Clinical characterization of membrane oxygenator exhaust capnometry during venovenous extracorporeal membrane oxygenation Authors: John W Stokes, Whitney D Gannon, Elizabeth Simonds, Ioannis A Ziogas, W Kelly Wu, Yatrik J Patel, Sean A Francois, Caitlin T Demarest, Matthew Bacchetta, Rei Ukita Published: 2026 ABSTRACT AbstractCapnometry is a cost-efficient method of monitoring carbon dioxide (CO2) in healthcare settings, especially during mechanical ventilation and cardiopulmonary bypass. However, the utility of capnometers for studying CO2 elimination in a longer-duration, clinical extracorporeal membrane oxygenation (ECMO) has been poorly characterized. This report presents data from 10 venovenous ECMO patients at Vanderbilt University Medical Center between April and June 2021. Exhaust CO2 was intermittently measured from a membrane oxygenator’s gas outlet, and arterial blood samples were simultaneously drawn from the patient. Clinical variables including ECMO and ventilator management, patient demographics, and outcomes were recorded. Exhaust CO2 and other study variables were recorded at 233 discrete time points from the 10 patients. The median duration of ECMO support was 9 days. Sweep gas flow rate demonstrated a strong, negative correlation with exhaust CO2 concentration, from both univariate and multivariate regression analyses (adjusted β = − 5.556 for sweep < 4 L/min, p < 0.001). Blood flow, oxygenator duration of use, and arterial CO2 demonstrated weaker relationships with exhaust CO2 (adjusted β = 2.341, − 0.258, 0.175, respectively), which agrees with the clinical understanding about CO2 management during ECMO. This descriptive study provides foundation for future research to precisely define the clinical role of ECMO exhaust capnometry in the management of the sweep gas flow and the overall therapy.Subject terms: Diseases, Health care, Medical research, Physiology Introduction Capnometry is a cost-efficient technology for non-invasively monitoring the concentration of CO2 in gas. In health care settings, it has been used to measure CO2 concentration in exhaled gas during invasive and non-invasive ventilation applications1,2, confirmation of advanced airway placement, and during cardiopulmonary resuscitation3,4. Analogous to exhaled gas from the native lung, the gas flow leaving the gas outlet of the membrane oxygenator (exhaust gas) in extracorporeal life support (ECLS) circuits has a measurable concentration of CO25,6. Several applications for exhaust capnometry during ECLS have been proposed. Exhaust CO2 measurements inform membrane oxygenation CO2 transfer7,8 thereby providing insight into the health and efficiency of oxygenators8,9. Changes in exhaust CO2 concentration may identify circuit thrombosis10, disruptions in sweep gas flow, patient factors, and condensation accumulation within the gas exchange chamber of membrane oxygenators7. Exhaust CO2 measurements may inform sweep gas flow titration11,12 and have been applied for autoregulatory control of sweep gas flow in large animals13. Newer ECLS devices are incorporating means of monitoring membrane oxygenator exhaust CO211,14. Despite the potential uses for this non-invasive measurement, the relationship between exhaust CO2 and the important variables that may inform its clinical application are inadequately understood. Changes in exhaust CO2 concentration over the duration of membrane oxygenator use and the effect of the partial pressure of CO2 in patients’ blood have not been described in human subjects receiving extracorporeal membrane oxygenation (ECMO). Exhaust capnometry is rarely monitored during ECMO or incorporated into clinical decision-making15, and data informing ECMO applications are not available. To identify the role of exhaust CO2 monitoring in the care of a patient receiving ECMO, these relationships must first be explored. This study measured exhaust CO2 concentration from membrane oxygenators (Fig. 1) and concomitantly recorded clinically relevant variables among patients receiving venovenous ECMO (VV-ECMO). The purpose is to examine the relationship between CO2 removal, and ECMO circuit and patient variables and assess the potential utility of oxygenator capnometry for longitudinal patient and/or device monitoring. We hypothesized that CO2 removal would decrease over the duration of oxygenator use and increase with higher arterial partial pressure of CO2, holding other relevant variables constant. Based on prior work, we also expected a negative and non-linear correlation between sweep gas flow rate and exhaust CO2 concentration11,16,17. Exhaust CO2 measurement equipment and integration with ECMO circuits. Exhaust gas was sampled in a side-stream fashion via a stop-cock connecter and ¼-inch tubing. A water trap and hydrophobic filter were integrated into the sampling tubing to avoid water contamination of the CO2 gas analyzer device. Results Participant and device characteristics Ten VV-ECMO patients were enrolled in this study. The median age of study participants was 36 years old (IQR 29-47), as shown in Table 1. Nine patients (90%) received ECMO for acute respiratory distress syndrome (ARDS) secondary to COVID-19. All patients were receiving invasive mechanical ventilation at every exhaust CO2 measurement time point, and the median duration of mechanical ventilation of participants was 27 days (IQR 12–70 days). Femoral-internal jugular dual-site configuration was the initial configuration in all 10 patients; one patient was reconfigured to an internal jugular vein single-site configuration during study participation. Among the 10 patients included in the study, 16 distinct oxygenators were used since 2 of the patients required repeated device exchange. Out of the 16 oxygenators studied, 8 were (50%) Cardiohelp devices, and 8 (50%) were Nautilus oxygenators coupled with CentriMag blood pumps. Median total oxygenator duration was 10 days (IQR 7–12). Median ICU and hospital length of stay were 29 days (IQR 11–74 days) and 41 days (IQR 21–78 days), respectively. A total of 6 patients (60%) survived to both decannulation and hospitalization. Table 1Characteristics of patients who enrolled in the study and their devices.All participants (N = 10)Patient characteristics Age (years)36 (29–47) Female5 (50) Body mass index (kg2)41.8 (35.7–49.2) Body surface area (m2)2.4 (2.2–2.6) ECMO indication Acute respiratory distress syndrome10 (100) COVID-199 (90) Aspiration pneumonitis1 (10) Initial dual-site V-V ECMO10 (100) Duration of mechanical ventilation (days)27 (12–70) Duration of ECMO (days)9 (5–63) Study oxygenators per patient1 (1–1) Measurement per patient11 (8–41) ICU length of stay (days)29 (11–74) Hospital length of stay (days)41 (21–78) Survival to ECMO decannulation6 (60) Hospital survival6 (60)All Oxygenators (N = 16)Oxygenator characteristics Measurement per oxygenator11 (8–15) Oxygenator duration (days)10 (7–12) Oxygenator type Cardiohelp8 (50.0) Cardiohelp duration (days)8 (6–13) Study time point measurements108 Nautilus8 (50.0) Study time point measurements125 Nautilus duration11 (9–12)Data are presented as frequency with percentage or median with interquartile range Characteristics of patients who enrolled in the study and their devices. Data are presented as frequency with percentage or median with interquartile range Longitudinal measurements of exhaust CO2 and concomitantly recorded variables Exhaust CO2 and other study variables were recorded at 233 discrete time points across 10 patients. Figure 2 shows the trajectories of ECMO blood flow, oxygenator pressure drop, and resistance over the days of oxygenator use, and Fig. 3 shows the sweep gas, exhaust CO2, and CO2 removal rate over the same course. The median exhaust CO2 concentration was 22.4 mmHg (IQR 18.1–33.0 mmHg). Median ECMO blood flow rate and sweep gas flow rate were 4.40 LPM (IQR 3.53–5.05 LPM) and 5.0 LPM (IQR 4.0–8.0 LPM), respectively. A total of 196 measurements (84.1%) were obtained while the fraction of delivered oxygen in the sweep gas flow was 1.0. Median oxygenator duration at the time of exhaust CO2 measurement was 6.02 days (IQR 2.81–10.34 days). The median partial pressure of arterial CO2 at the time of exhaust CO2 measurement was 50 mmHg (IQR 45–61 mmHg). Data stratified by the duration of oxygenator use longer or shorter than 6 days are presented in Table 2. Between the two strata, there were no significant differences in blood flow rate, sweep gas flow rate, or the fraction of delivered oxygen. Meanwhile, the oxygenator pressure drop and resistance were significantly higher in the group with duration of 6 days or longer (p < 0.001). In addition, both exhaust CO2 (p < 0.001) and oxygenator’s CO2 removal rate (p = 0.0189) were significantly lower in the longer duration group. Mechanical ventilation settings also were significantly different between the strata: the longer duration group had higher tidal volume (p = 0.027), respiratory rate (p < 0.001), minute ventilation (p < 0.001), and fraction of inspired oxygen (p = 0.031). Longitudinal trends in patient’s ventilation parameters are shown in Fig. 4. Longitudinal measurements of blood flow, oxygenator pressure drop, and resistance over the course of oxygenator use. A vertical dashed line represents the 6-day mark used for stratification in subsequent statistical analyses. Longitudinal measurements of sweep gas flow rate, exhaust CO2, and CO2 removal rate over the course of oxygenator use. A vertical dashed line represents the 6-day mark used for stratification in subsequent statistical analyses. Table 2Exhaust carbon dioxide measurements and concomitantly recorded variable.All time points (N = 233)Oxygenator duration shorter than 6 days (N = 116)Oxygenator duration 6 days or longer (N = 117)p-value*Exhaust CO2 concentration (mmHg)22.4 (18.1–33.0)27.8 (20.1–37.2)19.8 (16.9–26.2) < 0.001Blood flow rate (LPM)4.40 (3.53–5.05)4.28 (3.66–4.95)4.47 (3.46–5.06)0.617Sweep gas flow rate (LPM)5.0 (4.0–8.0)5.0 (3.5–8.0)6.0 (4.0–8.0)0.169CO2 Elimination Rate by the Oxygenator (mL/min)175 (133–210)180 (128–200)171 (148–215)0.0189ECMO device V/Q ratio1.26 (0.82–1.76)1.22 (0.74–1.85)1.29 (0.88–1.73)0.324Fraction of delivered oxygen1.0 (1.0–1.0)1.0 (1.0–1.0)1.0 (1.0–1.0)0.252Oxygenator pressure drop (mmHg)27 (20–31)24 (19–29)29 (24–35) < 0.001Oxygenator resistance (mmHg/L/min)5.96 (5.19–6.82)5.90 (4.42–6.37)6.09 (5.45–7.32) < 0.001Patient temperature (celsius)37.1 (36.8–37.3)37.2 (36.9–37.4)37.0 (36.8–37.3)0.146Tidal volume (mL)318 (211–430)281 (211–380)360 (211–454)0.027Respiratory rate (breaths per minute)24 (20–32)22 (18–30)28 (20–33) < 0.001Minute ventilation (LPM)8.14 (4.83–11.56)6.45 (4.28–10.51)9.00 (5.80–12.60) < 0.001Fraction of inspired oxygen0.6 (0.6–0.7)0.6 (0.5–0.65)0.6 (0.6–0.7)0.031Positive end-expiratory pressure (cmH2O)10 (8–10)10 (8–12)10 (8–10)0.051Arterial blood gas results pH7.38 (7.34–7.41)7.39 (7.35–7.42)7.36 (7.32–7.40)0.001 PaCO2 (mmHg)50 (45–61)50 (44–59)51 (45–62)0.199 PaO2 (mmHg)72 (64–87)78 (65–96)71 (64–80)0.009Serum hemoglobin (gm/dL)8.0 (7.5–8.8)8.6 (7.5–9.4)7.8 (7.4–8.3) < 0.001Serum bicarbonate (mEq/L)28.4 (25.0–36.0)28.8 (26–35.8)28.0 (25.0–37.0)0.369Serum lactate (mmol/L)1.13 (0.81–1.49)1.12 (0.75–1.48)1.13 (0.98–1.49)0.121Oxygenator duration at exhaust CO2 measurement (days)6.02 (2.81–10.34)2.75 (1.06–4.0)10.34 (7.88–17.88) < 0.001ECMO day at exhaust CO2 measurement (days)12 (4–29)4 (2–17)21 (10–32) < 0.001Nautilus device measurements125 (53.7)56 (48.3)69 (59.0)0.103CO2, Carbon dioxide; LPM, Liters per minute; PaCO2, Partial pressure of arterial carbon dioxide; PaO2, Partial pressure of arterial oxygen; ECMO, Extracorporeal membrane oxygenationData are presented as frequency with percentage or median with interquartile range*Wilcoxon rank-sum Exhaust carbon dioxide measurements and concomitantly recorded variable. CO2, Carbon dioxide; LPM, Liters per minute; PaCO2, Partial pressure of arterial carbon dioxide; PaO2, Partial pressure of arterial oxygen; ECMO, Extracorporeal membrane oxygenation Data are presented as frequency with percentage or median with interquartile range *Wilcoxon rank-sum Longitudinal measurements of tidal volume, respiratory rate, and minute ventilation over the course of ECMO therapy. Single variable correlations with exhaust CO2 concentration Sweep gas flow rate demonstrated a strong, negative correlation with exhaust CO2 concentration (r = − 0.77; p < 0.001), as anticipated from prior work16,17. The unadjusted relationship between exhaust CO2 and sweep gas flow rate is presented in Fig. 5. Correlations of exhaust CO2 with duration of oxygenator use (r = − 0.44; p < 0.001), partial pressure of arterial CO2 (r = 0.38; p < 0.001), and ECMO blood flow rate (r = 0.22; p = 0.02) were weak to moderate, as shown in Table 3. Non-ECMO respiratory support variables did not correlate with concentration of exhaust CO2, except for positive end-expiratory pressure (r = 0.33; p < 0.001), as shown in Table 3. Unadjusted graphical presentation of the relationship between exhaust CO2 and sweep gas flow rate at the time of exhaust CO2 measurement. The concentration of the patient’s arterial CO2 at the time exhaust CO2 measurement is represented with color gradation. Table 3Single variable correlations with exhaust CO2 concentration.All time points (N = 233)Spearman’s coefficientp-valueExhaust CO2 concentration (mmHg)22.4 (18.1–33.0)1.00Blood flow rate (LPM)4.40 (3.53–5.05)0.220.022Sweep gas flow rate (LPM)5.0 (4.0–8.0)− 0.77 < 0.001ECMO device V/Q ratio1.26 (0.82–1.76)− 0.81 < 0.001Fraction of delivered oxygen1.0 (1.0–1.0)− 0.080.399Transmembrane pressure drop (mmHg)27 (20–31)0.060.552Transmembrane pressure drop per Liter blood flow (mmHg/L)5.96 (5.19–6.82)− 0.150.122Patient temperature (celsius)37.1 (36.8–37.3)0.090.362Tidal volume (mL)318 (211–430)0.040.668Respiratory rate (breaths per minute)24 (20–32)− 0.070.471Minute ventilation (LPM)8.14 (4.83–11.56)− 0.070.468Fraction of inspired oxygen0.6 (0.6–0.7)− 0.110.274Positive end-expiratory pressure (cmH2O)10 (8–10)0.33 < 0.001Carbon dioxide clearance (mL/min)175.0 (133.5–210.0)− 0.020.861Arterial blood gas results pH7.38 (7.34–7.41)0.180.06 PaCO2 (mmHg)50 (45–61)0.38 < 0.001 PaO2 (mmHg)72 (64–87)− 0.020.864Serum hemoglobin (gm/dL)#8.0 (7.5–8.8)0.310.001Serum bicarbonate (mEq/L)28.4 (25.0–36.0)0.48 < 0.001Serum lactate (mmol/L)1.13 (0.81–1.49)− 0.010.912Oxygenator duration at exhaust CO2 measurement (days)6.02 (2.81–10.34)− 0.44 < 0.001ECMO day at exhaust CO2 measurement (days)12 (4–29)− 0.310.001CO2, Carbon dioxide; LPM, Liters per minute; PaCO2, Partial pressure of arterial carbon dioxide; PaO2, Partial pressure of arterial oxygen; ECMO, Extracorporeal membrane oxygenation. Single variable correlations with exhaust CO2 concentration. CO2, Carbon dioxide; LPM, Liters per minute; PaCO2, Partial pressure of arterial carbon dioxide; PaO2, Partial pressure of arterial oxygen; ECMO, Extracorporeal membrane oxygenation. Multivariable regression In a multivariable, mixed-effects linear regression model, exhaust CO2 decreased as sweep gas flow increased (Table 4). The magnitude of the effect of sweep gas flow on exhaust CO2 concentration was greater when sweep gas flow was less than 4 LPM (adjusted β, − 5.556; 95% CI, − 6.886 to − 4.226; p < 0.001) relative to 4–8 LPM (adjusted β, − 1.642; 95% CI, − 2.394 to − 0.897; p < 0.001) and greater than 8 LPM (adjusted β, − 1.391; 95% CI, − 2.599 to − 0.175; p = 0.003). Exhaust CO2 concentration increased as ECMO blood flow increased (adjusted β, 2.341; 95% CI, 1.041 to 3.633; p < 0.001) as shown in Table 4. Table 4Multivariable mixed-effects linear regression examining factors that affect exhaust carbon dioxide concentration.VariableEstimate (β)95% CIp-valueSweep gas flow rate (L/min) ≤ 4− 5.556− 6.886, − 4.226 < 0.001 > 4–8− 1.642− 2.394, − 0.897 < 0.001 > 8− 1.391− 2.599, − 0.1750.003ECMO blood flow rate (L/min)2.3411.041, 3.633 < 0.001Duration of oxygenator use at the time of the observation (days)− 0.258− 0.426, − 0.0990.002Arterial CO2 (mmHg)0.1750.061, 0.2810.002CardioHelp oxygenator (reference: Nautilus oxygenator)1.756− 2.751, 6.2550.45ECMO, Extracorporeal membrane oxygenation; CO2, Carbon dioxide. Multivariable mixed-effects linear regression examining factors that affect exhaust carbon dioxide concentration. ECMO, Extracorporeal membrane oxygenation; CO2, Carbon dioxide. Additionally, holding other variables constant, both duration of oxygenator use (adjusted β − 0.258; 95% CI, − 0.426 to − 0.099; p = 0.002) and partial pressure of arterial CO2 (adjusted β, 0.175; 95% CI, 0.061 to 0.281; p = 0.002) affected measured exhaust CO2 concentration, though the size of these effects were smaller than either the sweep gas flow rate or ECMO blood flow rate. Oxygenator type was not associated with differences in exhaust CO2 concentration, when controlled for other variables (adjusted β, 1.756; 95% CI, − 2.751 to 6.255; p = 0.45). Discussion To the authors’ knowledge, this is the first study reporting longitudinal membrane oxygenator exhaust CO2 measurements and the relationships between CO2 removal and other important clinical variables in human subjects receiving VV-ECMO across days to weeks of oxygenator use. These data comprise exhaust CO2 measurements over a broad range of sweep gas flow, oxygenator duration, and partial pressure of arterial CO2, which strengthened the ability to test the effect of these variables on exhaust CO2 concentration across values observed during VV-ECMO. Correlations between exhaust CO2, sweep gas flow rate, ECMO blood flow rate, and membrane oxygenator duration agreed with physiologic principles and prior mechanistic work8,11,16,17. Exhaust CO2 significantly decreased with increasing sweep gas flow rate, which is primarily explained by the shorter residence time with increasing sweep and hence reduced accumulation of CO2 along the gas pathway18,19. The findings also agreed with prior work that demonstrated sweep gas flow rate as the most important factor determining CO2 elimination for a given oxygenator type8,20,21. Increasingly, clinicians recognize that ECMO blood flow rate plays a role in CO2 elimination, particularly at higher rates of sweep gas flow11,17,22. Carbon dioxide removal is generally considered to be a sweep-limited process, so increasing sweep is the direct straightforward way of removing more CO2. Only when the sweep gas flow reaches maximum rate does this process start to be limited by the blood flow. The multivariate analysis also demonstrated that there was a measurable effect of ECMO blood flow rate on CO2 removal. For patients that are refractory to maximum sweep gas flow, blood flow may be increased to achieve additional CO2 clearance. However, CO2 clearance is not regularly measured or documented during clinical routine ECMO, so additional clinical data are needed to further support this claim. Additionally, multivariate analysis also showed that CO2 removal rate decreased with longer duration of oxygenator use and increasing minute ventilation rate. The authors speculate that this finding may be explained by two potential causes: oxygenator may be losing efficiency of CO2 removal due to thrombus burden, and/or mechanical ventilator and membrane oxygenator are competing for available CO2. Oxygenator pressure drop and resistance were significantly higher in the longer duration group than the shorter duration (Table 2), which aligns with our theory about the clot burden reducing oxygenator efficiency. Further studying and differentiating these potential causes in future studies will be necessary before reliably using exhaust CO2 and CO2 elimination rate to inform clinical management of VV-ECMO. Prior publications on the topic of exhaust CO2 monitoring studied it in the context of cardiopulmonary bypass to estimate the partial pressure of arterial CO212,23. There were also preclinical large animal ECMO models that used exhaust CO2 readings inform and autoregulate sweep gas flow titration13. These applications rely on identifying and quantifying the independent effect of partial pressure of CO2 in blood on the measured concentration of exhaust CO2. To this end, this study provides additional clinical data to further elucidate this relationship between exhaust and blood CO2. Unadjusted and adjusted analysis here indeed showed a significant relationship between PaCO2 and exhaust CO2 (Tables 3 and 4). Previous work has also examined the potential for exhaust CO2 monitoring to identify changes in membrane function earlier, either due to slow increases in oxygenator dead space or acute oxygenator thrombosis10. Our findings potentially add to this prior work by providing exhaust CO2 data with clinical VV-ECMO data correlates. The opportunity remains to develop a more thorough understanding of these relationships, so that clinicians can better isolate the etiology of an observed change in exhaust CO2 amidst the dynamic clinical circumstances and management of ECMO patients. There are multiple limitations to this study. First, there are a few limitations related to data acquisition methods. Sweep gas flow rate was recorded from a visual readout of the sweep gas flow rotameters, which is less precise than digital measurements. Additionally, the exhaust CO2 was measured by attaching an air pump to divert exhaust gas, and this reading could get diluted or contaminated by the ambient air. More advanced methods of sweep gas flow and exhaust CO2 measurement would improve the reliability of these readouts11,24. We also did not collect information that we know to affect extracorporeal CO2 removal. We did not measure pre-oxygenator blood partial pressure of CO2 or O2 at the time of exhaust CO2 measurements, as we do not routinely maintain pre-oxygenator access for blood sampling at our institution. Without this information, it is difficult to relate CO2 removal data to CO2 concentration gradient. Currently, the institution uses noninvasive perfusion monitoring technology that can measure oxyhemoglobin saturation at device inlet and outlet, so future studies can better correlate exhaust CO2 measurements with real-time values of blood oxygenation. We also did not record end-tidal CO2 readings from the mechanical ventilators for this study. Future studies will study and compare the rates of CO2 removal happening in the native lungs versus the extracorporeal artificial lungs. We did not measure or correct for the content of CO2 in sweep gas flow entering the oxygenator, since we assumed this to be zero. Most of the presented data were obtained while the fraction of delivered oxygen to device was 100%, but we acknowledge that there may be a nonzero amount of CO2 present, especially in room air blended with oxygen in the sweep. Finally, the reported values are only steady-state measurements, so there is future opportunity to investigate the clinical utility of analyzing transient changes in exhaust CO2. Conclusion Our findings suggest there may be a role for non-invasive monitoring of exhaust CO2 during VV-ECMO and other forms of ECLS. Significant relationships between sweep gas, blood flow, duration of use, minute ventilation, and CO2 removal rate were observed. These data lay a framework for additional investigation to clarify the multiple factors that affect measurements of exhaust CO2 to derive clinical utility of this cost-efficient, non-invasive measurement. Accurate exhaust CO2 and sweep gas flow sensors, combined with strict protocols for consistently obtaining exhaust CO2 measurements, are necessary to maximize precision of measurements if exhaust CO2 is to enhance clinical decision-making. As several variables affect exhaust CO2 concentration and thus CO2 clearance, exhaust CO2 measurements alone are unlikely to inform clinical diagnosis of the oxygenator efficiency or to guide sweep gas flow titration. Thus, monitoring trends over time could offer additional utility, but further study is needed to generate the data necessary to interpret longitudinal trends. Additionally, there may be potential for decision support algorithms and software that rapidly integrate multiple variables to aid clinical interpretation of isolated measurements or continuous trends of exhaust CO2. Robust data from larger patient cohorts are necessary to develop such systems. Methods Study design A single-center prospective cohort study was conducted among adult patients receiving VV-ECMO at Vanderbilt University Medical Center between April 1, 2021, and June 15, 2021. This study was approved by the Vanderbilt University Medical Center’s Human Research Protections Program and its Institutional Review Board (IRB, Protocol Number 210226). Given the impact of ECMO on the patient’s cognitive function, the study team obtained informed written consent from an appropriate surrogate for study participation. All methods were performed in accordance with the relevant federal and institutional guidelines and regulations. Location and participants All adults (18 years of age or older) receiving VV-ECMO at Vanderbilt University Medical center during the study period were eligible. Exclusion criteria were patients with more than one membrane oxygenator incorporated into their ECMO circuit and presence of an arterial reinfusion cannula. Eligible patients were enrolled immediately upon completion of informed consent at ECMO initiation. Patients whose ECMO support was initiated at a hospital outside of Vanderbilt were enrolled upon admission to an ICU at our institution. ECMO circuit equipment and management Initiation of VV-ECMO reflected Extracorporeal Membrane Oxygenation for Severe Acute Respiratory Distress (EOLIA) trial inclusion criteria25. All patients were supported with either a Cardiohelp HLS set advanced 7.0 (Getinge, Gothenburg, Sweden) circuit and oxygenator or a Nautilus oxygenator (Medtronic, Dublin, Ireland) integrated within a CentriMag (Abbott Laboratories, Chicago, IL) blood pump and circuit. Both the Cardiohelp HLS set advanced 7.0 and Nautilus oxygenators employed a polymethylpentene membrane with equivalent gas exchange surface areas of 1.8 square meters. Clinical decisions, including indications and timing of circuit exchanges, timing of ECMO decannulation, ECMO support settings, and non-ECMO respiratory support settings were determined by the clinical teams and not influenced by study interventions or measured data. Study interventions and data collection The study investigators intermittently measured exhaust CO2 concentration from the membrane oxygenators of study participants. Time points for exhaust CO2 measurements were selected to reflect steady-state conditions (i.e. measurements were not obtained immediately surrounding changes to ECMO or non-ECMO respiratory support settings). All exhaust CO2 measurements were obtained using a K33 ICB 10% CO2 gas analyzer kit (CO2METER.com, Ormond Beach, FL), which included a miniature air pump for gas sampling and CO2 gas analyzer rated up to 10% concentration. The device used for exhaust CO2 measurements is not approved by the Food and Drug Administration for clinical use. As such, it was clarified during the IRB approval process that measurements obtained during the study were not shared with the clinical ECMO team or used to influence clinical decision-making. The CO2 meter was attached to the oxygenators of study participants using ¼-inch tubing in a side-stream fashion from the exhaust port of the oxygenator, with additional tubing in-line with exhaust port to prevent room air contamination of measurements26, as shown in Fig. 1. Prior to measurements, excess water in the gas chamber was purged, by briefly increasing sweep gas flow to maximum for 1 min, bringing back down to original value, and waiting until the real-time exhaust CO2 reading reaches a steady state. Real-time CO2 concentration readings from the gas analyzer were displayed using Gas Lab software (CO2METER.com, Ormond Beach, FL), and steady-state time point measures of CO2 concentration were recorded in Research Electronic Data Capture (REDCap). In conjunction with exhaust CO2 concentration measurements, the following variables were also simultaneously collected: ECMO day, oxygenator duration, ECMO settings (blood flow rate, sweep gas flow rate, fraction of delivered oxygen, and transmembrane pressure), mechanical ventilator settings (tidal volume, respiratory rate, fraction of inspired oxygen, positive end-expiratory pressure), arterial blood gas values, serum bicarbonate, hemoglobin concentration, and patient temperature. Demographics, baseline characteristics, and outcomes were extracted from the electronic medical record. Data collections were conducted through REDCap. Statistical analysis Baseline characteristics and timepoint variables were first examined with descriptive statistics. Continuous variables were expressed as median and interquartile range and categorical variables were expressed as frequency and percentage. Exhaust CO2 measurements and associated variables collected at the time of data collection were stratified by less than or greater to 6 days of oxygenator use and by oxygenator type. The Wilcoxon rank-sum test was used to examine between-group differences in continuous variables. Linearity between exhaust CO2 and candidate effect variables was examined by scatter plot and by Shapiro–Wilk and Cameron-Trivedi decomposition tests for normality and homogeneity of residuals of simple linear regression plots. Correlation between collected variables and exhaust CO2 concentration were examined using Spearman’s correlation coefficient given predominance of non-linear or ordinal data. Then, multivariable mixed-effects linear regression was performed, which specified patient and oxygenator as random-effects, and exhaust CO2 as the outcome. Variables that were considered clinically meaningful and correlated with exhaust CO2 concentration in the single variable correlation analysis were selected as fixed-effect variables: sweep gas flow rate, blood flow rate, partial pressure of arterial CO2, and duration of oxygenator use at the time of exhaust CO2 measurement. We used linear splines with 2 knots to account for the nonlinear relationship between sweep gas flow and exhaust CO2, which was expected based on prior work11,16 and confirmed by previously described assessments for linearity. Analyses were performed using STATA 14.2 (College Station, TX). Abbreviations Acute respiratory distress syndrome Venovenous extracorporeal membrane oxygenation Coronavirus disease 2019 Carbon dioxide Intensive care unit Institutional review board Extracorporeal membrane oxygenation for severe acute respiratory distress syndrome Author contributions JS, WG, ES, YP, SF collected data JS, WG, IZ, WKW, CD, MB, RU analyzed data JS, WG, RU prepared the figures JS, WG, MB, RU wrote the manuscript All authors reviewed and approved the manuscript. Funding This work was supported by an Extracorporeal Life Support Organization (ELSO) Research Award given to John W Stokes. Data availability Data are available upon request. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement Data are available upon request.