The relationship between minute ventilation and end tidal CO2 in intubated and spontaneously breathing patients undergoing procedural sedation

Background Monitoring respiratory status using end tidal CO2 (EtCO2), which reliably reflects arterial PaCO2 in intubated patients under general anesthesia, has often proven both inaccurate and inadequate when monitoring non-intubated and spontaneously breathing patients. This is particularly important in patients undergoing procedural sedation (e.g., endoscopy, colonoscopy). This can be undertaken in the operating theater, but is also often delivered outside the operating room by non-anesthesia providers. In this study we evaluated the ability for conventional EtCO2 monitoring to reflect changes in ventilation in non-intubated surgical patients undergoing monitored anesthesia care and compared and contrasted these findings to both intubated patients under general anesthesia and spontaneously breathing volunteers. Methods Minute Ventilation (MV), tidal volume (TV), and respiratory rate (RR) were continuously collected from an impedance-based Respiratory Volume Monitor (RVM) simultaneously with capnography data in 160 patients from three patient groups: non-intubated surgical patients managed using spinal anesthesia and Procedural Sedation (n = 58); intubated surgical patients under General Anesthesia (n = 54); and spontaneously breathing Awake Volunteers (n = 48). EtCO2 instrument sensitivity was calculated for each patient as the slope of a Deming regression between corresponding measurements of EtCO2 and MV and expressed as angle from the x-axis (θ). All data are presented as mean ± SD unless otherwise indicated. Results While, as expected, EtCO2 and MV measurements were negatively correlated in most patients, we found gross systematic differences across the three cohorts. In the General Anesthesia patients, small changes in MV resulted in large changes in EtCO2 (high sensitivity, θ = -83.6 ± 9.9°). In contrast, in the Awake Volunteers patients, large changes in MV resulted in insignificant changes in EtCO2 (low sensitivity, θ = -24.7 ± 19.7°, p < 0.0001 vs General Anesthesia). In the Procedural Sedation patients, EtCO2 sensitivity showed a bimodal distribution, with an approximately even split between patients showing high EtCO2 instrument sensitivity, similar to those under General Anesthesia, and patients with low EtCO2 instrument sensitivity, similar to the Awake Volunteers. Conclusions When monitoring non-intubated patients undergoing procedural sedation, EtCO2 often provides inadequate instrument sensitivity when detecting changes in ventilation. This suggests that augmenting standard patient care with EtCO2 monitoring is a less than optimal solution for detecting changes in respiratory status in non-intubated patients. Instead, adding direct monitoring of MV with an RVM may be preferable for continuous assessment of adequacy of ventilation in non-intubated patients.


Methods
M1winute Ventilation (MV), tidal volume (TV), and respiratory rate (RR) were continuously collected from an impedance-based Respiratory Volume Monitor (RVM) simultaneously with capnography data in 160 patients from three patient groups: non-intubated surgical patients managed using spinal anesthesia and Procedural Sedation (n = 58); intubated surgical patients under General Anesthesia (n = 54); and spontaneously breathing Awake Volunteers (n = 48). EtCO 2 instrument sensitivity was calculated for each patient as the slope of a Deming regression between corresponding measurements of EtCO 2 and MV and expressed as angle from the x-axis (θ). All data are presented as mean ± SD unless otherwise indicated. PLOS

Introduction
Whereas it is standard practice to both control and monitor ventilation during general anesthesia, it is equally important to monitor ventilation in non-intubated patients undergoing procedural sedation. End tidal CO 2 (EtCO 2 ) monitoring with capnography has become the standard of care in intubated patients for both confirming endotracheal tube placement and monitoring adequacy of ventilation [1,2]. Capnography with an endotracheal tube in place is considered a reliable method to non-invasively reflect arterial PaCO 2 [3,4], however, measuring EtCO 2 in spontaneously breathing patients can be inaccurate in certain settings, particularly during procedural sedation [5] and post-operatively in the post-anesthesia care unit [6][7][8]. Variables such as sensor positioning, changes in respiratory patterns, and changes in oxygen supplementation often distort EtCO 2 measurements in non-intubated patients, rendering them unreliable. As a result, it is common for healthcare providers to overlook or discount information obtained from the capnography waveform [9]. Furthermore, since EtCO 2 is an indirect indicator of respiratory status, it reflects ventilatory changes later than a direct measurement of ventilation, like minute ventilation (MV). This difference is especially important in spontaneously breathing subjects [10]. In current clinical applications, non-invasive measurements of EtCO 2 do not allow health care practitioner to identify subtle changes in ventilation over a background of confounding factors such as ventilation/perfusion mismatch, partial airway obstruction, or metabolic derangement. For these reasons, capnography has never achieved wide clinical adoption in non-intubated patients [11].
Given the limitations of EtCO 2 monitoring in precisely reflecting the respiratory status of patients, more emphasis may need to be placed on methods of volumetrically monitoring ventilation in non-intubated patients. With the introduction of a non-invasive Respiratory Volume Monitor (RVM) that can provide accurate measurements of MV, tidal volume (TV), and respiratory rate (RR) in non-intubated patients [12,13], direct monitoring of ventilation in non-intubated patients has become available both inside and out of the operating room. Here we studied the ability for conventional EtCO 2 monitoring to reflect changes in ventilation in non-intubated surgical patients undergoing spinal anesthesia and procedural sedation. We computed the instrument sensitivity of a EtCO 2 monitor when detecting changes in MV in these non-intubated surgical patients and compared and contrasted this sensitivity to both intubated patients under general anesthesia and spontaneously breathing volunteers.

Experimental design
Continuous respiratory data (MV, TV, and RR) were collected from an impedance-based RVM (ExSpiron, Respiratory Motion, Inc., Waltham, MA) simultaneously with capnography data (EtCO 2 ) in patients in three groups: patients under Procedural Sedation (1), patients under General Anesthesia (2), and Awake Volunteers (3). The Procedural Sedation group was of primary interest in this manuscript and the other two groups were used effectively as "control" groups providing limiting conditions based on the level of sedation ranging from "none" in the Awake Volunteers to "deep" in the intubated and mechanically-ventilated patients in the General Anesthesia group. Procedural Sedation cohort. In this group, patients underwent elective joint replacement surgery with spinal anesthesia and procedural sedation. EtCO 2 data were collected from a sampling nasal cannula with oral scoop sampling port (Covidien Smart CapnoLine Plus Oral/ Nasal, Boulder, CO) using a ventilator (Dräger Apollo, Andover, MA). Anesthesia was initiated immediately prior to surgery, typically with an intrathecal dose of bupivacaine 0.5% (1.5-4.0 ml) and supplemented with midazolam, propofol, and fentanyl for sedation. Additional intraoperative opioids such as hydromorphone were rarely used in Spinal Anaesthesia cases. Typically, patients undergoing knee surgery also received a femoral nerve block, consisting of either 20 ml ropivacaine 0.2% or 20 ml bupivacaine 0.25%, administered in preoperative holding.
General Anesthesia cohort. In this group, patients underwent elective joint replacement surgery under general anesthesia. MV and EtCO 2 data were collected from the endotracheal tube using a ventilator (Dräger Apollo, Andover, MA). Anesthesia was initiated immediately prior to surgery, with various doses of a muscle relaxant (rocuronium, vecuronium, or cisatracurium), in conjunction with sedatives (midazolam, propofol, and ketamine), and opioids (fentanyl, hydromorphone, meperidine, remifentanil, and morphine). Typically, patients undergoing knee surgery also received a femoral nerve block, consisting of either 20 ml ropivacaine 0.2% or 20 ml bupivacaine 0.25%, administered in preoperative holding. A detailed summary of relevant medications used intra-operatively in the Procedural Sedation and General Anaesthesia cohorts, with frequency and dosage, can be found in Table 1.
Awake Volunteers cohort. In this group, spontaneously breathing subjects performed a total of six breathing trials at varying prescribed respiratory rates for a total of 13 min. In the first and last trials, subjects were instructed to breathe normally, while in the middle four trials, subjects alternated between fast (25 bpm) and slow (5 bpm) as set by a metronome. EtCO 2 data were collected from a sampling nasal cannula with oral scoop sampling port (Covidien Smart CapnoLine Plus Oral/Nasal, Boulder, CO) using a dedicated capnograph (Capnostream 20, Covidien, Boulder, CO). All subjects responded to an Institutional Review Board-approved advertisement.
Equipment. In all three cohorts, the RVM collected bio-impedance traces via an electrode padset placed in the recommended positions: sternal notch, xiphoid, and right mid-axillary line at the level of the xiphoid. The skin was prepped and the padset applied in a fashion similar to that used in standard ECG electrode placement. At the beginning of the study, the RVM was calibrated against a ventilator in the General Anesthesia group, a Wright sprirometer (Mark 14, nSpire Health, Inc., Longmont, CO) in the Procedural Sedation group, and a heated pneumoatchometer (Heated FVL, Morgan Scientific, Haverhill, MA) in the Awake Volunteers group.
Institutional review board and consent. Inclusion criteria for the all three cohorts were English-speaking men and women aged 18 years to 99 years. Exclusion criteria for the Procedural Sedation and General Anesthesia groups were pregnant females, patients with an electronic implantable device, and surgery positions other than supine or lateral. Exclusion criteria for the Awake Volunteers group were hospitalization within 30 days before the study and pregnant females. The study for the Procedural Sedation and General Anesthesia cohorts was approved by the Partners Institutional Review Board, Boston, MA (2011P002898). The study for the Awake Volunteers group was approved by the Schulman Associates Institutional Review Board, Cincinnati, OH (201102306). All patients gave informed written informed consent.

Data and statistical analysis
The ability of EtCO 2 to reflect changes in MV (instrument sensitivity) was calculated for each patient. Specifically, instrument sensitivity was defined as the slope calculated by a Deming regression between individual corresponding measurements of EtCO 2 and MV. The slopes of the regression were presented as angles from the x-axis (y ¼ tan À 1 DEtCO 2 DMV À Á ). A steep correlation line (i.e., θ % -90˚) corresponds to high instrument sensitivity, indicating a small change in MV leads to a large change in EtCO 2 . A flatter correlation line (i.e., θ % 0˚) corresponds to low instrument sensitivity, indicating a small change in MV results in almost no change in EtCO 2 . The stated EtCO 2 accuracies of the ventilator and capnograph used in this study were ±3.8 mm Hg [14] and ±2.0 mm Hg [15], respectively. Previous work suggests that the minimally acceptable instrument sensitivity for clinically-relevant EtCO 2 monitoring is -4.0 mmHg / L/min (i.e., θ = -76˚) [10]. For each patient, MV was calculated as a percent of their individual predicted MV (MV PRED ), based on each patient's body surface area and sex [16,17], which has been shown to be a better predictor of actually observed MV during spontaneous respiration than MV PRED based on Ideal Body Weight (IBW) [18]. Unbalanced one-way ANOVAs were used to compare demographics, instrument sensitivities, as well as average EtCO 2 and MV measurements across cohorts. All data are presented as mean ± SD unless otherwise indicated.

Results
Data were collected from 160 patients across the three cohorts (Table 2). Height and weight were not significantly different across the three cohorts (p = 0.12 and p = 0.17, respectively), however, BMI and age were significantly lower in the Awake Volunteers group compared to the Procedural Sedation and General Anesthesia groups (BMI: p < 0.02; age: p < 0.0001 for both comparisons). Procedural Sedation and General Anesthesia patients tended to have more comorbidities than the Awake Volunteers. All patients in the Procedural Sedation and General Anesthesia groups received supplemental oxygen and both groups had a similar average FiO 2 delivered throughout the procedure (p = 0.86). Patients in the three cohorts had similar MV PRED (p = 0.25, Table 3). During the course of the surgical procedure, General Anesthesia patients had an average MV of 80.9% MV PRED , suggesting a decreased metabolic function resulting from the anesthesia. In contrast, the Procedural Sedation patients had a significant higher average MV of 148.4% MV PRED , due to both an increased TV and RR (p < 0.0001), indicative of their lightly sedated state. In comparison, during normal breathing trials, Awake Volunteers maintained close to their predicted MV (106.3% MV PRED ).
In a given patient, a plot of EtCO 2 measurements against corresponding MV measurements produced a negatively correlated distribution: as MV increased, EtCO 2 generally decreased (Fig 1). In a representative intubated patient under General Anesthesia (blue), small changes in MV (from 5.3 to 6.7 L/min) triggered large changes in EtCO 2 (from 43.7 to 34.1 mmHg). Specifically, a 1 L/min increase in MV resulted in a 13.2 mmHg decrease in EtCO 2 , yielding a high EtCO 2 instrument sensitivity of 13.2 mmHg/L/min, equivalent to θ GA = -85.7˚(nearly vertical line, as shown in Fig 1). In contrast, in an Awake Volunteer (green), a ten times larger change in MV (from 4.3 to 25.2 L/min) was required to trigger a similar change in EtCO 2 (from 26.1 to 36.0 mmHg). For this patient, a 1 L/min increase in MV resulted in a 0.27 mmHg decrease in EtCO 2 , yielding a low EtCO 2 instrument sensitivity of 0.27 mmHg/L/min, equivalent to θ AV = -14.9˚(nearly horizontal line, as shown in Fig 1). Interestingly, patients under Procedural Sedation (red), fell between the General Anesthesia and Awake Volunteer patients. Specifically, in the example patient, a 1 L/min increase in MV led to 2.0 mmHg decrease in EtCO 2 , yielding an instrument sensitivity of 2.0 mmHg/L/min, equivalent to θ PS = -63.5˚. EtCO 2 instrument sensitivity was calculated for each patient and the distributions of the instrument sensitivities for each of the three groups were analyzed (Fig 2). The median EtCO 2 instrument sensitivities were -85.1˚, -38.1˚, and -20.2˚for the General Anesthesia, Procedural Sedation, and Awake Vounteer cohorts, respectively. The General Anesthesia and Awake Volunteers cohorts had unimodal distributions of EtCO 2 instrument sensitivity, well described by single Gaussian functions fit to these data. EtCO 2 instrument sensitivities were significantly higher in the intubated patients under General Anesthesia (θ = -83.6 ± 9.9˚) compared to nonintubated Awake Volunteers (θ = -24.7 ± 19.7˚, p < 0.0001). Interestingly, the distribution of EtCO 2 instrument sensitivity for the Procedural Sedation cohort was clearly bimodal, illustrating a lack of uniformity in this group. A mixed-model of two Gaussian was therefore fit to these data, showing that 47% (27/54) of the patients experienced high EtCO 2 instrument sensitivity (θ = -96.6 ± 15.0˚), similar to the General Anesthesia patients, while the remaining patients had low instrument sensitivity (θ = -1.2 ± 22.4˚), similar to the Awake Volunteers. Importantly, while the majority of patients in the General Anesthesia group (43/54, 80%) had EtCO 2 instrument sensitivity which showed changes in EtCO 2 in the clinically-relevant range (shaded gray area), less than half of Procedural Sedation patients (24/58, 41%) and no patients in the Awake Volunteers cohort had clinically-relevant EtCO 2 instrument sensitivity of -76c orresponding to -4 mmHg / L/min . The average EtCO 2 measurement over the length of monitoring was calculated for each patient and the distributions of these average measurements for the three groups were analyzed (Fig 3). Measured EtCO 2 values were higher in the General Anesthesia (37.2 ± 4.3 mmHg) group than the Procedural Sedation (23.3 ± 4.8 mmHg) and Awake Volunteers (31.4 ± 5.2 mmHg) groups (p < 0.0001). Interestingly, the average measured EtCO 2 of the Awake Volunteers across the prescribed breathing patterns was higher than the Procedural Sedation group (p<0.0001). The majority of General Anesthesia patients (37/54, 69%) had an average EtCO 2 measurement within the normal range of EtCO 2 (35-45 mmHg, shaded yellow area). In contrast, only 27% (13/48) of the Awake Volunteers and only 5% (3/58) of the Procedural Sedation patients had an average EtCO 2 measurement within the normal range.

Discussion
In this study, we assessed and quantified the ability of capnography to measure and reflect real-time changes in respiratory status, specifically ventilation (MV), in non-intubated patients undergoing procedural sedation. First, we quantified EtCO 2 instrument sensitivity for each patient as the slope of a Deming regression between corresponding measurements of EtCO 2 and MV. Next, we compared the instrument sensitivity between patients under Procedural Sedation and two control groups: General Anesthesia and Awake Volunteers. In the intubated patient under General Anesthesia, we found a strong relationship between MV and EtCO 2 (median EtCO 2 instrument sensitivity of -85.1˚). This EtCO 2 instrument sensitivity was better than the clinically-relevant EtCO 2 instrument sensitivity of -76˚, confirming EtCO 2 measurements in intubated patients could adequately reflect changes in MV. In contrast, in the nonintubated patients (i.e., both the Procedural Sedation and Awake Volunteer groups), the relationship between MV and EtCO 2 is much weaker (median EtCO 2 instrument sensitivities of -38.1˚and -20.2˚, respectively) and better than the clinically-relevant instrument sensitivity of -76˚in only 23% (24/106) of the non-intubated patients. This finding indicates that the EtCO 2 instrument sensitivity in non-intubated and spontaneously breathing individuals may not be adequate for detecting meaningful changes in MV in over three-fourth of patients.
We found EtCO 2 instrument sensitivity in Procedural Sedation patients exhibited a bimodal distribution with approximately half of the patients having high EtCO 2 instrument sensitivity and other half exhibiting low EtCO 2 instrument sensitivity. We saw very similar low EtCO 2 instrument sensitivity to that demonstrated in the Awake Volunteer cohort. One potential explanation for this observation is that more deeply sedated Procedural Sedation patients behaved similarly to patients under General Anesthesia and therefore showed high EtCO 2 instrument sensitivity, while less deeply sedated Procedural Sedation patients behaved more similarly to the Awake Volunteers and maintained their ability to modulate MV in response to changes in EtCO 2 . Within the Procedural Sedation group, there was not a significant difference between average supplemental oxygen FiO 2 delivered to patients with high EtCO 2 instrument sensitivity compared to patients with low EtCO 2 instrument sensitivity (p = 0.70).
In spontaneously breathing patients, an increase in partial pressure of carbon dioxide in the arterial blood (PaCO 2 ) triggers an increase in ventilation in order to maintain a relatively constant level of PaCO 2 within the physiological range. In spontanously breathing patients under the awake volunteers. Clinically-relevant EtCO 2 is indicated by the shaded gray area. The majority of General Anesthesia patients (43/54, 80%) had clinically-relevant EtCO 2 instrument sensitivity. In contrast, less than half of Procedural Sedation patients (24/58, 41%) and no patients in the Awake Volunteer cohort demonstrated clinically-relevant EtCO 2 instrument sensitivity of -76˚. anesthesia, this respiratory drive to increase ventilation in response to hypercapnia is blunted through depressed drive from both central and peripheral muscular chemoreceptors [19]. These decreases in MV results in a buildup of PaCO 2 , and in turn EtCO 2 , without a compensatory increase in ventilation.
In mechanically ventilated patients, measurements of EtCO 2 provide a clinically useful surrogate for the PaCO 2 , and capnography is the standard of care in this setting [3,4]. Recently, capnography was proposed as a tool to detect respiratory depression in non-intubated patients earlier than pulse oximetry [20][21][22][23]. However, despite the initial enthusiasm, capnography has not achieved wide clinical adoption in hospital settings such as the post-anesthesia care unit and general hospital floor [6,11]. Even in the more controlled setting of the procedure room, capnography has proved to be less reliable than anticipated due to cannula dislodgement, patient noncompliance, and complexity in interpreting CO 2 waveforms [9,11,24]. Other factors which decrease the accuracy and utility of capnography in non-intubated patients include mouth versus nose breathing, changes in flow of oxygen, procedures requiring oral intervention or lack of access to the head of the bed to ensure proper cannula placement [25]. Furthermore, EtCO 2 does not consistently reflect PaCO 2 , particularly in patients with cardiac and respiratory failure and in patients with a high ventilation-perfusion ratio [6][7][8][26][27][28][29]. Even when reliable EtCO 2 measurements are obtained, they provide a lagging indicator of respiratory performance rather than direct measure of changes in respiratory volumes [30].
Recent work has shown that capnography has poor instrument sensitivity to changes in MV in a cohort of spontaneously breathing volunteers [10]. Here we evaluated the capability of capnography to detect changes in respiratory status in a group of patients undergoing procedural sedation for a surgical procedure and demonstrated the surprisingly variable instrument sensitivity of capnography, often outside of the clinically-relevant range. The data suggest that the RVM can provide more clinically useful information than capnography during procedural sedation. These results are in-line with the findings that the RVM provides an indication of respiratory depression in advance of changes in pulse oximetry in patients following orthopedic procedures [31] and can also be used to identify and quantify respiratory depression following the administration of midazolam peri-operatively [32]. Furthermore, Holley et al. demonstrated the superiority of MV moniting over monitoring RR alone during procedural sedation for upper endoscopic procedures [33]. During these procedures, the RVM detected decreases in MV in response to sedatives and also identified increases in MV following airway maneuvers such as chin lifts and jaw thrusts [34,35]. In addition to changes in MV, the RVM is also able to detect periods of airway obstruction [12]. Combining our findings with these previous reports supports the conclusion that monitoring respiratory volumes directly in non-intubated patients under procedural sedation delivers earlier and more reliable assessment of respiratory status than capnography or pulse oximetry, providing a better alternative for use in adjusting sedation to maintain both patient safety and comfort. sensitivity (middle vertical line), the box extends from the 25th to 75th percentile, the whiskers extend to the most extreme non-outlier data points, and statistical outliers are plotted individually (plus signs). EtCO 2 values for all three groups were unimodal and a single normal distribution was fit to each group (black lines). The average EtCO 2 in the General Anesthesia cohort (37.2 ± 4.3 mmHg) was significantly higher than in the Awake Volunteers (31.4 ± 5.2 mmHg) which in turn was higher than the Procedural Sedation groups (23.3 ± 4.8 mmHg) (p < 0.0001). The normal range of EtCO 2 (35-45 mmHg) is indicated by the shaded yellow area. The majority of General Anesthesia patients (37/54, 69%) had an average EtCO 2 measurement within this normal range. In contrast, only 27% (13/48) of Awake Volunteers across the range of prescribed respiratory patterns and 5% (3/58) of Procedural Sedation patients had an average EtCO 2 measurement within the normal range. https://doi.org/10.1371/journal.pone.0180187.g003 The most challenging aspect of this study was the establishment of clinically-relevant instrument sensitivity in the Awake Volunteers control group. Whereas the General Anesthesia cohort was very similar to the Procedural Sedation cohort, the Awake Volunteers were in some ways out of place in this study. In both Procedural Sedation and General Anesthesia groups all measurements of EtCO 2 and MV were done in a purely observational manner, the Awake Volunteers had to be instructed to breathe over a range of breathing patterns in order to provide a wide measurement range over which EtCO 2 instrument sensitivity could be accurately estimated. The Awake Volunteers were also younger, had lower BMIs, and fewer comormidites than the General Anesthesia and Procedural Sedation patients. Therefore, the combination of these factors suggest that the capacity for exhalation of CO 2 is likely less in the General Anesthesia and Procedural Sedation cohorts compared to the Awake Volunteers group. Nonetheless, the addition of this cohort provided us with data from non-anesthetized patients to help bracket and better understand the group of lightly sedated Procedural Sedation patients. Interestingly, it also helped to profile the fact that in general, the Procedural Sedation group was hyperventilating, with an average MV of 148% MV PRED .
There are several other limitations to this study. First, the anesthetic and sedation regimens were chosen by the anesthesiologist and individualized for each patient. The level of sedation in the Procedural Sedation cohort was not standardized. Since individual patients respond significantly differently to similar levels of opioids and sedatives, even with a standardized sedation protocol, sedation levels still might vary greatly from patient to patient. Depth of anesthesia was not monitored with a bispectral monitor or with sedation scores as part of the standard of care. Future studies where depth of anaesthesia is also monitored are needed to test if the bimodal distribution of EtCO 2 instrument sensitivies is explained by sedation level.

Conclusions
While EtCO 2 is a useful indicator of respiratory status in patients under General Anesthesia, its sensitivity to changes in ventilation is greatly reduced in non-intubated patients. Therefore, agumenting standard patient care with EtCO 2 monitoring is a suboptimal solution for monitoring respiratory status in non-intubated patients undergoing Procedural Sedation. The addition of direct monitoring of MV with an RVM may be preferable for primary continuous assessment of adequate ventilation of non-intubated patients undergoing procedural sedation.