact that carbon dioxide can diffuse through body tissue and skin and can be detected by a sensor at the skin surface. Warming of the sensor induces a local hyperaemia which increases the supply of arterial blood to the dermal capillary bed. Carbon dioxide passing through the gas-permeable membrane of the ear clip induces changes in the pH of the electrolyte solution which are detected by the measurement of potential between a miniature glass pH electrode and a reference electrode by the Stow-Severinghaus sensor. While capnography has been proposed as a potential non-invasive method of indirectly measuring arterial carbon dioxide tension, its accuracy is limited in sick patients with “shunting” or low ventilation–perfusion ratios. This does not apply to transcutaneous monitoring which also has the advantage of being possible in patients breathing room air. It has the potential to provide useful information in a range of patients in the ED with both respiratory and metabolic conditions as well as increasing the safety of monitoring in procedural sedation.

The mean difference found between the measurements in our study was 0.02 kPa (0.15 mm Hg) with limits of agreement of ±0.9 kPa (6.84 mm Hg). The limits of agreement are larger than in some earlier small studies. Eberhard et al10 found a mean (SD) value of 1.22 (3.69) mm Hg and Kagawa et al11 reported a value of 2.3 (2.5) mm Hg. It is closer, however, to the result from the study by Bendjelid et al12 of 1.2 (6.0) mm Hg in critically ill ITU patients and falls within the range of 7.5 mm Hg which they described as being “clinically acceptable”. A recent study by Bolliger et al13 on the accuracy of the Tosca monitor in intubated patients in theatre and the ITU found limits of agreement of ±1.6 kPa, which was outside their predetermined range of ⩽1 kPa. They concluded that it was not of benefit in these clinical settings, but could not identify any obvious cause of the inaccuracy such as hypothermia, vasopressor use or haemodynamic instability. The manufacturer’s literature suggests, however, that excessive illumination such as surgical lamps may be one possible source of inaccurate measurement. The results of our study do fall within their “clinically useful” range, and we feel this degree of accuracy is reasonable in patients in the ED. We specified a relatively short time period from application of the monitor to blood sampling (5 min). This may have adversely affected the efficacy of the monitor, but this figure was chosen to reflect the need for a monitor to provide rapid information in patients in the ED.

Low readings may be due to insecure application of the ear clip allowing exposure to the air;14 however, this study was conducted by senior ED physicians who had undergone training in the use of the monitor. Excessively high values may be due to an “overshoot” phenomenon described by Kagawa et al,11 although the Quickstart process (initial heating to 45° for rapid arterialisation) is designed to minimise this.

The results of two patients were excluded from the analysis. In one case the membrane of the ear clip was damaged during removal of its protective backing paper prior to application. In the other, an unknown sensor application problem was suspected as the monitor displayed a “low perfusion index” message despite the patient being well with normal vital signs. The recruitment of 49 patients instead of the intended minimum of 50 slightly reduces the precision of the estimate. However, this reduction in precision does not invalidate our conclusions.

For ethical reasons it was not possible to recruit patients who could not give informed consent. Inevitably, this led to the exclusion of some patients with higher degrees of hypercarbia. Application of the findings of our study to such patients should be done with caution.

Anecdotally, the monitor was popular with patients, particularly those with COPD who had experience of numerous previous arterial punctures. There were no complaints of discomfort from the sensor and no local complications. The manufacturers advise that they can be safely left in place for up to 8 h before the location needs to be changed.

While the monitor can never replace arterial blood gas sampling completely, it appears to be accurate enough to continuously monitor trends and alert staff to potentially dangerous changes in carbon dioxide levels, guiding the appropriate use of arterial puncture. The monitor costs in the region of £7000 and, while the cost per patient depends on the frequency of use, it is estimated at around £10.

CONCLUSION

The TOSCA monitor was found to be an accurate and practical means of non-invasive monitoring of Paco₂ in patients in the ED. It may be of particular benefit to those patients requiring continuous monitoring to avoid complications associated with high Paco₂, but further studies are required to investigate the clinical benefits of the monitor in such patients.