Comparing the novel microstream and the traditional mainstream method of end-tidal CO2 monitoring with respect to PaCO2 as gold standard in intubated critically ill children

The objective of this study was to evaluate a novel microstream method by comparison with PaCO2 and the more standard mainstream capnometer in intubated pediatric patients. We hypothesized that the novel microstream method would superior compared to the traditional mainstream method in predicting PaCO2. This was a prospective single-center comparative study. The study was carried out on 174 subjects with a total of 1338 values for each method. Data were collected prospectively from mainstream and microstream capnometer simultaneously and compared with PaCO2 results. Although both mainstream PetCO2 (mainPetCO2) and microstream PetCO2 (microPetCO2) were moderately correlated (r = 0.63 and r = 0.68, respectively) with PaCO2 values, mainPetCO2 was in better agreement with PaCO2 in all subjects (bias ± precision values of 3.8 ± 8.9 and 7.3 ± 8.2 mmHg, respectively). In those with severe pulmonary disease, the mainPetCO2 and microPetCO2 methods were highly correlated with PaCO2 (r = 0.80 and r = 0.81, respectively); however, the biases of both methods increased (14.8 ± 9.1 mmHg and 16.2 ± 9.0 mmHg, respectively). In cases with increased physiologic dead space ventilation, the agreement levels of mainPetCO2 and microPetCO2 methods became distorted (bias ± precision values of 20.9 ± 11.2 and 25.0 ± 11.8 mm Hg, respectively) even though mainPetCO2 and microPetCO2 were highly correlated (r = 0.78 and r = 0.78, respectively). It was found that the novel microstream capnometer method for PetCO2 measurements provided no superiority to the traditional mainstream method. Both capnometer methods may be useful in predicting the trend of PaCO2 due to significant correlations with the gold standard measurement in cases with severe pulmonary disease or increased physiological dead space –despite reduced accuracy.


Population.
The study evaluated all children aged between 1 month to 17 years that had been intubated with cuffed ETT due to a definite indication for mechanical ventilation. Among these, those who accepted invasive monitoring of arterial blood pressure and provided informed consent (parents or legal guardians) were included in the study. The presence of any one of the following characteristics was defined as grounds for exclusion from the study: patients with tracheostomy, sampling performed with venous blood, non-compliance to study protocol (premature discontinuation of measurement, signal abnormality [absence of waveform or presence of interrupted waveform]), use of uncuffed endotracheal tubes, patients with known congenital heart and lung defects, need for high-frequency oscillatory ventilation or extracorporeal life support, determination of any type of air leakage in the lung (pneumothorax, pneumomediastinum etc.).
Monitoring. The intubations were performed with single-lumen cuffed ETT that was appropriately sized for age and weight. CO 2 in the exhaled air of patients was monitored simultaneously with mainstream (Mainstream EtCO 2 ; Philips Capnostat M25O1A, Germany) and microstream (Microstream EtCO 2 ; Medtronic Cap-nostream35, USA) capnometers. The dimensions of the airway adapters to be used were based on the manufacturer's guidelines. The airway adaptors of both methods of measurements were kept in the same location and insertions were performed sequentially between the airway circuit and the proximal ETT. ABG were analyzed at the bedside using an ABL 90 FLEX blood gas analyzer (Radiometer, Medical ApS, Copenhagen, Denmark) within 3 min of collection and without any delay. No additional ABG was performed for the data collection of consecutive samples.
Study protocol and recording. ABG analysis, mainPetCO 2 and microPetCO 2 values and mechanical ventilator parameters were recorded simultaneously. Prior to obtaining each arterial blood gas sample, a researcher checked whether the capnometer adapters were blocked by secretions or moisture. Capnometer adapters were replaced with new ones in the event of any type of blockage. Both capnometer methods were analyzed using continuous steady waveforms of expired CO 2 through the ventilator cycle, in order to ensure the accuracy of readings. A minimum of 4 and a maximum of 8 simultaneous PCO 2 measurements (PaCO 2 , mainPetCO 2 , microPetCO 2 ) were planned to be taken from each patient. Patients with a measurement number less than 4 for various reasons (death, extubation, interruption of monitoring etc.) were excluded from the study.
For subgroup analyses, patients were grouped with regard to the severity of pulmonary disease and physiological dead space ventilation levels. Severe pulmonary disease was defined as an OI of ≥ 10 and mild-to-moderate pulmonary disease was defined as an OI of < 10 18,24 . Determination of Vd/Vt ratio ≥ 0.4 was defined as increased dead space ventilation 23,25 . The consistency of PetCO 2 monitoring (mainPetCO 2 and microPetCO 2 ) within each patient and subgroup was assessed by examining the relationships between the changes in PaCO 2 and the two PetCO 2 methods in consecutive samples.
Statistical analysis. Analyses were performed by using the IBM Statistical Package for the Social Sciences version 21 (SPSS, Inc., Chicago, IL) or Med Calc version 19.1 (Med Calc Software, Ostend, Belgium). Patient characteristics are described using qualitative variables (using frequencies and percentages) and quantitative variables (using means and standard deviation [SD] or median with interquartile range [IQR], depending on type of distribution). Simple linear regression analysis was performed and Spearman correlation coefficients were calculated for the assessment of relationships between PaCO 2 , main-PetCO 2 and micro-PetCO 2 . We assessed the agreement between these measurements (bias [mean difference] and precision [SD of the differences]) by the Bland-Altman technique. The results were considered statistically significant in tests resulting in a P value lower than 0.05.

Power analysis.
Power analysis was conducted using the Power Analysis Sample Size (PASS) for Windows version 11.0 Package Program. Group sample sizes of 174 were determined to achieve 97% power to detect a difference of − 3.6 between the null hypothesis that both group means were 3.8, and the alternative hypothesis that the mean of group 2 was 7.4 (with estimated group standard deviations of 9.0 and 8.3), and with a significance level (alpha) of 0.05 using a two-sided two-sample t-test.
Ethics approval and consent to participate. The study was approved by the institutional review board of our center. The parents of all patients signed an informed consent form before inclusion into the study.

Conference presentation.
This study presented at the 15th National Congress of Pediatric Emergency and Critical Care, 18-20 October 2018, Turkey.

Results
The study was performed in 174 patients that provided 1338 measurements for each method. The median age and interquartile range (IQR) of the included subjects was 42 months (IQR: 12-108 mo.). Table 1 shows the characteristics of the study group. Conventional invasive mechanical ventilator modes were used in all patients included in the study (Galileo Mechanical Ventilator, Hamilton Medical AG, Rhäzüns, Switzerland).
The median (range) levels of PaCO 2 , mainPetCO 2 , and microPetCO 2 Table 2 and illustrated in Fig. 1 in all subject groups-and also according to severity of pulmonary disease. In all subjects (1338 pairs), the mean difference (bias) and SD of the differences (precision) for mainPetCO 2 was 3.8 ± 8.9 mm Hg (95% limits of agreement -13.7 to 21.4 mm Hg) with moderate correlation (r = 0.63, p < 0.001) (Fig. 1A). The mean bias and precision for microPetCO 2 was 7.3 ± 8.2 mm Hg (95% limits of agreement − 8.8 to 23.6 mm Hg) with moderate correlation (r = 0.68, p < 0.001) (Fig. 1B). Although both PetCO 2 measurement methods were moderately correlated, mainPetCO 2 was more accurate compared to the microPetCO 2 method overall (in the whole subject group) . Additionally, when we evaluated the correlation between mainPetCO 2 and microPetCO 2 throughout all patients, the methods demonstrated a strong level of correlation (r = 0.84, p < 0.001) (Fig. 2).
Study subjects were also compared based on the presence of lung pathology (Table 2). In the mild-to-moderate pulmonary disease group, 1242 measurements from each end-tidal CO 2 method were compared. In this group, the mean bias and precision for mainPetCO 2 was 2.9 ± 8.4 mm Hg (95% limits of agreement − 13.4 to 19.4 mm Hg) with moderate correlation (r = 0.64, p < 0.001) (Fig. 1C). The mean bias and precision for microPetCO 2 was 6.7 ± 7.8 mm Hg (95% limits of agreement − 8.5 to 22.1 mm Hg) with moderate correlation (r = 0.68, p < 0.001) (Fig. 1D). Although both PetCO 2 measurement methods were moderately correlated, mainPetCO 2 was more accurate than microPetCO 2. In the severe lung disease group, we compared 96 results from each method of measurement. For the mainPetCO 2 and PaCO 2 comparison, the mean bias and precision was 14.8 ± 9.1 (95% limits of agreement − 3.0 to 32.7 mm Hg) (Fig. 1E). Whereas the mean bias and precision between microPetCO 2 and PaCO 2 was 16.2 ± 9.0 mm Hg (95% limits of agreement − 1.4 to 33.9 mm Hg) (Fig. 1F). In the severe lung disease group, almost all PaCO 2 values were higher than PetCO 2 measurements (Fig. 1E,F) www.nature.com/scientificreports/ mainPetCO 2 and microPetCO 2 were highly correlated with PaCO 2 (r = 0.80, p < 0.001 and r = 0.81, p < 0.001, respectively); however, the biases of both methods increased.
To determine whether the accuracy of the non-invasive PCO 2 measurement methods were altered in the presence of high physiologic dead space, we compared the mainPetCO 2 and microPetCO 2 values with regard

Laboratory values, median (IQR)
Arterial blood gas analysis  In the Vd/Vt < 0.4 (the normal physiologic dead space) group, the comparison of mainPetCO 2 and PaCO 2 values showed a mean bias and precision of 3.0 ± 8.0 mm Hg, with moderate correlation (r = 0.63, p < 0.001). Whereas the mean bias and precision between microPetCO 2 and PaCO 2 was 6.5 ± 7.0 mm Hg, again with moderate correlation (r = 0.68, p < 0.001). In the Vd/Vt ≥ 0.4 (increased physiologic dead space) group, both mainPetCO 2 and microPetCO 2 were highly correlated (r = 0.78, p < 0.001 and r = 0.78, p < 0.001, respectively) with increased PetCO 2 -PaCO 2 gradient (bias ± precision values of 20.9 ± 11.2 and 25.02 ± 11.8 mm Hg, respectively). Although both non-invasive PCO 2 measurement methods were highly correlated with PaCO 2 , mainPetCO 2 was more accurate than microPetCO 2 in both the normal and increased dead space ventilation groups.

Discussion
To our knowledge, this is the largest cohort study including 174 pediatric patients who received mechanical ventilation in the PICU. The evaluation of 1338 measurements for each method and the comparison of two different PetCO 2 monitoring methods with accuracy determined according to simultaneous PaCO 2 measurements are among the other strengths of this study. Although different PetCO 2 measurement methods have distinct advantages, the accuracy and correlation of these methods in comparison to ABG measurements is without doubt the most vital feature of any method. The microstream capnometer requires in-depth analysis to prove that it contributes to or surpasses available methods by analyzing whether the advantageous properties expressed in the literature are indeed superior in the real-life follow-up of intubated pediatric patients.
Although there are many studies evaluating the accuracy and correlation of various non-invasive PetCO 2 measurement methods, the majority of these studies were performed in non-intubated patient groups [26][27][28][29][30] . In intubated patients, the studies on PetCO 2 monitoring are mostly compared with the ABG analysis of a single method and often evaluate the relationship between the severity of lung disease and the accuracy of the method. In our study, two different PetCO 2 monitoring methods were evaluated simultaneously, and both mainPetCO 2 and microPetCO 2 measurements were found to be moderately correlated with PaCO 2 .
Rozycki et al. reported that mainPetCO 2 measurements were highly correlated with PaCO 2 in intubated newborns, with a mean bias of -6.9 mm Hg 13 . Similar results have been found in other studies using the mainstream technology in intubated newborns 31,32 . Microstream is preferred especially in the neonatal age group due to the use of very low flow rates (50 mL/min), causing smaller dead space and allowing measurement from the distal part of the ETT. In the study by Kugelman et al. 24 microPetCO 2 was found in adequate agreement with PaCO 2 , which indicated closer agreement than seen in the current research. Although similar 'close' results have been obtained in other studies 18,33 , Singh and colleagues found similar results to ours in terms of agreement between microPetCO 2 and PaCO 2 15 . In intubated patients, PetCO 2 measurements can be performed from the proximal or distal part of the ETT. To compare the advantages of different PetCO 2 measurement technologies in our study, it was thought that the measurements obtained from the same locations would be more guiding. Therefore, in order for one of the methods to gain no advantage due to localization, both PetCO 2 measurements were obtained from the same location (proximal part of ETT). In various studies comparing PetCO 2 measurements obtained from the distal and proximal parts of the ETT, it has been suggested that distal measurements provide more accurate results; however, several other studies have demonstrated comparable accuracy between proximal and distal measurements 21,33-35 .
The first study comparing two different PetCO 2 measurement methods in intubated patients was performed by Kugelman and colleagues. This study, which was comprised of 27 infants, showed better correlation between PetCO 2 and PaCO 2 with distal sampling of expired air using microstream technology against the mainstream method through a proximal port using double lumen ETT 18 . The measurements made in this study were obtained from different locations of the ETT and this situation may have led to an advantage for the microstream method. In our study, although the correlation coefficients of both methods were similar, the agreement level of main PetCO 2 measurements was better.
There are various studies investigating the relationships between pulmonary disease and PaCO 2 -PetCO 2 values. These studies have defined pulmonary disease severity according to various parameters, such as OI, arterial-alveolar PO 2 gradient and PaO 2 /FiO 2 ratio. In this study we used the OI value to define severe pulmonary disease. Sivan et al. reported in their study that mainPetCO 2 and PaCO 2 compatibility decreased as lung disease severity increased in neonatal patients 36   www.nature.com/scientificreports/ intubated newborn patients and found that microPetCO 2 and PaCO 2 differences were higher in the pulmonary disease group compared to controls 33 . Different results were reported by other investigators. Tingay et al. 37 found that the PetCO 2 bias was independent of severity of lung disease and similarly Rozycki et al. 13 reported that the degree of lung disease had little influence on the degree of discrepancy between measurement. Kugelman and colleagues reported that although the accuracy of microPetCO2 decreased with lung disease it still remained good correlation as a useful measure of PaCO2 in conditions of severe lung disease 18 . The study by McDonald et al. found an overall moderately correlation between PaCO 2 and mainPetCO2 for all included patients, but the investigators concluded that significant lung disease (defined by PaO 2 /FiO 2 < 200) had a negative effect on the correlation 12 . In our study, it was concluded that both PetCO 2 measurement methods highly correlated in patients with severe lung disease, albeit with a significant decrease in measurement accuracy. The most important parameter contributing to the PetCO 2 -PaCO 2 gradient is the increase in physiological dead space due to ventilation-perfusion mismatch 38,39 . Physiologic dead space ventilation is the sum of anatomical dead space from the conducting airways and alveolar dead space from disease processes and/or therapies employed. The increased gradient between PetCO 2 and PaCO 2 with high PaCO 2 levels are directly proportional to the degree of physiologic dead space. Although typical alveolar CO 2 concentrations are slightly greater than of ABG, PetCO 2 normally 2-5 mmHg lower than PaCO 2 due to mixing of CO 2 containing alveolar gas with exhaled gas devoid of CO 2 from the anatomical dead space. In a patient with lung disease, the addition of alveolar dead space further dilutes PetCO 2 relative to PaCO 2 . As a result, PetCO 2 measurements depict greatly reduced results compared to PaCO 2 . The normal physiologic dead space to tidal volume ratio (Vd/Vt) is established to be 0.20-0.35 23 . In this study, we provide evidence that physiologic dead space ventilation is a major factor in determining the relationship between capnographic monitoring of PetCO 2 and PaCO 2 . Despite multiple earlier publications comparing PetCO 2 and PCO 2 in presence of pulmonary disease and hypercarbia, few studies have examined the effect of change in physiologic dead space on the relationship between PetCO 2 and PaCO 2 across an increased range of Vd/Vt ratios in mechanically ventilated pediatric patients 23 . Our study is the first to investigate the correlations between two different capnometers in patients with increased physiological dead space ventilation.
In patients with a low calculated physiologic dead space to tidal volume ratio (Vd/Vt < 0.4), there is a moderate correlation between both PetCO 2 (measured noninvasively by capnography) measurements and PaCO 2 value. Despite the high correlation between PetCO 2 and PaCO 2 values in patients with high physiologic dead space to tidal volume ratio (Vd/Vt ≥ 0.4), the accuracy of measurements was greatly reduced. Therefore, in the presence of severe pulmonary diseases with increased physiological dead space, it is much more reliable to use PetCO 2 results as a measure of trend rather than absolute value. It is also critical to note that further problems in accuracy may arise with smaller infants or newborns (which were not included in the study population) and reduced volumes or I:E values. In a study including 56 intubated pediatric patients by McSwain et al., it was found that, while the strength of the association diminished slightly as the dead space ratio increased, the correlation still remained strong between the methods. The PaCO 2 -PetCO 2 gradient was increased predictably with increasing Vd/Vt 23 . Our findings show that increased physiological dead space as a result of severe pulmonary disease will increase the gradient between PaCO 2 and PetCO 2 in favor of PaCO 2 values, making almost all PaCO2 results greater than those recorded by PetCO 2 . These findings were similar to the outcomes of previous studies performed in newborns and children with pulmonary disease 33,36 .
To our knowledge, there are no other studies investigating the relationships between PetCO 2 measurements and increased physiological dead space ventilation. There are however, various studies investigating PetCO 2 correlations with hypercarbia as a proxy for increased dead space ventilation. In the study conducted by Kugelman et al., microPetCO 2 was reported as a useful measure of PaCO 2 , whereas mainPetCO 2 was distorted on the high range of PaCO 2 level 18 . Rosycki et al. 13 did not find any effect of increased PaCO 2 on mainPetCO 2 measurements.
Our study has several limitations. Non-consecutive ABGs were used for data collection and inadvertent selection bias may have been introduced. In our study, we used proximal measurement method for both PetCO 2 methods. In subsequent studies, the relationship between concurrent microPetCO 2 measurements obtained from the proximal and distal part of the ETT may reveal differences in results which could be crucial for physicians and patients in intensive care units. Although our study reached the highest number of patients and samples in the literature, the number of samples in the subgroups of severe pulmonary disease and increased physiologic dead space ventilation, were relatively low; thus limiting the generalizability of those results. Due to the low number of patients with ARDS, we could not group patients as mild, moderate, severe ARDS with regard to the criteria put forth by the Pediatric Acute Lung Injury Consensus Conference (PALICC); thus, subgroup analyses concerning these groups could not be performed. Also, the number of cases with increased physiologic dead space (Vd/Vt ≥ 0.4) was low, leading to a lack of further subgroup analysis.

Conclusion
It was found that the novel microstream capnometer has no superiority to the traditional mainstream method. Although the mainstream and microstream capnometer measurements had similar correlation values with ABG results, the agreement level of the mainstream method was higher. Although the absolute gradient between both PetCO 2 methods and PaCO 2 results demonstrated a consistent increase in the presence of severe pulmonary disease and increased dead space ventilation, both methods showed significant correlations with PaCO 2 values. Therefore, in the presence of severe pulmonary disease and/or increased dead space ventilation, it is possible that both PetCO 2 monitoring methods may be helpful in predicting the trend of PaCO 2 despite limitations in accuracy.