Venoarterial extracorporeal membrane oxygenation (VA ECMO) is a life-saving care modality for bridging patients with severe cardiopulmonary failure to definitive therapy or illness resolution.1 However, because of a combination of patient disease and possible complications of VA ECMO, prolonged exposure to this care modality is frequently accompanied by end-organ complications. For example, one study estimated the incidence of neurologic complications after extracorporeal membrane oxygenation (ECMO) to be 13.3%.2 Gaseous microemboli (GME) are postulated to play a role in end-organ damage and neurologic dysfunction after cardiopulmonary bypass (CPB),3,4 and literature describing and quantifying GME burdens predominantly involves CPB,3–11 while only one study provides evidence of GME during VA ECMO.12 To the extent that GME may occur during VA ECMO, the lack of an arterial filter,5 pre-existing disease burden, and long periods of extracorporeal circulation may make them particularly deleterious in this critically ill patient population. We hypothesized that a largely unrecognized GME burden exists during VA ECMO owing to iatrogenic introduction of venous air during routine patient care and the lack of arterial filtration on ECMO circuits.
Therefore, we performed a pilot observational clinical study of adult VA ECMO patients to characterize GME burdens upstream and downstream of the ECMO membrane oxygenator. We used noninvasive Doppler probes, as required by our institutional review board (IRB), and performed in vitro experiments to correlate Doppler signals with definitive Emboli Detection and Classification (EDAC) quantifier (Terumo Cardiovascular, Ann Arbor, MI) measurement of bubble counts and sizes.13
In Vivo Extracorporeal Membrane Oxygenation Gaseous Microemboli Measurement Methods
Noninvasive measurement of GME in VA ECMO circuits was approved by the Human Investigations Committee of Yale University, including a waiver of consent. While we would have preferred to use EDAC to measure GME in vivo, this would have necessitated off-label use of EDAC cuvettes for greater than their 6 hour Food and Drug Administration (FDA) approval, an option that was considered inappropriate in the context of this research study. Therefore, we used two Doppler probes (Medasonics Versatone D8, 2.4 MHz 3.3 cm probe; Cooper Surgical, Trumbull, CT) to measure GME activity noninvasively in pre- and postmembrane locations of VA ECMO circuits. The devices were fixed to blood tubing via custom-made polycarbonate braces (Figure 1A). The Doppler signal was processed by a custom analog envelope detector then digitized and recorded electronically (60 Hz sample rate) using Windaq (DATAQ Instruments, Akron, OH). Circuit elements are shown in Figure 1B.
A total of 15 hours of dual-site Doppler recordings were collected during routine intensive care unit (ICU) care of four adult VA ECMO patients (Table 1). A dedicated human observer was present during data collection to record clinical interventions and ECMO parameters. Embolic activity was categorized based on temporal association—either during or within 60 seconds of 1) background, 2) arterial, 3) movement, 4) intravenous (IV) injection, or 5) IV infusion as follows. “Background” was composed of periods of no identifiable clinical intervention. “Arterial” involved blood draws and flushing of the arterial line. “Movement” included moving or examining the patient, or movement of any extracorporeal circuit or IV element without direct exposure to air. “IV injection” included flushes of fluids and medications and connections/disconnections of equipment that directly exposed the IV apparatus to air. Although patients were continually exposed to IV infusions, infusions of red blood cells (RBCs), platelets, and vancomycin produced ongoing embolic showers differing from background and were thus described in a separate category termed “IV infusion.”
In vivo Doppler data were visually examined to exclude obvious signal artifact. Systematic signal analysis of the Doppler waveform was performed using Python 3.2 (Python Software Foundation, Wilmington, DE). Doppler peaks were counted at any local maxima (defined as a data point higher than the single points preceding and following) in the Doppler signal that were at least six standard deviations above a nearby GME-free baseline region to differentiate GME signals from noise. A combination of automated peak detection and visual inspection of the record was used to divide the Doppler signal into contiguous showers. Numbers of pre- and postmembrane peaks were compared using a paired, two-tailed Student’s t-test. Shower sizes (number of postmembrane Doppler peaks) associated with the clinical intervention categories were compared using a nonparametric Analysis of Variance (ANOVA) (Kruskal–Wallis). Post hoc comparison of each category against background was performed with a nonparametric multiple comparison (Dunn) test (Prism 6; Graphpad, La Jolla, CA).
In Vitro Doppler Calibration Experiment Methods
Because of the fact that the Doppler voltage signals provide only semi-quantitative information about GME burden, we calibrated them in vitro to the gold-standard EDAC to facilitate estimation of the numbers and volumes of GME experienced by the study patients. An ECMO circuit was filled with heparinized banked human RBCs and fresh-frozen plasma (discarded product, hematocrit = 33.9–35.6%, Hartford Hospital Transfusion Services). Patient metabolism was simulated using a Capiox RX-25 oxygenator (Terumo Cardiovascular, Ann Arbor, MI) supplied with hypobaric carbon dioxide sweep gas at 0.1 atmospheres absolute (ata) as we have previously described.13 Circuit elements are shown in Figure 1C. To simulate IV injections witnessed during patient care, small quantities (1–10 cc) of saline and blood were injected into the blood path upstream of the premembrane recording devices via the 16 g distal port of a 16 cm triple lumen catheter (TLC; Arrow International, Reading, PA). The technique and rate of injection were varied in the attempt to mimic the range of routine patient care interventions witnessed in the ICU setting.
EDAC emboli counts and volumes were obtained directly from the EDAC software (version 2.5.1; Terumo Cardiovascular, Ann Arbor, MI) using 10 μm bin sizes in 1 second time intervals. Data from the two independent modalities of GME measurement were correlated using linear and logarithmic least-squares regression (Prism 6). Microembolic counts and volumes represented by Doppler signals recorded in vivo were estimated using the regression equations derived from the in vitro data.
Fifteen hours of pre- and postmembrane Doppler recordings from VA ECMO circuits in four patients showed characteristic signals consistent with gas embolization (Figure 2).13 Doppler peaks in the premembrane trace preceded Doppler peaks in the postmembrane trace by 3–4 seconds, presumably representing the time required for a traveling embolus to traverse the ECMO oxygenator. Premembrane Doppler peaks led to larger numbers of postmembrane peaks (414% of premembrane values; p < 0.0001), consistent with larger emboli breaking into smaller emboli as they traverse the oxygenator.14 We have previously shown that Doppler peaks of this nature are efficiently eliminated by the use of hypobaric oxygenation in extracorporeal circuits, confirming their identity as GME.13
Five hundred three total GME showers were observed during the study (Table 2). We observed 309 showers (8% of total postmembrane Doppler peaks) during background periods with no observed clinical intervention, no showers during arterial interventions, 36 showers (11% of peaks) during patient movement, 22 showers (37% of peaks) during IV injection, and 136 showers (44% of peaks) during IV infusion. Compared with showers during background periods (median: 1; range: 0–285), the number of postmembrane peaks per shower was much larger during clinical interventions involving patient movement (median: 4; range: 0–708), IV injection (median: 300; range: 1–1,616), and IV infusion (median: 37; range: 0–1,107) (all p < 0.0001 for comparison with background; Figure 3).
Recognizing that Doppler probes outside the circuit tubing lack sensitivity and ability to discriminate high embolic rates, we sought to understand the magnitude of in vivo GME activity by comparing Doppler to the gold-standard EDAC quantifier in an in vitro ECMO circuit (Figure 1C). When the circuit was challenged with an array of injections meant to simulate patient care activities, the EDAC detected GME with close temporal correlation to the Doppler in both pre- and postmembrane locations (Figure 4A). The EDAC size distribution of GME observed in the unfiltered ECMO circuit appeared larger than that reported from EDAC monitoring of filtered CPB circuits in vivo, as expected (Figure 4B).15 Four regression equations were fit to describe the relationship between Doppler signals and EDAC bubble counts as follows: During modest GME rates where the Doppler signal was not saturated, the relationship between Doppler and EDAC peak counts was highly linear at both the pre- (r2 = 0.81) and postmembrane (r2 = 0.97) locations (Figure 5, A and C). In events with stronger GME signals during which the Doppler signal was saturated at the upper extent of the measurable voltage range, the relationship between the Doppler signal and the EDAC signal in the premembrane data remained strongly linear (r2 = 0.81; Figure 5B). In the postmembrane data, the saturated Doppler signal was most accurately correlated to the EDAC data by a logarithmic equation (r2 = 0.92; Figure 5D). Four similar regression equations described the relationship between Doppler peaks and EDAC GME volumes.
Based on the regression equations, GME bubble counts and volumes for each embolic shower recorded during the in vivo sessions were estimated. Across 15 hours of recording, approximately 359,000 estimated GME were returned toward the patient’s arterial circulation from the VA ECMO circuit, accounting for an estimated 60 μl of volume (Table 2). Three percent of estimated GME accounting for 1% of estimated embolized volume occurred during the 70% of time considered “background” because of the absence of clinical intervention (Figure 6). In contrast, 68% of estimated GME and 88% of estimated volume occurred during the 4% of time attributed to IV injection, whereas 6% of GME and 3% of volume occurred during patient movement and 23% of GME and 8% of volume occurred during IV infusion.
Here, we demonstrate the presence of a substantial embolic burden experienced during the routine clinical care of VA ECMO patients. Previously, ECMO circuits were thought to be relatively immune from GME because the circuit lacks an open reservoir and has no surgical blood inputs. However, iatrogenic introduction of venous air during routine clinical care provides a potential source of air to ECMO circuits. One study has used transcranial Doppler to document the occurrence of arterial GME during ECMO but did not characterize how frequently they occur.12 Our data support the conclusions that 1) showers of GME are common during routine clinical care of patients receiving VA ECMO and 2) clinical interventions, specifically iatrogenic venous air introduction, trigger a majority of embolic events.
When we used an in vitro calibration experiment with EDAC in an effort to understand the magnitude of embolization represented by the in vivo Doppler recordings, we were careful to use the same ECMO circuit design and pump settings, human blood, and an array of clinically realistic interventions for testing. We believe the nonlinear correlation between saturated Doppler signals and EDAC measurements at the postmembrane location exists because the true embolic rate, and thus the difference between the true rate and the maximum 30/s Doppler peak rate, rises as the length of the saturated signal increases at this location. Despite the tight relationships between the Doppler and gold-standard EDAC data that we observed in vitro, the estimated GME counts and volumes that we report are only estimates and certainly contain some inaccuracy. Nonetheless, the estimated average embolic rate of 24,000 emboli per hour and 4 μl/hr in this study suggests that hourly GME rates during ECMO may rival embolic rates recorded using EDAC during CPB for coronary artery bypass and valve procedures.15
The potentially harmful effects of GME and their common occurrence during VA ECMO raise the question whether measures to limit GME may improve end-organ outcomes in this vulnerable patient population. Although this study was designed as a pilot analysis to describe the quantity and timing of GME experienced by VA ECMO patients during routine ICU care, renal and neurologic outcomes of the study patients are listed in Table 1 for the sake of descriptive completeness. The possible association of the listed outcomes in these study patients with quantities of observed GME described above is not meant to be implied. Although GME have been observed in other vulnerable populations including pediatric patients undergoing CPB, the quantity and size that lead to significant end-organ dysfunction in the context of modern elimination strategies remain a topic of continued research.16 The addition of arterial filters to CPB circuits in the 1990s was accompanied by fewer emboli measured by transcranial Doppler, fewer embolic insults in the brain, and fewer clinically evident neurologic complications. Indeed, the use of arterial filters in the context of CPB has become standard practice in the field.5,17,18 The use of 0.2 μm pore size bubble filters at venous access points has been suggested, but this would only represent a partial solution to GME observed in our study given their lack of utility during blood draws and transfusions of blood products.12 The addition of arterial filtration to the ECMO circuit, which is routinely done in CPB, using filters as small as 20 microns,19 would limit GME delivery but may complicate management of anticoagulation. The potential future use of hypobaric oxygenation may limit GME without impacting anticoagulant management if it proves compatible with diffusion membrane oxygenators.13 Future studies would be needed to define the effectiveness of these interventions in improving outcomes among this most fragile of patient populations.
The authors thank Hartford Hospital Transfusion Services for donating blood products.
1. Stretch R, Sauer CM, Yuh DD, Bonde P. National trends in the utilization of short-term mechanical circulatory support: Incidence, outcomes, and cost analysis. J Am Coll Cardiol 2014.64: 1407–1415.
2. Cheng R, Hachamovitch R, Kittleson M, et al. Complications of extracorporeal membrane oxygenation
for treatment of cardiogenic shock and cardiac arrest: A meta-analysis of 1,866 adult patients. Ann Thorac Surg 2014.97: 610–616.
3. Gallagher EG, Pearson DT. Ultrasonic identification of sources of gaseous microemboli
during open heart surgery. Thorax 1973.28: 295–305.
4. Grocott HP, Homi HM, Puskas F. Cognitive dysfunction after cardiac surgery: Revisiting etiology. Semin Cardiothorac Vasc Anesth 2005.9: 123–129.
5. Pugsley W, Klinger L, Paschalis C, Treasure T, Harrison M, Newman S. The impact of microemboli
during cardiopulmonary bypass on neuropsychological functioning. Stroke 1994.25: 1393–1399.
6. Baker AJ, Naser B, Benaroia M, Mazer CD. Cerebral microemboli
during coronary artery bypass using different cardioplegia techniques. Ann Thorac Surg 1995.59: 1187–1191.
7. Clark RE, Brillman J, Davis DA, Lovell MR, Price TR, Magovern GJ. Microemboli
during coronary artery bypass grafting. Genesis and effect on outcome. J Thorac Cardiovasc Surg 109: 249–257, 1995; discussion 257.
8. Feerick AE, Church JA, Zwischenberger J, Conti V, Johnston WE. Systemic gaseous microemboli
during left atrial catheterization: A common occurrence? J Cardiothorac Vasc Anesth 1995.9: 395–398.
9. Jones TJ, Deal DD, Vernon JC, Blackburn N, Stump DA. Does vacuum-assisted venous drainage increase gaseous microemboli
during cardiopulmonary bypass? Ann Thorac Surg 2002.74: 2132–2137.
10. Padayachee TS, Parsons S, Theobold R, Linley J, Gosling RG, Deverall PB. The detection of microemboli
in the middle cerebral artery during cardiopulmonary bypass: A transcranial Doppler ultrasound investigation using membrane and bubble oxygenators. Ann Thorac Surg 1987.44: 298–302.
11. Sylivris S, Levi C, Matalanis G, et al. Pattern and significance of cerebral microemboli
during coronary artery bypass grafting. Ann Thorac Surg 1998.66: 1674–1678.
12. Zanatta P, Forti A, Bosco E, et al. Microembolic signals and strategy to prevent gas embolism during extracorporeal membrane oxygenation
. J Cardiothorac Surg 2010.5: 5.
13. Gipson KE, Rosinski DJ, Schonberger RB, et al. Elimination of gaseous microemboli
from cardiopulmonary bypass using hypobaric oxygenation. Ann Thorac Surg 2014.97: 879–886.
14. Gill MC, Dando H, John D. Is the air handling capability of the Quadrox D pump dependent within an ECMO circuit? An in vitro
study. J Extra Corpor Technol 2010.42: 203–211.
15. Lynch JE, Wells C, Akers T, et al. Monitoring microemboli
during cardiopulmonary bypass with the EDAC quantifier. J Extra Corpor Technol 2010.42: 212–218.
16. Matte GS, Connor KR, Liu H, DiNardo JA, Faraoni D, Pigula F. Arterial limb microemboli
during cardiopulmonary bypass: Observations from a congenital cardiac surgery practice. J Extra Corpor Technol 2016.48: 5–10.
17. Murphy GS, Hessel EA 2nd, Groom RC. Optimal perfusion during cardiopulmonary bypass: An evidence-based approach. Anesth Analg 2009.108: 1394–1417.
18. Padayachee TS, Parsons S, Theobold R, Gosling RG, Deverall PB. The effect of arterial filtration on reduction of gaseous microemboli
in the middle cerebral artery during cardiopulmonary bypass. Ann Thorac Surg 1988.45: 647–649.
19. Jabur GN, Willcox TW, Zahidani SH, Sidhu K, Mitchell SJ. Reduced embolic load during clinical cardiopulmonary bypass using a 20 micron arterial filter. Perfusion 2014.29: 219–225.
Keywords:Copyright © 2018 by the American Society for Artificial Internal Organs
extracorporeal membrane oxygenation; microemboli