Article Text

Download PDFPDF

End tidal carbon dioxide is as reliable as transcutaneous monitoring in ventilated postsurgical neonates
  1. David Gerald Tingay1,2,3,
  2. Kwok Sean Mun1,
  3. Elizabeth Jean Perkins1,2
  1. 1Department of Neonatology, Royal Children's Hospital, Melbourne, Victoria, Australia
  2. 2Neonatal Research, Murdoch Children's Research Institute, Melbourne, Victoria, Australia
  3. 3Department of Paediatrics, University of Melbourne, Melbourne, Australia
  1. Correspondence to Dr David Gerald Tingay, Department of Neonatology, Royal Children's Hospital, Flemington Road, Parkville, Melbourne, Victoria 3052, Australia; david.tingay{at}


Objectives To compare the agreement, precision and repeatability of end tidal carbon dioxide (Embedded Image) and transcutaneous carbon dioxide (Embedded Image) with partial pressure of arterial CO2 (Embedded Image) in postoperative neonates.

Patients Fifty mechanically ventilated neonates without lung disease, and with no contraindications for either Embedded Image or Embedded Image monitoring.

Interventions Paired Embedded Image and Embedded Image values were recorded with three consecutive Embedded Image measurements within the first 48 h of surgery.

Main outcome measures Embedded Image, Embedded Image and Embedded Image triplets were compared using Bland-Altman plots.

Results One hundred thirty-two triplet measures of CO2 were recorded with mean Embedded Image 43.5 (7.3) mm Hg, Embedded Image38.8 (6.4) mm Hg and Embedded Image 43.8 (8.8) mm Hg (p<0.0001 for Embedded Image against Embedded Image; paired t test). The Embedded ImageEmbedded Image bias±2SD was 4.1±9.0 mm Hg and −0.8±13.0 mm Hg for Embedded ImageEmbedded Image. 56.1% of Embedded Image, and 60.6% of Embedded Image values were within ±5 mm Hg of paired Embedded Image.

Conclusions In postoperative neonates, Embedded Image and Embedded Image demonstrated a clinically acceptable agreement with Embedded Image.

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.

What is already known on this topic

  • Transcutaneous carbon dioxide (Embedded Image) is widely used to monitor partial pressure of arterial CO2 (Embedded Image) trends in the neonatal intensive care unit (NICU) but rarely during or after surgical procedures.

  • End tidal carbon dioxide (Embedded Image) monitoring is an accepted and reliable method of indicating Embedded Image in the anaesthetised patient.

  • Embedded Image is rarely used in the NICU due to variable accuracy in preterm neonates and neonates with severe respiratory failure.

What this study adds

  • Transcutaneous carbon dioxide (Embedded Image) demonstrated good agreement with partial pressure of arterial CO2 (Embedded Image) in postsurgical neonates without lung disease, but lacked precision over time.

  • End tidal carbon dioxide (Embedded Image) underestimated Embedded Image by about 4 mm Hg but demonstrated better precision than Embedded Image.

  • This study suggests that both Embedded Image and Embedded Image are feasible methods of tracking Embedded Image trends over time in the postsurgical neonate.


Non-invasive monitoring carbon dioxide (CO2) is frequently advocated in the care of the ventilated neonate.1 Transcutaneous CO2 (Embedded Image) monitoring and end tidal CO2 (Embedded Image) are the two most commonly used techniques. Although Embedded Image monitoring is a standard of care during anaesthesia,2 the use in infants is limited due to conflicting results regarding accuracy.3–7 In contrast, Embedded Image is known to provide a good agreement with partial pressure of arterial CO2 (Embedded Image)4 and accurately trend over time.4 Consequently, Embedded Image monitoring is used more in the neonatal intensive care unit (NICU) environment than Embedded Image, although neither has gained universal acceptance, mainly due to technical limitations.

Previous studies of side-stream Embedded Image identified a clinically unacceptable underestimation of Embedded Image.4 ,6 ,7 These studies predominantly involved preterm neonates with lung disease. Embedded Image sensors cannot identify an alveolar CO2 plateau in the large ventilation-perfusion mismatching and fast rate, small tidal volume states characteristic of this population.1 ,2 Neonates are also ventilated after surgery, often without lung disease, and also need strict control of the postoperative course to minimise long-term morbidity. No study has specifically investigated continuous CO2 monitoring in the surgical neonate despite the suitability of this population to non-invasive techniques.

The aim of this study was to compare the agreement, precision and repeatability of sidestream Embedded Image and Embedded Image with Embedded Image in mechanically ventilated postoperative neonates without lung disease.


This study was performed at the Royal Children's Hospital (Melbourne, Australia), a regional surgical NICU, and approved by our research ethics committee. Mechanically ventilated neonates, without primary respiratory failure, <12 h postoperative were studied if they had an indwelling arterial line. Neonates with clinical states known to limit the accuracy of either device were not studied, including congenital diaphragmatic hernia and cyanotic heart disease, fragile skin, significant shock, hypotension, metabolic acidosis, inotropes likely to significantly impair skin perfusion or endotracheal tube leak >20%. Mechanical ventilation was applied using either synchronised intermittent mandatory ventilation or synchronised intermittent positive pressure ventilation with or without volume-targeted modes at the discretion of the clinical team.

On completion of surgery, a microstream FilterLine neonatal sidestream Embedded Image system (Oridion Medical Inc., Needham, Massachusetts, USA; deadspace <0.5 ml, sampling rate 50 ml/min) was incorporated into the ventilator circuit distal to the flow sensor. A Embedded Image probe (TINA, Radiometer Medical, Brønshøj, Denmark) was secured to the abdomen or chest, preferably the right upper chest. The sensor was allowed at least 20-min to achieve thermostability at 43°C before arterial sampling, and repositioned at least 4 hourly. This system does not allow manual calibration to Embedded Image. The capnographic waveform and numerical values for Embedded Image and Embedded Image were displayed with the arterial pressure waveform in real-time (MP70 monitor, Philips Medical, Boeblingen, Germany). Clinicians were not blinded to either measure and were allowed to make ventilation changes based on the displayed Embedded Image values. Arterial blood gas samples were taken when clinically indicated. The results were recorded and the alveolar to arterial oxygen tension (A/a) ratio (Partial pressure of alveolar oxygen Embedded Image; where 0.8=respiratory quotient) calculated. Severe lung disease was defined as an A/a ratio<0.3.

At the time of blood gas analysis, the highest Embedded Image and Embedded Image during the 10 consecutive inflations immediately before and after arterial sampling were recorded. Only Embedded Image values with a distinct alveolar plateau on the capnographic waveform were included. This was repeated for three consecutive arterial samples unless the arterial line was removed, extubation or ineligibility criteria occurred.

Statistical analysis

The Bland-Altman technique8 was used to determine Embedded ImageEmbedded Image and Embedded ImageEmbedded Image agreement. Limits of agreement (precision) defined as two SDs of the bias. A bias of ±5 (10) mm Hg was deemed clinically acceptable.4 ,5 Subgroup analysis to determine the influence of tidal volume, Embedded Image site, time since Embedded Image re-position and order of arterial sampling was made.


Fifty neonates were studied and summarised in table 1. A total of 132 arterial blood gases (56% postductal) were performed, 45 neonates had two or more measurements with 37 neonates having three. The sensor site or duration since re-position did not influence the Embedded Image results.

Table 1

Subject characteristics

The respective mean (SD) Embedded Image, Embedded Image and Embedded Image values were 43.6 (7.0) mm Hg, 39.4 (6.3) mm Hg and 44.3 (8.8) mm Hg, with Embedded Image being lower than Embedded Image and Embedded Image (both p<0.0001, paired t test). Overall, Embedded Imageunderestimated Embedded Image, with a Embedded ImageEmbedded Image bias (2SD) of 4.1 (9.0) mm Hg (figure 1A). Embedded Imageapproximated Embedded Image but with wider limits of agreement: bias (2SD) −0.8 (13) mm Hg (figure 1B). These biases were independent ofEmbedded Image. 56.1% of Embedded Image values, and 60.6% of Embedded Image were within ±5 mm Hg of the paired Embedded Image (p=0.533, Fisher's exact test). Only 27.3% and 35.6% of Embedded Image and Embedded Image values were within ±2 mm Hg of Embedded Image (p=0.420).

Figure 1

Bland-Altman plots of the difference between partial pressure of arterial carbon dioxide (Embedded Image) and end tidal carbon dioxide (Embedded Image) (A) and transcutaneous carbon dioxide (Embedded Image) (B) against average CO2. Values obtained at the first arterial blood gas are shown with closed circle, second with diamonds and third with open circles. The solid line represents the overall bias with ±2 SD being shown in the dashed lines. The Embedded ImageEmbedded Image bias was 4.1 mm Hg and Embedded ImageEmbedded Image bias −0.8 mm Hg.

Subgroup analysis found that tidal volumes ≥10 ml, or ≥4.5 ml/kg (table 2), and sampling order resulted in better agreement between both Embedded Image and Embedded Image with Embedded Image (table 3).

Table 2

Relationship between CO2 measurements and expiratory tidal volume (VT).

Table 3

Bias (limits of agreement) of CO2 measurements at each arterial blood gas sample


The benefit of continuous CO2 monitoring in ventilated neonates is well established,1 but not universally applied. This study suggests that both Embedded Image and sidestream Embedded Image are reliable methods to describe Embedded Image trends in the postoperative period. To our knowledge, this is the largest comparative study of non-invasive CO2 monitoring in ventilated neonates, and the first to selectively identify a population in which neither Embedded Image nor Embedded Image has significant technical disadvantages.

It is not surprising that there was less discrepancy between Embedded Image and Embedded Image than Embedded Image,4 ,6 Embedded Image is known to be a reliable proxy of Embedded Image in neonates.4 ,7 ,9 The precision of Embedded Image, however, was variable, as is evident by the wide limits of agreement. The explanation for this is unclear but likely related to the Embedded Image system rather than sensor placement1 application temperature10 and skin perfusion.

Embedded Imageunderestimated Embedded Image compared with Embedded Image. Previous studies in neonates found Embedded Image underestimated Embedded Image by 6.8–11.2 mm Hg.4–7 The better agreement found in our study likely reflects study populations as previous studies included infants with lung disease. Embedded Image is known to be more accurate in infants with an A/a ratio >0.3.3 We intentionally choose to limit our investigation, and interpretation, to surgical neonates without lung disease. We found that a sidestream Embedded Image system, designed for small volume states, showed better precision between the three sample epochs thanEmbedded Image. The intrasubject repeatability of the Embedded ImageEmbedded Image bias over these samples suggests that monitoring CO2 trends would be feasible using Embedded Image in this population.

The improved agreement between Embedded Image and Embedded Image at higher tidal volumes illustrates the importance of achieving true capnographic alveolar plateau.2 The relatively large circuit deadspace, high rates and low tidal volumes characteristic of neonatal ventilation may not result in the sampled gas representing true Embedded Image.1 ,2 In neonates requiring low tidal volumes Embedded Image maybe a better alternative. By contrast, the surgical neonate provides some unique challenges that may limit Embedded Image monitoring, including wound dressings, drapes, drains and other monitoring devices limiting access to well perfused Embedded Image sites. As neither Embedded Image nor Embedded Image were clearly superior in this study, our results suggest that the consideration of each system's limitations, and the clinical environment, will be important in determining the most appropriate method of CO2 monitoring for an individual neonate.


To minimise unnecessary blood loss the timing of arterial sampling was not standardised. Ventilation strategy were also at the discretion of the treating clinician. Not every neonate contributed three arterial gases. This was a pragmatic choice to represent clinical practice but potentially introduces bias, and may explain the Embedded Image differences over time. We contend that the fact that Embedded Image agreement did not differ between samples suggests that the Embedded Image differences were related to other factors. Overall, the postsurgical population is a significant contributor to NICU occupancy, but specific diagnoses are rare and this resulted in a heterogeneous population, limiting subgroup analysis.


In postsurgical neonates without lung disease, Embedded Image underestimated Embedded Image more than Embedded Image but provided greater precision over repeated arterial blood gases, however it was less accurate at smaller tidal volumes. Both techniques are feasible methods of non-invasively monitoring CO2 trends in the postsurgical neonate. The clinician should be aware of the specific advantages and disadvantages of each.


The authors wish to thank Dr Neil Patel for assistance in preparing this manuscript.



  • Contributors All authors made substantial contributions to conception and design, data acquisition, analysis and interpretation. All authors were involved in the drafting of the submitted manuscript and approve of the manuscript in current form. DGT authored the first draft of the manuscript.

  • Financial support DGT is supported by a National Health and Medical Research Council Clinical Research Fellowship (Grant ID 491286) and the Victorian Government Operational Infrastructure Support Programme.

  • Competing interests None.

  • Ethics approval Human Research Ethics Committee of the Royal Children's Hospital, Melbourne, Victoria, Australia.

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