Arterial and end-tidal carbon dioxide difference in pediatric intensive care

Abstract

Background and Aim:

Arterial carbon dioxide tension (PaCO₂) is considered the gold standard for scrupulous monitoring in pediatric intensive care unit (PICU), but it is invasive, laborious, expensive, and intermittent. The study aims to explore when we can use end-tidal carbon dioxide tension (P_ETCO₂) as a reliable, continuous, and noninvasive monitor of arterial CO₂

Materials and Methods:

Concurrent P_ETCO₂, fraction of inspired oxygen, PaCO₂, and arterial oxygen tension values of clinically stable children on mechanical ventilation were recorded. Children with extra-pulmonary ventriculoatrial shunts were excluded. The P_ETCO₂ and PaCO₂ difference and its variability and reproducibility were studied.

Results:

A total of 624 concurrent readings were obtained from 105 children (mean age [SD] 5.53 [5.43] years) requiring invasive bi-level positive airway pressure ventilation in the PICU. All had continuous P_ETCO₂ monitoring and an arterial line for blood gas measurement. The mean (SD) number of concurrent readings obtained from each child, 4-6 h apart was 6.0 (4.05). The P_ETCO₂ values were higher than PaCO₂ in 142 observations (22.7%). The PaCO₂–P_ETCO₂ difference was individual admission specific (ANOVA, P < 0.001). The PaCO₂–P_ETCO₂ difference correlated positively with the alveolar-arterial oxygen tension [P(A-a)O₂] difference (ρ = 0.381 P < 0.0001). There was a fixed bias between the P_ETCO₂ and PaCO₂ measuring methods, difference +0.66 KPa (95% confidence interval: +0.57 to +0.76).

Conclusions:

The PaCO₂–P_ETCO₂ difference was individual specific. It was not affected by the primary disorder leading to the ventilation.

Introduction

The measurement of partial pressure of carbon dioxide in arterial blood (arterial carbon dioxide tension [PaCO₂]) is an essential diagnostic and monitoring tool used in current clinical practice. However, its measurement remains invasive and intermittent.[1] The graphic display of CO₂ concentration (or partial pressure) during the respiratory cycle (capnography) has offered many uses in adult[2,3,4] and pediatric clinical practice.[5,6] The absolute numerical measurement of CO₂ concentration (or partial pressure) during the respiratory cycle has proved less useful. This is because the partial pressure of CO₂ at the end of exhalation (end-tidal carbon dioxide tension [P_ETCO₂]) does not reliably correlate with PaCO₂. Our study was designed to investigate the relationship between P_ETCO₂ and PaCO₂ in children during steady state mechanical ventilation in the pediatric intensive care unit (PICU).

Materials and Methods

All patients receiving treatment in our PICU have their vital signs data automatically recorded and saved in an electronic patient record system (MetaVision, iMD-Soft systems, Schiessstraße 55, 40549, Düsseldorf, Germany). The laboratory investigation results and blood gases are saved in the same record. Mechanically ventilated patients also have their fraction of inspired oxygen (FiO₂) and P_ETCO₂ measured concurrently and automatically recorded. The FiO₂ was measured by an oxygen electrode (calibrated daily) located at the inspiratory limb of the ventilator (Draeger Evita 4 XL ventilator, Drägerwerk AG & Co. KgaA, Moislinger Allee 53-55, 23558 Lübeck, Germany) whereas P_ETCO₂ was measured by micro-stream sampling (Philips server, model M3015A/B, attached to the Philips MP30 and MP70 IntelliVue monitors, calibrated with each device check, Philips Healthcare, P.O. Box 10.000, 5680 DA Best, The Netherlands).

We collected above data and clinical diagnoses of children who were mechanically ventilated in the PICU over a 12-month period (January–December 2012). Children who had (a) an end tidal CO₂ recording; (b) an arterial line for blood gas measurements and (c) were “clinically stable” (i.e, not in a resuscitation phase, and hemodynamically stable); and (d) on invasive bi-level positive airway pressure (BiPAP) ventilation were recruited. The maximum number of concurrent readings obtained from any single patient, if available, was capped to the first 16. All children had cuffed endotracheal tubes and minimal air leaks. A specific data collection form was used. They also had intermittent arterial blood gas measurements (corrected to 37°C) and continuous pulse oximetry (SpO₂). We retrieved concurrently recorded FiO₂, SpO₂, P_ETCO₂, PaCO₂, PaO₂, data every 4-6 h from all qualifying children using the “Metavision” electronic database. Children with suspected intracardiac shunting and/or poor cardiac output states were excluded. Furthermore, children on high-frequency oscillation (HFO) were excluded as they had no P_ETCO₂ measurement.

The standard alveolar-gas equation (Eq. 1) was used to calculate the alveolar oxygen tension (PAO₂).[7] The alveolar-arterial oxygen tension difference [P(A-a)O₂] was estimated by subtracting the concurrent measured PaO₂ from the calculated PAO₂ in mmHg.

Equation 1: The alveolar gas equation: PAO₂ = alveolar oxygen tension in mmHg; FiO₂ = fractional inspired oxygen concentration; P_atm = atmospheric pressure in mmHg; P_H2O = partial pressure of water (47 mmHg at 37°C); and R = 0.8

PAO₂ = (FiO₂× [P_atm − P_H2O]) − (PaCO₂÷ R)      (1)

Statistics

The P(A-a)O₂ difference was interpreted as indicative of the degree of intra-pulmonary shunt. Partial pressures were expressed in mmHg where necessary using a standard conversion (1 KPa = 7.5006 mmHg). The variability of the P_ETCO₂ and PaCO₂ difference (between children and between disorders) was studied using one-way ANOVA. The performance of P_ETCO₂ and PaCO₂ measurement methods were compared using a Bland-Altman plot.[8]

Ethics

This was an observational study of routinely monitored respiratory and hemodynamic parameters of children requiring mechanical ventilation in the PICU. There was neither any direct intervention in patient management nor a need to use any patient identifiable information. Therefore, according to local guidelines, an Ethics Committee review was not required. The study was registered as a service evaluation project.

Results

We obtained 624 concurrent readings from 105 children requiring invasive bi-level ventilation in the PICU. All children were in a clinically steady state and had an arterial line in situ for blood gas measurements and continuous P_ETCO₂ monitoring. The demographic data of the study population are shown in Table 1.

Table 1.

Demographic data of the study population

The P_ETCO₂ values recorded were higher than the concurrent PaCO₂ (i.e, PaCO₂ /P_ETCO₂ ratio was <1) in 142 observations (22.7%). The proportions with PaCO₂ /P_ETCO₂ ratio <1 in the subcategories were, liver disorder 26.0%, respiratory disorder 20.8%, central nervous system disorder 16.5%, septic shock (multiple organ failure) 33.3%, and others 26.7% respectively.

The PaCO₂–P_ETCO₂ difference was not significantly influenced by the disorders leading to mechanical ventilation (df 4, ANOVA, P = 0.6) [Table 1]. However, the mean PaCO₂–P_ETCO₂ difference was individual specific (df 103, ANOVA, P < 0.0001) [Figure 1]. This finding is limited by the fact that, in larger group analyses, ANOVA in SPSS (IBM United Kingdom Limited, PO Box 41, North Harbour, Portsmouth, Hampshire, PO6 3AU) only tells you at least one group in the analysis is different from at least one other. For the confirmation of this finding follow-up data are needed in the same groups and this is obviously not realistic in our cohort.

Figure 1.

Individual variability of arterial carbon dioxide tension (PaCO₂)–end-tidal carbon dioxide tension (P_ETCO₂) difference during a single pediatric intensive care unit admission

The PaCO₂–P_ETCO₂ difference positively correlated with P(A-a)O₂ difference (two-tailed Pearson's correlation, ρ = 0.381, P < 0.0001) [Figure 2]. Please note scarcity of data at high FiO₂ concentrations. This is because these children were on HFO and excluded. The PaCO₂–P_ETCO₂ difference did not correlate with age when averaged.

Figure 2.

The arterial carbon dioxide tension (PaCO₂)–end-tidal carbon dioxide tension (P_ETCO₂) difference correlates positively with alveolar-arterial oxygen tension [P(A-a)O₂] difference (two-tailed Pearson correlation 0.381 P < 0.0001)

We compared PaCO₂ and P_ETCO₂, without bias, using the Bland-Altman method.[8] We found that the average reading of PaCO₂ was consistently +0.66 KPa (95% confidence interval: +0.57 to +0.76) higher than that of P_ETCO₂ [Figure 3].

Figure 3.

A Bland-Altman plot comparing concurrent arterial carbon dioxide tension (PaCO₂) and end-tidal carbon dioxide tension (P_ETCO₂) measurements (bias = 0.66, standard deviation = 1.22, limits of agreement = −1.72, 3.05, bias confidence interval [CI] 95% CI = 0.57-0.76, lower limit of agreement CI 95% CI = −1.89 to −1.56, upper limit of agreementCI 95% CI = 2.88–3.21)

Discussion

P_ETCO₂ allows global evaluation of three main bodily functions: Metabolism, circulation, and ventilation.[9] Our study was designed to investigate the relationship of P_ETCO₂ and PaCO₂ in PICU patients, during different disease states. Circulation, ventilation and metabolic parameters were considered to be in a steady state whilst on invasive bi-level mechanical ventilation. We have shown that PaCO₂–P_ETCO₂ difference is consistent in each individual. Thus, in a clinically “steady” state of mechanical ventilation, the P_ETCO₂ and PaCO₂ difference can be expected to be static but unique to the individual in that setting. We also observed that the variability of PaCO₂–P_ETCO₂ was not affected by the primary disorder leading to the PICU admission.

The study demonstrates that in clinically stable children on mechanical ventilation, irrespective of their disease status, the P_ETCO₂ and PaCO₂ difference is consistent but individual specific. The corollary of this finding is that if an unstable difference between PaCO₂ and P_ETCO₂ becomes stable, this may suggest a clinical improvement as far as the ventilation/perfusion (V/Q) status of the lungs is concerned. Through attention to PaCO₂–P_ETCO₂ difference in PICU patients, unnecessary blood gas measurements can be avoided. A trend analysis of PaCO₂–P_ETCO₂ difference may therefore be valuable in the PICU setting.

Efforts to study P_ETCO₂ against PaCO₂ are not new. In order to have a noninvasive measure of PaCO₂, several authors have investigated the relationship between PaCO₂ and P_ETCO₂ with mixed results. Sharma found the P_ETCO₂ -PaCO₂ mean difference to be stable and patient specific but variable between individuals.[10] Our findings go a step further. P_ETCO₂ alone is not sufficiently reliable to replace arterial blood sampling based measurements,[11] but becomes useful once the PaCO₂–P_ETCO₂ difference is known.

Capnography provides a numerical measurement of inspired and end tidal CO₂. Previous investigators have reported an average P_ETCO₂ -PaCO₂ difference of 4-6 mmHg in patients with normal lungs.[11,12,13,14,15] In patients with respiratory failure, the average PaCO₂–P_ETCO₂ difference was 18 mmHg.[16] We noted a difference of 4.9 mmHg (0.66 KPa) in our study.

Intra-pulmonary shunting in lung disease due to alveolar collapse/consolidation and/or diffusion abnormalities minimally affects CO₂ elimination in the early stages.[17] With worsening lung disease, right-to-left shunting increases and results in poor oxygenation and carbon dioxide elimination. Therefore, for any given degree of arterial oxygen de-saturation there is an associated and obligatory PaCO₂–P_ETCO₂ difference. Using the shunt equation, we have shown a positive correlation between the PaCO₂–P_ETCO₂ difference and the P(A-a)O₂ difference, and this provides an explanation to the above observations. As the FiO₂ requirement increases, the greater the PaCO₂–P_ETCO₂ difference.

Approximately 20% of our P_ETCO₂ values read higher than the concurrent PaCO₂ measurement, resulting in a negative PaCO₂–P_ETCO₂ difference. Similar negative PaCO₂–P_ETCO₂ differences have been observed more than 30 years ago by Nunn and Hill during anesthesia but no explanation was offered.[12] There does not appear to be any correlation between this negative difference and the primary disease. This suggests the factors contributing to this difference may be more technical and not related to the primary disease. For example, McSwain et al. has recently shown an increase in the PaCO₂–P_ETCO₂ difference with rising dead space/tidal volume ratio.[18] Fletcher and Jonson observed negative or zero PaCO₂–P_ETCO₂ values in 12% of normal subjects during intermittent positive pressure ventilation with large tidal volumes and low frequencies under anesthesia.[19] Negative PaCO₂–P_ETCO₂ values were also observed during anesthesia in 50% of pregnant subjects, in 8.1% of patients after postcardiac bypass and in 50% of infants.[20,21,22,23,24,25]

The following hypotheses may explain the observed negative PaCO₂–P_ETCO₂ differences. Low tidal volume and high frequency ventilation settings may result in poor ventilation of dependent but well-perfused alveoli, resulting in worse V/Q matching. The gas emptying from these slow alveoli may remain in the airways during small frequent breaths. Under these circumstances, the low V/Q areas (alveoli with higher PCO₂) make a higher contribution to overall gas exchange. The net effect of these factors is to enable the terminal part of phase III of the end tidal CO₂ wave to exceed mean PaCO₂, resulting in a negative PaCO₂–P_ETCO₂ .

Alternatively, very elevated mixed venous PCO₂, as in exercise or septic shock, may contribute to a negative PaCO₂–P_ETCO₂ difference. In healthy adults, PaCO₂–P_ETCO₂ difference was shown to be inversely related to the frequency of breathing and directly related to tidal volume and CO₂ output.[26] A negative PaCO₂–P_ETCO₂ difference in pregnancy has been attributed to higher CO₂ production. The same explanation may hold true for children and infants with relatively higher metabolic rates. This study demonstrates that 33% of children in septic shock have negative PaCO₂–P_ETCO₂ observations.

The study had several limitations. For example, the standard alveolar gas equation is a simplification of the actual relationship between FiO₂, PaCO₂, and PaO₂ and may deviate up to 10 mmHg (when FiO₂ = 1.0) from the more rigorous, full calculation. In addition, values used in the equation may not be precisely known, particularly the value of respiratory quotient of 0.8 (R), which may shift depending upon the relative utilization of carbohydrate, protein, and fat. The breath-to-breath variability of respiratory rate or tidal volume that may occur during invasive BiPAP mode ventilation has also not been accounted for.

Conclusion

The PaCO₂–P_ETCO₂ difference is individual specific in mechanically ventilated children. It is not affected by the disorder leading to mechanical ventilation. The P(A-a)O₂ difference has a minor but a significant influence upon PaCO₂–P_ETCO₂ difference. In a “steady” state of mechanical ventilation the PaCO₂–P_ETCO₂ difference is static.