The accuracy of non-invasive carbon dioxide monitoring: A clinical evaluation of two transcutaneous systems

Transcription

1 doi: /j x APPARATUS The accuracy of non-invasive carbon dioxide monitoring: A clinical evaluation of two transcutaneous systems D. Bolliger, 1 L. A. Steiner, 2 J. Kasper, 1 O. A. Aziz, 3 M. Filipovic 2 and M. D. Seeberger 2 1 Research fellow, 2 Consultant Anaesthetist, 3 Medical Student, Department of Anaesthesia and Intensive Care Unit, University of Basel Hospital, CH-4031 Basel, Switzerland Summary We determined the accuracy of two transcutaneous carbon dioxide monitoring systems (SenTec Digital Monitor with V-Sign Sensor and TOSCA 500 with TOSCA Sensor 92) for the measurement of single values and trends in the arterial partial pressure of carbon dioxide in 122 adult patients during major surgery and in 50 adult patients in the intensive care unit. One or several paired measurements were performed in each patient. The first measurement was used to determine the accuracy of a single value of transcutaneous carbon dioxide; the difference between the first and the last measurements was used to analyse the accuracy and to track trends. We defined a 95% limit of agreement of 1 kpa as being clinically useful. There was insufficient agreement between transcutaneous carbon dioxide partial pressure values derived from the two systems and arterial carbon dioxide values for both single values and trends as defined by our suggested limit of agreement. We conclude that these systems cannot replace conventional blood gas analysis in the clinical setting studied.... Correspondence to: Dr D. Bolliger dabolliger@uhbs.ch Accepted: 2 September 2006 The arterial partial pressure of carbon dioxide (P a CO 2 ) is an important indicator of respiratory function and ventilation that is monitored in patients in the operating theatre, in the intensive care unit (ICU) and during sedation [1]. Arterial blood gas analysis is the gold standard for P a CO 2 measurement but this method has major disadvantages because each determination of P a CO 2 requires withdrawal of arterial blood and laboratory analysis. Other disadvantages are that assessment of P a CO 2 is intermittent, relatively expensive and there is a time delay in obtaining results. In addition, the need for invasive monitoring is not without complications such as infection, haemorrhage and vessel occlusion. There are alternative methods for P a CO 2 measurement but they are currently too inaccurate to replace blood gas analysis. One such method is the monitoring of end-tidal carbon dioxide values (F É CO 2 ). This can be performed non-invasively and continuously in intubated patients but typically underestimates P a CO 2 ; its dependency on ventilation-perfusion ratio and cardiac output further decreases its value as an indicator of P a CO 2 [2]. Recently, new miniaturised systems for continuous and non-invasive carbon dioxide monitoring have been introduced into clinical practice. Based on Stow- Severinghaus electrodes, these systems measure the transcutaneous partial pressure of carbon dioxide (P tc CO 2 ) after arterialisation of the skin by local warming. The sensors are attached to the ear lobe and provide continuous, non-invasive on-line information about P tc CO 2 in addition to transcutaneous oxygen saturation. Using this method, good agreement has been reported between P tc CO 2 and P a CO 2 [3 5]; however, one study only evaluated volunteers [4], and two studies were not independent of the manufacturers [3, 4]. We therefore conducted a prospective study to test the accuracy and clinical usefulness of two of these transcutaneous systems in patients undergoing major surgery and in intubated patients in the ICU. We defined the required precision of the transcutaneous 394 Journal compilation Ó 2007 The Association of Anaesthetists of Great Britain and Ireland

2 D. Bolliger et al. Æ Transcutaneous carbon dioxide monitoring systems for clinical usefulness as being 95% limit of agreement 1 kpa [6]. Methods The study protocol was approved by the Local Research Ethics Committee, and all subjects gave written, informed consent. The study consisted of two parts, one performed in the operating theatre and the other in the ICU. Patients were eligible for the operating theatre part of the study if they were scheduled for major surgery that required placement of an indwelling arterial catheter and repeated arterial blood gas analysis. Patients were eligible for the ICU part of the study if their lungs were mechanically ventilated because of a planned, delayed tracheal extubation after major surgery or for any nonsurgical reason. Four of these latter patients were included in the study after written, informed consent had been obtained from their relatives, they also gave written, informed consent themselves after tracheal extubation. The ear lobes of each patient were inspected before and after the study. Pre-existing lesions on the ear lobe were an exclusion criterion. All measurements in the operating theatre and the ICU were performed during stable cardiovascular conditions and during nasopharyngeal or tympanic normothermia, defined as a body temperature of C. In all patients in the operating theatre and in 32 (64%) patients in the ICU, measurements were performed during stable volume-controlled ventilation. In 18 (36%) patients in the ICU, measurements were performed during weaning from volume-controlled ventilation to assisted spontaneous breathing. Paired measurements could be obtained in all of these latter 18 patients with the SenTec monitor (SenTec Digital Monitor System with V-Sign TM Sensor, Sentec AG, Therwil, Switzerland) and in 15 patients with the TOSCA monitor (TOSCA 500 with TOSCA Sensor 92, Linde Medical Sensors, Basel, Switzerland). Before starting the measurements, the patients haemodynamic values, body mass index, laboratory values, medical history and medications were noted. The SenTec and TOSCA measurement systems have been described previously [3, 7, 8]. The sensors are automatically calibrated using one-point dry gas calibration with 8% carbon dioxide (SenTec) and 7% carbon dioxide (TOSCA). To induce local vasodilation and to enhance skin permeability to carbon dioxide to improve gas diffusion at the measurement sites, the sensor in the SenTec system is heated to 42 C. The TOSCA sensor is heated to 44 C for the first 20 min to prevent initial P tc CO 2 overshoot and heating then continued at 42 C [9]. The sensors were attached to the inner aspect of the ear lobe with the monitor-specific clip. Before sensor placement, the skin was cleaned with alcohol and then dried, and one drop of contact gel was applied to the centre of the sensor. If a monitor indicated bad signal quality, the sensor was repositioned and the system was recalibrated. End-tidal carbon dioxide and P tc CO 2 were measured at the same time as arterial blood gas sampling in both the operating theatre (ADU, Datex-Ohmeda, Bromma, Sweden) and the ICU (Evita 4, Dräger Medical, Lübeck, Germany). End-tidal carbon dioxide values during singlelung ventilation were excluded from analysis, as were values obtained during or after cardiopulmonary bypass in cardiac surgical patients. Arterial blood gas analyses were performed either by an ABL 700 series (Radiometer, Copenhagen, Denmark) or a GEM Premier 3000 (Instrumentation Laboratory, Barcelona, Spain) system. Arterial carbon dioxide values were compared with simultaneously obtained P tc CO 2 values from the SenTec and TOSCA systems, as well as with F É CO 2 values. Haemodynamic and respiratory parameters and the use of vasopressors at the time of each simultaneous carbon dioxide analysis were noted. For the analysis of the accuracy of the measurement of single values in the operating theatre or the ICU, the first measurement with a good signal quality in each patient was included. The proportion of P tc CO 2 values that were within the required precision of ± 1 kpa was determined. The trend, defined as the difference in P a CO 2 over time, was compared with the corresponding trends in P tc CO 2 and F É CO 2 in patients having multiple measurements. To calculate the trend, the first and the last carbon dioxide measurements from the SenTec and TOSCA systems with good quality contact as indicated by the systems were used. Receiver operator characteristic (ROC) curves for P tc CO 2 were constructed to determine the accuracy of both systems to detect hypercapnia, defined as P a CO 2 > 5.5 kpa. These ROC curves were compared with the analogous ROC curve for F É CO 2. The optimal cut-off points were determined. The proportion of hypercapnic P a CO 2 values detected by the cut-off points for each of the two systems was determined. The authors performed all measurements and analyses. The bias and 95% limit of agreement (defined as mean difference: 1.96 SD of the difference) were calculated as described by Bland and Altman [10]. The two P tc CO 2 measurement systems were compared by using the Chi-squared test. Other results are expressed as mean (SD) or as median (IQR [range]). Data were analysed using SPSS 13.0 (SPSS Inc., Chicago, IL). Results The part of the study performed in the operating theatre included 122 patients. Of these patients, 93 (76%) underwent cardiac surgery, 28 (23%) lung or mediastinal Journal compilation Ó 2007 The Association of Anaesthetists of Great Britain and Ireland 395

3 D. Bolliger et al. Æ Transcutaneous carbon dioxide monitoring Anaesthesia, 2007, 62, pages Table 1 Patients characteristics. Values are mean (SD) or number (%). Operating theatre patients, n ¼ 122 Intensive care patients, n ¼ 50 Age; years 63.8 (11.8) 66.7 (12.5) Weight; kg 76.1 (16.3) 78.2 (15.8) Height; cm (9.3) (8.6) Body mass index; kg.m ) (4.3) 26.8 (3.9) Sex; M : F 89 : : 10 ASA physical status (8%) 0 ASA physical status (89%) 38 (76%) ASA physical status 4 4 (3%) 12 (24%) (kpa) SenTec surgery and one (1%) neurosurgery. The ICU part of the study included 50 patients, of which 46 (92%) were admitted for lung ventilation after cardiac surgery, two (4%) for severe sepsis and two (4%) for neurological reasons (one ischaemic stroke and one subarachnoid haemorrhage). A total of 42 cardiac surgical patients were studied both in the operating theatre and in the ICU. Table 1 shows the patients characteristics. To maintain haemodynamic stability in the operating theatre, 60 (49%) patients required pharmacological support within half an hour of arterial blood gas sampling: ephedrine was given to 15 patients at a median (IQR [range]) dose of 5 (2.5 5 [2.5 20]) mg and phenylephrine to 52 patients at a median (IQR [range]) dose of 0.2 ( [5 0.75]) mg; 12 of these patients were given both ephedrine and phenylephrine. Five of these patients were also given noradrenaline in a dose of 1 lg.kg )1.min )1, and one patient was given both noradrenaline and adrenaline in doses of 4 lg.kg )1.min )1. To maintain cardiovascular stability in the ICU, 26 (52%) patients were given continuous pharmacological support during the study period: noradrenaline was given to 23 (46%) patients at continuously adjusted, varying doses of up to lg.kg )1.min )1, adrenaline to 17 (34%) patients at doses of up to 0.15 lg.kg )1.min )1 ; 14 of these patients received both adrenaline and noradrenaline. The monitoring equipment used in the investigation was well tolerated by all patients, and no adverse events such as skin lesions or thermal injury to the ear lobes were observed. Some loss of data resulted from inadvertent detachment of sensors from the ear lobes, insufficient signal quality of one or both systems at the time of arterial blood sampling or failure and unavailability of a monitoring system. Figures 1 and 2 show the agreement between the P tc CO 2 values indicated by the SenTec and the TOSCA systems and P a CO 2 values measured in the operating theatre. A total of 101 and 112 paired measurements with arterial blood gas analysis were obtained for the SenTec and TOSCA systems, respectively. Bias ± 95% limit of agreement (kpa) SenTec)/2 1 Figure 1 Bland-Altman plot for single values (first measurement) from the SenTec Digital Monitor with V-Sign Sensor in the operating theatre. (kpa) TOSCA (kpa) TOSCA)/2 1 Figure 2 Bland-Altman plot for single values (first measurement) from the TOSCA Sensor 92 in the operating theatre. between P a CO 2 and P tc CO 2 was ) 0.57 ± 2 kpa for the SenTec system and ) 0.29 ± 1.43 kpa for the TOSCA system. These results were similar in the equally large groups of patients with and without administration of vasopressors (data not given in detail). In total, 76% of the measurements from the SenTec system and 85% of the measurements from the TOSCA system were within the required precision of ± 1 kpa of the corresponding P a CO 2 value (v 2 ¼ 0.10). Figures 3 and 4 show the agreement between the SenTec and the TOSCA systems and P a CO 2 values in the ICU. A total of 49 paired measurements were obtained for each system. bias ± 95% limit of agreement between P a CO 2 and P tc CO 2 was ) 0.35 ± 1.53 kpa for the 396 Journal compilation Ó 2007 The Association of Anaesthetists of Great Britain and Ireland

4 D. Bolliger et al. Æ Transcutaneous carbon dioxide monitoring (kpa) SenTec trend (kpa) trend PtcCO 2 (kpa) SenTec Operating theatre ICU (kpa) SenTec)/2 Figure 3 Bland-Altman plot for single values (first measurement) from the SenTec Digital Monitor with V-Sign Sensor in the intensive care unit. (trend (kpa) + trend (kpa) SenTec)/2 Figure 5 Bland-Altman plot for trend (difference between the last minus the first measurement) from the SenTec Digital Monitor with V-Sign Sensor. Data from the operating theatre and the intensive care unit were pooled. (kpa) TOSCA trend (kpa) trend PtcCO 2 (kpa) TOSCA Operating theatre ICU (kpa) TOSCA)/2 Figure 4 Bland-Altman plot for single values (first measurement) from the TOSCA 500 with TOSCA Sensor 92 in the intensive care unit. SenTec system and )0.29 ± 1.25 kpa for the TOSCA system. The results in the ICU were also similar in the equally large groups of patients with and without administration of vasopressors (data not given in detail). In total, 82% of the measurements from the SenTec system and 86% of the measurements from the TOSCA system were within the required precision of ± 1 kpa of the corresponding P a CO 2 value (v 2 ¼ 0.79). Figures 5 and 6 show the agreement in the trend between P a CO 2 and the two transcutaneous systems. The data from the operating theatre and the ICU parts of the study were pooled for this analysis. For the SenTec system, 24 paired measurements were performed in the (trend (kpa) + trend PtcCO 2 (kpa) TOSCA)/2 Figure 6 Bland-Altman plot for trend (difference between the last minus the first measurement) from the TOSCA 500 with TOSCA Sensor 92. Data from the operating theatre and the intensive care unit were pooled. operating theatre and 47 in the ICU. Bias ± 95% limit of agreement for the trends of P a CO 2 and the SenTec P tc CO 2 was 0.44 ± 1.82 kpa. For the TOSCA system, 26 paired measurements were performed in the operating theatre and 46 in the ICU. Bias ± 95% limit of agreement for the trends of P a CO 2 and the TOSCA P tc CO 2 was ) 0.41 ± 1.39 kpa. Bias ± 95% limit of agreement was similar when patients whose lungs were ventilated in a volumecontrolled mode at both measurement points and patients who had been switched to assisted spontaneous breathing before the second measurement point were analysed separately. In the latter patient group, bias ± 95% limit Journal compilation Ó 2007 The Association of Anaesthetists of Great Britain and Ireland 397

5 D. Bolliger et al. Æ Transcutaneous carbon dioxide monitoring Anaesthesia, 2007, 62, pages of agreement for the trends of P a CO 2 and P tc CO 2 was 0.58 ± 1.23 kpa for the SenTec system (n ¼ 18) and ) 0.26 ± 1.49 for the TOSCA system (n ¼ 15). Bias ± 95% limit of agreement for single values of F É CO 2 was 0.78 ± 0.85 kpa (n ¼ 92) and 0.64 ± 1.25 kpa (n ¼ 50) in the operating theatre and in the ICU, respectively. Based on 11 paired values from the operating theatre and 49 values from the ICU, bias ± 95% limit of agreement for the trends of P a CO 2 and F É CO 2 was )0.26 ± 1.17 kpa. Receiver operator characteristic curves were constructed to study the ability of the transcutaneous systems to detect hypercapnia (P a CO 2 > 5.5 kpa). The data from the operating theatre and the ICU parts of the study were pooled for this analysis. Hypercapnia was present in 46 (27%) patients at the time of the first P a CO 2 determination. The area under the curve was 0.84 and 0.83 for the SenTec and TOSCA systems, respectively. The optimal cut-off point for the SenTec system was 5.58 kpa with a positive predictive value of 0.53 and a negative predictive value of The optimal cut-off point for the TOSCA system was 5.55 kpa with a positive predictive value of 0.50 and a negative predictive value of With these cut-off points, 93% and 86% of hypercapnic P a CO 2 levels were detected by the SenTec and TOSCA systems, respectively (v 2 ¼ 0.30). The same analysis was performed to analyse the value of F É CO 2 for detecting hypercapnia. The area under the curve was 0.75, the optimal cut-off point was 4.35 kpa, and the positive and negative predictive values were 0.40 and 0.93, respectively. This cut-off point detected 94% of hypercapnic P a CO 2 values. There was no significant difference in detection of hypercapnia compared with the transcutaneous systems. Discussion Our study found that the accuracy of the two new transcutaneous systems for non-invasive monitoring of P a CO 2 was below the level that we had defined as being clinically useful, i.e. 95% limit of agreement of ± 1 kpa [6]. The transcutaneous systems did not provide any additional information in ventilated patients with intubated tracheas because the P tc CO 2 values were no more accurate than the routinely available F É CO 2 values. We conclude that the two transcutaneous systems lack sufficient accuracy for monitoring P a CO 2 in ventilated surgical patients with intubated tracheas or critically ill patients. The gold standard for P a CO 2 monitoring, arterial blood sampling and laboratory analysis, has itself a degree of imprecision that is dependent on the P a CO 2 level. For example, the 95% limit of agreement of the ABL 700 series machine used in this study is ± 0.24 kpa at a physiological P a CO 2 of 5.5 kpa. Further imprecision can result from incorrect handling such as delayed transportation, high temperature or air in the syringe. We tried to eliminate these additional sources of imprecision by using only a small number of people to draw and immediately process the arterial blood samples according to a standardised procedure. Nevertheless, we postulated a 95% limit of agreement of ± 0.5 kpa for the arterial blood gas analysis and defined twice this range as the maximally acceptable 95% limit of agreement between P a CO 2 and P tc CO 2, i.e. ± 1.0 kpa. The findings of our study are in conflict with those of several published studies [3, 4, 11]. However, our results are in agreement with those of two recent studies performed in critically ill patients that reported 95% limits of agreement between P a CO 2 and P tc CO 2 of ± 1.6 kpa with the TOSCA system [6] and ± 1.2 kpa with the SenTec system [12]. However, we do not share the same conclusion as these investigators that these 95% limits of agreement indicate sufficient accuracy of the systems for reliable clinical use [6, 12]. The difference between our study and previous studies that found better agreement between P a CO 2 and P tc CO 2 may have several causes. One study was conducted in healthy volunteers [4], whereas all our patients underwent major surgery or were severely ill. Two previous studies were not conducted independently of the manufacturers [3, 4], whereas in our study design, data collection and manuscript preparation were. One might suspect that the use of vasopressors in our patients was another reason for the poor accuracy of the transcutaneous systems and the conflicting results, as the difference between P tc CO 2 and P a CO 2 is influenced by skin perfusion and skin temperature [13, 14]. However, we found that the use of vasopressors did not significantly affect P tc CO 2 measurements, which is in agreement with several previous studies [6, 11, 12, 15]. Moreover, fewer than half of the study patients in the ICU were given vasopressors, and higher doses of noradrenaline (> 0.15 lg.kg )1.min )1 ) were only necessary in 11% of these patients. Thus, the use of vasopressor agents did not seem to influence our P tc CO 2 measurements significantly. Hypothermia can also be excluded as a reason for differing results in our study because body temperatures were continuously monitored and all patients were normothermic. Finally, a low cardiac output also can be excluded because all patients were haemodynamically stable. The differences in P tc CO 2 measurements between the two transcutaneous systems were small. Overall, TOSCA showed slightly better, although still insufficient, accuracy in the operating theatre and in the ICU. End-tidal carbon dioxide monitoring is routinely performed to estimate P a CO 2 in intubated patients and was thus also analysed in our study. Although F É CO 2 has the disadvantages 398 Journal compilation Ó 2007 The Association of Anaesthetists of Great Britain and Ireland

6 D. Bolliger et al. Æ Transcutaneous carbon dioxide monitoring of usually underestimating P a CO 2 and being strongly dependent on ventilation-perfusion ratio and cardiac output [2], the F É CO 2 trend can be useful during stable conditions [16, 17]. Neither of the two transcutaneous systems was better than F É CO 2 in assessing single values or trends, or for detecting hypercapnia. In conclusion, the P tc CO 2 monitor systems we tested failed to indicate P a CO 2 reliably in ventilated surgical patients with intubated tracheas or in critically ill patients, and the systems were not beneficial in providing additional information when compared with routinely available F É CO 2 values. Therefore, arterial blood gas analysis still needs to be performed whenever precise assessment of P a CO 2 is needed. Acknowledgements SenTec AG, Therwil, Switzerland, provided a SenTec Digital Monitor System with V-Sign TM Sensor as well as all disposable parts for conduct of the study. Linde Medical Sensors AG, Basel, Switzerland, provided a TOSCA 500 system with a TOSCA Sensor 92 as well as all disposable parts. There are no conflicts of interest to declare. The authors thank Kai Monte RNA, for technical assistance and help with data collection in the operating theatre and Joan Etlinger BA, for editorial assistance. References 1 American Society of the Anesthesiologist Task Force. Practice guidelines for sedation and analgesia by nonanesthesiologists. Anesthesiology 2002; 96: Wahba RW, Tessler MJ. Misleading end-tidal CO 2 tensions. Canadian Journal of Anaesthesia 1996; 43: Eberhard P, Gisiger PA, Gardaz JP, Spahn DR. Combining transcutaneous blood gas measurement and pulse oximetry. Anesthesia and Analgesia 2002; 94: S Hayoz J, Rohling R, Tschupp A. World s first combined digital pulse oximetry and carbon dioxide tension ear sensor. Anesthesia and Analgesia 2002; 94 1S: Senn O, Clarenbach CF, Kaplan V, Maggiorini M, Bloch KE. Monitoring carbon dioxide tension and arterial oxygen saturation by a single earlobe sensor in patients with critical illness or sleep apnea. Chest 2005; 128: Bendjelid K, Schutz N, Stotz M, Gerard I, Suter PM, Romand JA. Transcutaneous PCO 2 monitoring in critically ill adults: clinical evaluation of a new sensor. Critical Care Medicine 2005; 33: Heuss LT, Chhajed PN, Schnieper P, Hirt T, Beglinger C. Combined pulse oximetry cutaneous carbon dioxide tension monitoring during colonoscopies: pilot study with a smart ear clip. Digestion 2004; 70: Kocher S, Rohling R, Tschupp A. Performance of a digital PCO 2 SPO 2 ear sensor. Journal of Clinical Monitoring and Computing 2004; 18: Kagawa S, Otani N, Kamide M, Gisiger PA, Eberhard P, Severinghaus JW. Initial transcutaneous PCO 2 overshoot with ear probe at 42 degrees C. Journal of Clinical Monitoring and Computing 2004; 18: Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1: Janssens JP, Howarth-Frey C, Chevrolet JC, Abajo B, Rochat T. Transcutaneous PCO 2 to monitor noninvasive mechanical ventilation in adults: assessment of a new transcutaneous PCO 2 device. Chest 1998; 113: Rodriguez P, Lellouche F, Aboab J, Brun Buisson C, Brochard L. Transcutaneous arterial carbon dioxide pressure monitoring in critically ill adult patients. Intensive Care Medicine 2006; 32: Burki NK, Albert RK. Noninvasive monitoring of arterial blood gases. A report of the ACCP section on respiratory pathophysiology. Chest 1983; 83: Hazinski TA, Severinghaus JW. Transcutaneous analysis of arterial PCO 2. Medical Instrumentation 1982; 16: Palmisano BW, Severinghaus JW. Transcutaneous PCO 2 and PO 2 : a multicenter study of accuracy. Journal of Clinical Monitoring 1990; 6: Hillen T, Sumpelmann R, Strauss JM. [Intraoperative changes in arterial end-tidal CO 2 partial pressure difference in interventions with constant ventilation-perfusion ratio]. Anästhesiologie und Reanimation 1999; 24: Sharma SK, McGuire GP, Cruise CJ. Stability of the arterial to end-tidal carbon dioxide difference during anaesthesia for prolonged neurosurgical procedures. Canadian Journal of Anaesthesia 1995; 42: Journal compilation Ó 2007 The Association of Anaesthetists of Great Britain and Ireland 399