Month-long Respiratory Support by a Wearable Pumping Artificial Lung in an Ovine Model

Abstract

Background:

A wearable artificial lung could improve lung transplantation outcomes by easing implementation of physical rehabilitation during long-term pretransplant respiratory support. The Modular Extracorporeal Lung Assist System (ModELAS) is a compact pumping artificial lung currently under development. This study evaluated the long-term in vivo performance of the ModELAS during venovenous support in awake sheep. Feedback from early trials and computational fluid dynamic (CFD) analysis guided device design optimization along the way.

Methods:

The ModELAS was connected to healthy sheep via a dual-lumen cannula in the jugular vein. Sheep were housed in a fixed-tether pen while wearing the device in a holster during support. Targeted blood flow rate and support duration were 2-2.5 L/min and 28-30 days, respectively. Anticoagulation was maintained via systemic heparin. Device pumping and gas exchange performance and hematologic indicators of sheep physiology were measured throughout support.

Results:

CFD-guided design modifications successfully decreased pump thrombogenicity from initial designs. For the optimized design, 4 of 5 trials advancing past early perioperative and cannula-related complications lasted the full month of support. Blood flow rate and CO₂ removal in these trials were 2.1 ± 0.3 L/min and 139 ± 15 mL/min, respectively, and were stable during support. One trial ended after 22 days of support due to intradevice thrombosis. Support was well tolerated by the sheep with no signs of hemolysis or device-related organ impairment.

Conclusions:

These results demonstrate the ability of the ModELAS to provide safe month-long support without consistent deterioration of pumping or gas exchange capabilities.

INTRODUCTION

Lung disease is the fourth-leading cause of death in the United States and remains a primary health concern.¹ Respiratory support via mechanical ventilation (MV) or extracorporeal membrane oxygenation (ECMO) is typically required during severe respiratory failure as a bridge to recovery or lung transplantation. Although these forms of support are sufficient for short-term use, neither are ideal for longer-term use in an awake patient. Prolonged use of MV can exacerbate lung injury via barotrauma and volutrauma.^2-4 Accordingly, the use of MV as a bridge to lung transplantation is associated with negative outcomes.^5,6

ECMO is an attractive alternative to MV due to its ability to provide respiratory support independent of the lungs and without intubation. Outcomes associated with pretransplant support via ECMO relative to MV vary across reports but have generally shown similar or, more frequently, improved survival for ECMO.^7-10 The successful implementation of long-term ECMO, however, is challenging. Conventional ECMO circuits composed of multiple components connected via tubing are cumbersome and overly complex. Such systems typically render a patient bedridden, thereby resulting in progressive atrophy that can negatively impact long-term patient outcomes.^11,12 More recently, the introduction of portable ECMO systems and improved dual-lumen cannulae has reduced the burden of ambulating patients during support. Retrospective analysis has shown that the implementation of physical rehabilitation during pretransplant ECMO results in improved patient outcomes.^13-15 Although newer ECMO systems are compact enough to be portable, ambulation during ECMO is still viewed as challenging. Despite the evidence for its benefits, ambulatory ECMO is only performed at 22-34% of centers.^16,17 Further integration and size reduction of system components may enable a wearable device that eases physical rehabilitation during support thereby improving patient outcomes.

Our group is developing the Modular Extracorporeal Lung Assist System (ModELAS) as a wearable artificial lung that can be used for long-term support (target of 1-3 months) and in multiple respiratory assist applications. The ModELAS integrates a custom centrifugal blood pump with a highly efficient hollow fiber membrane (HFM) gas exchanger design. The pump can be configured with HFM bundles of varying sizes to accommodate respiratory support in adult or pediatric patients as well as low-flow extracorporeal CO₂ removal (ECCO₂R). The bundle design was optimized for high gas exchange efficiency to minimize the surface area required for therapeutic gas transfer.¹⁸ This minimizes blood-contacting surface area that may initiate thrombosis and produces a compact overall device form factor. The integrated centrifugal pump was designed to generate a large pressure head for compatibility with high-resistance, dual-lumen cannulae. Prior in vitro results have demonstrated the ability of the ModELAS design to achieve targeted blood flow and gas exchange rates while maintaining low hemolysis under conditions anticipated for adult, pediatric and ECCO₂R respiratory support applications.^19-21 Additionally, successful short-term sheep studies assessing performance in the setting of pediatric and adult support have also been reported.^22,23

This study evaluated the performance of the ModELAS in the adult support configuration during 1 month of continuous support in a healthy sheep model. The primary goal was to assess the capability of the device to maintain consistent performance during long-term support. Additionally, potential effects of device support on normal physiology and end organ function were assessed. Multiple device design iterations were evaluated. Feedback from initial 1-month trials was used along with computational fluid dynamic (CFD) modeling to iteratively improve device design.

MATERIALS AND METHODS

Device Design

A schematic of the ModELAS is shown in Figure 1a. The cylindrical HFM bundle (0.65 m² surface area) was fabricated from uncoated polymethylpentene fibers (OXYPLUS, 3M, Saint Paul, MN) as previously described.¹⁸ The integrated pump utilizes a closed impeller design with embedded permanent magnets that couple to an external motor. The impeller uses a simple contact bearing system with ball pivots supported by ultra-high molecular weight polyethylene cups. Bearing design evolved based on early in vivo trials within this work, resulting in 3 different designs (Figure 1b) being used during in vivo trials. Details regarding bearing design progression and CFD methods are provided in the SDC, Materials and Methods. Functional ModELAS prototypes (Figure 1c) were machined from acrylic and polycarbonate. The current prototype weighs approximately 1.8 lb and has a priming volume of 135 mL. Device weight is expected to be substantially reduced upon transitioning to injection molding for a finalized design.

Figure 1:

(a) Schematic of ModELAS showing the blood flow path in red arrows. (b) Schematics showing cross-sectional views of the impeller bearing designs. (c) Photograph of functional ModELAS prototype used during in vivo trials.

Surgical Procedure

In vivo device evaluation was performed in adult sheep (40-80 kg) at the McGowan Institute Center for Preclinical Studies. All animals received humane care in accordance with the Guide for Care and Use of Laboratory Animals. The surgical and animal care protocol was approved by the Institutional Animal Care and Use Committee of the University of Pittsburgh. Sheep were induced with subcutaneous atropine (0.22 mg/kg) and butorphonal (0.25 mg/kg) followed by intravenous (IV) ketamine (4 mg/kg) and midazolam (0.4 mg/kg). Anesthesia during surgery was maintained with isoflurane inhalation (1.5-3%). Prior to cannulation, an activated clotting time (ACT) of 300 seconds was targeted via a 180 IU/kg heparin bolus. A 27 Fr Avalon Elite Bi-Caval Dual Lumen Catheter (DLC) (Maquet, Rastatt, Germany) was placed in the right external jugular with the outflow hole at the level of the tricuspid valve. Positioning of the cannula was confirmed via fluoroscopy before connecting it to a primed ModELAS (4 IU/mL heparinized saline). Due to DLC fractures in early trials, the exterior portion of the DLC was reinforced with a custom molded silicone sleeve. The DLC was secured at the vessel via the purse string technique and externally sutured to the neck around the silicone reinforcement. A photograph of the cannulation site and fluoroscopic images showing cannula positioning are provided in Figure S1.

Animal Management

Recovered sheep were housed in a fixed-tether pen with free access to food and water while wearing the device in a holster. Sheep received prophylactic antibiotics (Baytril, 5 mg/kg) daily, an analgesic (Banamine, 1 mg/kg) 2-3 times per day for the first 3 postoperative days, and an appetite stimulant (Reglan, 10 mg) as needed. Continuous heparin infusion was used to maintain a targeted ACT of 1.5-2 times baseline values. Following recovery from surgery, pump speed was incrementally increased from 1200 to approximately 1400 revolutions/min (corresponding to a blood flow rate of ~2-2.5 L/min) over the first 48 hours while monitoring for intravascular suction. Prior short-term studies showed this to be the highest flow rate achievable in this experimental model without frequent suction complications.²² Pump speed was lowered during support if flow limitations were repeatedly observed. Sweep gas (100% O₂) flow was primarily maintained at twice the blood flow rate. End of study criteria were defined as 28-30 days of support or a decrease in bundle permeability greater than or equal to 50%. At trial conclusion, explanted devices were passively rinsed with saline prior to being photographed and removing HFM samples for scanning electron microscopy. A necropsy was performed to inspect the heart, lungs, kidneys, liver and spleen and to collect samples for histopathology.

Data Collection and Analysis

Blood flow rate and motor torque were continuously recorded and averaged over daily intervals. Motor torque is dependent on pump speed, and thus a normalized torque change is presented to account for manual pump speed changes during support. The normalized value is the percent change in torque relative to that measured during the initial 24 hours at a given pump speed. Gas exchange rates were measured 2-3 times per week and calculated based on blood and sweep gas measurements according to previously described methods.^20,21 Sweep gas flow rate was increased to 15 L/min during measurement of CO₂ removal to ensure it was not limited by gas flow rate. A normalized CO₂ removal rate is presented to account for varying blood flow rates between measurements. The normalized value is equal to the measured value divided by the flow rate-specific CO₂ removal rate acquired from benchtop studies conducted over the entire flow rate range. The effective Darcy permeability (k) of the HFM bundle was assessed 2-3 times per week as a measure of intrabundle flow occlusion. Bundle permeability is inversely proportional to flow resistance and takes into account changes in blood viscosity. Further details regarding benchtop CO₂ removal measurements used for normalization and permeability calculations are described within the SDC, Materials and Methods. Device performance results for completed trials were aggregated whereas results for trials that ended prematurely are discussed separately.

Blood samples were collected prior to the trial (i.e., at baseline), intraoperatively, throughout support, and at trial conclusion. Plasma free hemoglobin (pfHb) and circulating platelet activation (indicated by surface P-selectin expression) were measured using previously described methods.^24-26 Platelet function was assessed by measuring P-selectin expression of platelets stimulated with platelet activating factor (PAF). Soluble P-selectin was quantified using an immunosorbent assay (MBS737302, MyBioSource, San Diego, CA). A comprehensive metabolic panel, coagulation profile, manual platelet count, and complete blood count were completed by an outside laboratory (IDEXX Laboratories, INC., Westbrook, ME). Platelet count for Trial 3.7 on postoperative day (POD) 21-25 could not be measured due to an error in sample collection. Platelet P-selectin expression measurement for Trial 3.5 on POD 21-25 was not feasible due to the timeframe of the premature trial termination.

Statistical analysis was performed using SPSS 26 (IBM Corporation, Armonk, NY). A previously described mixed linear model analysis of variance with the restricted maximum likelihood estimate method was used due to the instances of missing data.²⁷ Data that violated the assumption of normality were normalized prior to statistical analysis using a previously described approach.²⁸ Appropriate pairwise comparisons were made using least significant difference analysis. Subsequently reported values are given as mean ± standard deviation unless otherwise specified.

RESULTS

Study Summary and Design Evolution

Study outcomes from 19 trials performed across the 3 different ModELAS designs are summarized in Table 1. Consistent bearing thrombosis during design 1 trials led to substantial increases in motor torque (>50%) and hemolysis (pfHb > 50 mg/dL) 2-3 weeks into the support period and motivated bearing design modification. Figure 2 shows CFD-predicted shear rates and velocity magnitudes within the blood in the regions near the bearings for all designs. Modifications introduced for design 2 increased modeled shear rates at the lower bearing by 2 orders of magnitude and eliminated the local regions of flow stasis shown for design 1. These modifications successfully reduced adherent lower bearing thrombosis and eliminated the occurrence of associated hemolysis. Transient minor increases in motor torque (<20%), however, were still observed 2-3 weeks into support and coincided with observations of bearing thrombosis. Upper bearing ring thrombus was consistently observed in the region of modeled flow separation shown in Figure 2 (indicated by dotted circle). Inverting the pivot-cup configuration in design 3 eliminated flow separation and improved surface washing. A similar inversion at the lower bearing along with changes to the flushing port geometry increased local shear rates. In summary, CFD results indicate decreased blood stagnation and improved surface washing in the regions near the impeller bearings due to the implementation of the described design modifications.

Table 1:

Summary of In Vivo Trial Outcomes

Device Design	Trial No.	Duration [days]	Termination Summary	Significant Observations
1	1.1	30	Study Completion	Thrombosis at lower and upper bearing
1	1.2	28	Study Completion	Thrombosis at lower and upper bearing, significant hemolysis
1	1.3	8	Fractured Cannula	-
1	1.4	27	Motor Failure	Thrombosis at lower and upper bearing, significant hemolysis
2	2.1	2	Animal Welfare	Uncontrollable bleeding at arterial line
2	2.2a	30	Study Completion	-
2	2.3a	23	Decreasing Bundle Permeability	Thrombosis at lower and upper bearing
2	2.4	30	Study Completion	-
2	2.5a	30	Study Completion	Thrombosis at upper bearing
2	2.6	30	Study Completion	Thrombosis at lower and upper bearing
2	2.7a	3	Fractured Cannula	-
3	3.1a	30	Study Completion	-
3	3.2	4	Animal Welfare	Internal bleeding
3	3.3	30	Study Completion	-
3	3.4a	10	Increasing Motor Torque	Suspected perioperative thrombus ingestion
3	3.5	22	Decreasing Bundle Permeability	Thrombosis at lower bearing
3	3.6	30	Study Completion	-
3	3.7	28	Study Completion	Thrombosis at upper bearing
3	3.8a	14	Animal Welfare	Uncontrollable bleeding at cannulation site

^aThese trials required a single device exchange within the immediate postoperative period. These exchanges are discussed further within the Discussion section.

Figure 2:

CFD-predicted shear rates (s⁻¹) at the lower impeller bearing (left column) and velocity magnitudes (m/s) at the upper impeller bearing (right column) for all evaluated designs. Schematics show cross-sectional views of the device bearing designs with colored space corresponding to blood-filled pathways. Shear rates or velocity magnitudes of blood flow within these spaces are indicated via color according to the scales included within each image.

Design 3 was utilized in 8 trials. Trials 3.2 and 3.8 ended prematurely due to perioperative or cannula-related bleeding. During trial 3.2, internal bleeding that appeared to be a result of damage to the superior vena cava sustained during the operative or recovery period required euthanasia at POD 4. During trial 3.8, external bleeding at the cannulation site required blood transfusions at POD 9 and 10 and ultimately led to euthanasia at POD 14. Data for the remaining trials is discussed in the following sections. Four of the 6 remaining trials lasted for the entire month-long support period. Trial 3.4 ended on POD 10 due to increasing motor torque while trial 3.5 ended on POD 22 due to decreasing bundle permeability. Further details for trials 3.4 and 3.5 will be discussed below. Subcutaneous bleeding at the cannulation site during trial 3.3 required a blood transfusion at POD 7 but subsided thereafter.

As noted in Table 1, multiple trials required a device exchange within 2 hours of connection due to bundle thrombosis. This was unexpected based on the absence of device exchanges during prior in vivo trials.^19,22 Investigation revealed that toxic intradevice ethylene oxide concentrations (> 30 ppm) remained after our passive degassing process following sterilization. Following trial 3.4, devices were degassed under nitrogen gas flow for 24 hours prior to use. The only subsequent device exchange (trial 3.8) occurred due to insufficient heparinization at cannulation resulting in a measured ACT below the targeted value. No other device exchanges were made during support.

Device Performance and Hematological Results

Motor torque is shown in Figure 3a while Figure 3b shows normalized torque change which accounts for variations in pump speed. Average normalized motor torque change for completed trials ranged between −1 and 2% and exhibited no significant change with time (p=0.52). Persistent intravascular cannula suction events required sustained pump speed reductions for trial 3.7 on POD 18-20. Pump speed was maintained at 1350-1400 revolutions/min following initial adjustments for all other trials. As shown in Figure 3b, Trial 3.4 exhibited increased and variable torque from the beginning of support. Closer inspection revealed a step change in motor torque within the first 24 hours postsurgery followed by erratic growth. In our experience, this is indicative of device ingestion of particulates or thrombus from upstream during the perioperative period. Due to the early onset and suspected nature of irregularities within trial 3.4, it was excluded from subsequent analyses.

Figure 3:

(a) Raw motor torque and (b) normalized motor change during support. Data points represent mean values and error bars represent standard deviations.

Device blood flow rate is shown in Figure 4. Average flow across completed trials did not change significantly with time (p=0.38). Blood flow rate in trial 3.7 decreased to approximately 1.0 L/min due to the aforementioned intravascular cannula suction and led to increased variability following POD 18. All other trials remained between 1.9-2.4 L/min. HFM bundle permeability is shown in Figure 5. Average bundle permeability across completed trials decreased slightly within the first 10 days but remained steady thereafter. Average permeability across completed trials did not change significantly with time (p=0.68). Bundle permeability in trial 3.5 rapidly decreased by approximately 50% over 24 hours on POD 22, thereby leading to trial termination. Device oxygen transfer rate was 143 ± 25 mL/min across all measurements and blood oxygen saturation at the device outlet was always 100%. Device CO₂ removal rate (Figure 6a) was 139 ± 14 mL/min across all measurements. Blood pCO₂ at the bundle inlet was 44 ± 4 mmHg and CO₂ removal rate variability was primarily due to differences in blood flow rate at the time of measurement. Figure 6b shows normalized CO₂ removal rates accounting for variation in blood flow rate. Benchtop CO₂ removal results used for normalization are given in Figure S2. Average normalized CO₂ removal rate for completed trials was 91-97% over the course of support. There was a significant effect of time on normalized CO₂ removal rate for completed trials (p < 0.01) with measurements after POD 5 being statistically different from baseline values.

Figure 4:

Device blood flow rate during support. Data points represent mean values and error bars represent standard deviations.

Figure 5:

Effective HFM bundle permeability during support. Data points represent mean values and error bars represent standard deviations.

Figure 6:

Raw CO₂ removal rate (a) and normalized CO₂ removal rate (b) during support. Data points represent mean values and error bars represent standard deviations.

The ACT (normalized to baseline) across all trials was 1.8 ± 0.2. Additional hematological data are summarized in Table 2. It should be noted that trials 3.5 and 3.7 exhibited elevated baseline soluble P-selectin relative to other trials (trial 3.5 = 8.6 ng/mL, trial 3.7= 6.2 ng/mL, remaining trials = 3.4 ± 0.3 ng/mL).

Table 2:

Summary of Hematological Results

	POD
Measurements	Baseline	0	1 to 5	6 to 10	11 to 15	16 to 20	21 to 25	26 to 30
PfHb [mg/dL]	18 ± 3	13 ± 1	16 ± 3	15 ± 4	15 ± 1	16 ± 3	16 ± 5	14 ± 2a
LDH [IU/L]	401 ± 31	315 ± 24b	446 ± 44	341 ± 70b	361 ± 83	357 ± 64	357 ± 39	367 ± 24a
Creatine Kinase [U/L]	97 ± 45	153 ± 106	151 ± 148	87 ± 52	91 ± 54	87 ± 38	90 ± 33	88 ± 48a
WBC [K/uL]	9 ± 2	7 ± 2b	12 ± 3b	12 ± 2b	11 ± 2b	12 ± 2b	11 ± 2	10 ± 1a
Hct [%]	35 ± 3	23 ± 3b	26 ± 3b	24 ± 3b	26 ± 5b	26 ± 4b	26 ± 4b	29 ± 4a,b
Hgb [g/dL]	12 ± 1	8 ± 1b	9 ± 1b	8 ± 1b	8 ± 2b	9 ± 1b	9 ± 1b	9 ± 1a,b
Platelets [K/uL]	602 ± 129	378 ± 123b	417 ± 152	542 ± 135	687 ± 118	791 ± 139	630 ± 344a	972 ± 820a
P-Selectin Positive Platelets [%]	6 ± 3	4 ± 2	6 ± 3	6 ± 2	10 ± 8	7 ± 3	8 ± 4a	8 ± 5a
P-Selectin Positive Platelets - PAF activated [%]	75 ± 5	66 ± 11b	72 ± 5	79 ± 12	82 ± 10	78 ± 11	80 ± 7a	78 ± 3a
Soluble P-Selectin [ng/mL]	5 ± 2	4 ± 1	4 ± 2	4 ± 2	5 ± 2	5 ± 2	4 ± 2	5 ± 2a
ALT [U/L]	15 ± 4	11 ± 3	14 ± 3	13 ± 5	13 ± 6	12 ± 6	12 ± 7	13 ± 9a
AST [U/L]	77 ± 12	60 ± 12b	66 ± 13b	48 ± 10b	49 ± 6b	44 ± 8b	45 ± 10b	51 ± 3a,b
BUN [mg/dL]	11 ± 3	11 ± 4	13 ± 3	11 ± 2	10 ± 2	11 ± 3	9 ± 3	10 ± 2a
Creatinine [mg/dL]	1.0 ± 0.1	0.8 ± 0.1b	1.1 ± 0.2	0.9 ± 0.1	0.8 ± 0.1b	0.8 ± 0.1b	0.9 ± 0.1b	0.9 ± 0.1a

Each measurement had n = 5 unless otherwise specified.

^aThese measurements had n = 4

^bSignificantly different relative to corresponding baseline value (p < 0.05). PfHb = plasma free hemoglobin, LDH = lactate dehydrogenase, WBC = white blood cells, Hct = hematocrit, Hgb = hemoglobin, PAF = platelet activating factor, ALT = alanine aminotransferase, AST = aspartate aminotransferase, BUN = blood urea nitrogen.

End of Study Findings

Figure 7 shows macroscopic images and scanning electron micrographs of regions within the device where clotting was observed during some trials. Explanted devices from Trials 3.1, 3.3, and 3.6 appeared similar and exhibited no significant intradevice thrombus (Figure 7, top row). The explanted device from trial 3.5 (Figure 7, bottom row) showed minor thrombus at the lower impeller bearing and small, dark thrombi deposited on the majority of the bundle inlet face. Bundle thrombus was limited to the first 10% of HFM layers at the inlet. The explanted device from trial 3.7 exhibited minor ring thrombus around the upper impeller pivot and nonadherent thrombi at the impeller vanes and bundle inlet face (remainder of bundle was clean). The cannula from Trial 3.7 exhibited significant thrombus that partially or completely occluded inflow and outflow holes. Histological examination revealed bacterial colonies in the recovered thrombus and a positive blood culture for trial 3.7 (cultures from all other trials were negative).

Figure 7:

Photographs and scanning electron micrographs of the lower impeller bearing, bundle inlet, and bundle middle for trials 3.6 (top row) and 3.5 (bottom row). The lower bearing images (left most column) show the bearing cup (which supports the end of the rounded pivot during use) and the 6 flushing ports emanating radially outward from the cup. Micrographs of the bundle show the hollow fiber membranes (oriented generally vertically) and the woven weft threads that maintain fiber spacing/positioning (oriented generally horizontally).

Inspection of the heart, lungs, liver, kidneys, and spleen showed no gross abnormalities for trials 3.1, 3.3, and 3.5. The sheep from trial 3.6 exhibited a small region of atelectasis at the lingula of the left lung but no other abnormalities. The sheep from trial 3.7 exhibited a small region of emphysema in the right lung apical lobe, slight hepatomegaly, and multiple thrombi within the right ventricle and atrium. Histopathological examination of organ tissues revealed primarily normal morphology.

DISCUSSION

Active rehabilitation during long-term respiratory support is widely recognized as being critically important to reducing patient morbidity and mortality.^12-14 Ambulation during ECMO, however, remains logistically challenging with currently available systems and is not implemented at most centers.^16,17 A wearable artificial lung could ease implementation of ambulatory support and potentially improve long-term patient outcomes. The ModELAS is a versatile, compact pump-lung being developed for use during adult, pediatric, and low-flow ECCO₂R respiratory support.^19-23 This work used studies in healthy sheep to evaluate the ability of the ModELAS to provide month-long venovenous support while maintaining normal physiology. CFD-guided design modifications to the impeller bearings during the evolution from device design 1 to the final device design 3 successfully decreased the tendency for pump thrombosis and improved long-term device reliability. In vivo results with the optimized ModELAS demonstrated its ability to provide up to 1 month of continuous support. Events adversely affecting animal physiology were limited to cannula-related and perioperative complications. Four of 5 trials advancing past these early complications ran to study completion without signs of deteriorating performance while 1 trial ended after 22 days of support due to device thrombosis.

Other researchers have also pursued the development of integrated pumping artificial lungs. Griffith and colleagues developed a wearable artificial pump lung (APL) consisting of a custom annular HFM bundle (0.8 m²) integrated with a blood pump. The APL was used with central cannulation and exhibited primarily positive results during 30-day sheep studies.^29,30 The Mobybox (Hemovent, Aachen, Germany) is a device currently under development that integrates a pneumatically-driven pump and HFM bundle (1.6 m²) into a single housing. A recent study of the Mobybox showed positive safety and efficacy during 7-day sheep studies but long-term performance has yet to be evaluated.³¹ The ModELAS similarly integrates a pump and gas exchanger into a single device, but utilizes a smaller HFM bundle while still maintaining a rated flow of approximately 4 L/min. The optimized ModELAS bundle design results in minimal stagnation and highly efficient gas exchange even at blood flow rates as low as 500 mL/min.²¹ Additionally, the integrated centrifugal pump has sufficient pumping capacity for use with higher-resistance dual-lumen cannulae. Thus, the ModELAS is capable of a wide range of respiratory assist applications including use during low-flow ECCO₂R using dialysis-like cannula sizes.

In vivo trials in this study were hampered by adverse cannula-related and perioperative complications, but demonstrated the ability of the ModELAS to provide month-long support with minimal deterioration in device performance. Four of the 5 trials advancing past early complications ran to completion and exhibited stable device pumping and gas exchange. The absence of significant changes in blood flow rate and normalized motor torque in these trials indicates maintained pumping capabilities. Stable function of the integrated gas exchanger was demonstrated by a lack of significant decreases in bundle permeability and CO₂ removal. Blood flow rates in this study are below the rated flow of the ModELAS (~4 L/min) and thus oxygen transfer is perfusion-limited and not a reliable indicator of bundle performance. In contrast, normalized CO₂ removal is sensitive to performance deterioration. Although there was a statistically significant effect of time on normalized CO₂ removal, the magnitude of change (4%) was modest and rates remained stable thereafter. Explanted devices from completed trials were primarily free of thrombus. During trial 3.7, the extensive thrombus surrounding the cannula was most likely the source of nonadherent thrombi observed within the device as well as within the right heart. Trial 3.5 was the only long-term trial that exhibited a decline in device performance and did not complete the targeted 1-month duration of support. The distribution of discrete thrombi on the bundle inlet face and presence of similar thrombus within the lower bearing suggested transient clots were generated at the bearing and shed to the bundle. Scanning electron micrographs of the bundle inlet also appear to be consistent with thromboemboli trapping. The absence of a substantial decrease in normalized CO₂ removal rate for this trial indicates that the majority of HFM surface area was not impaired and supports the observation that thrombus was limited to the bundle inlet face.

ModELAS support was generally well tolerated by the sheep. There were no signs of hemolysis during support. Markers of tissue and cell damage (LDH, creatine kinase) exhibited a transient increase following surgery before returning to baseline levels or lower. White blood cell count was elevated during the beginning of support but returned to baseline levels during the final 10 days. The decrease in hematocrit and hemoglobin on POD 0 is likely due to fluid infusion and minor blood loss during surgery and is consistent with other similar studies.^29,32 Cannula-related bleeding in trial 3.3 also likely contributed to lowered levels during the first 10 days of support. Both parameters, however, generally trended upward for the remainder of support. Platelet count was reduced on the day of the surgery but subsequently stabilized at or above baseline values consistent with a lack of platelet consumption by the device. Platelet activation also did not significantly differ from baseline during support. Baseline levels of soluble P-selectin for trials 3.5 and 3.7, however, were approximately twice those of all other trials. Notably, these trials were the only 2 (of the 5 not terminated from early complications) to exhibit any intradevice thrombosis. It is possible that sheep from these trials had a preexisting condition (such as infection or inflammation) that may have contributed to a hypercoagulable state. Markers indicating function of the kidneys (BUN, creatinine) and liver (ALT, AST) did not exhibit any consistent increase and thus did not reflect any functional impairment. Similarly, examination of kidney and liver tissues following support did not show any consistent signs of device-related damage. The only meaningful abnormalities observed for the spleen, heart or lungs was thrombi observed within the right heart during trial 3.7 which is suspected to have originated at the cannula.

In conclusion, these studies demonstrated promising potential for the ModELAS during long-term support. Complications were primarily limited to cannula-related challenges that are specific to experiments in sheep and not expected during clinical use. The ModELAS demonstrated stable long-term pumping and gas exchange without signs of blood trauma. Device thrombosis affecting support was limited to a single trial and may have been promoted by a preexisting procoagulant state in this sheep. Future work will focus on the implementation of a thromboresistant coating within the ModELAS to enable longer support with reduced likelihood of thrombosis.

ABBREVIATIONS

ACT

activated clotting time

ALT

alanine aminotransferase

AST

aspartate aminotransferase

BUN

blood urea nitrogen

CFD

computational fluid dynamics

CK

creatine kinase

DLC

dual-lumen catheter

ECCO₂R

extracorporeal CO₂ removal

ECMO

extracorporeal membrane oxygenation

Hct

hematocrit

HFM

hollow fiber membrane

Hgb

hemoglobin

IV

intravenous

ModELAS

modular extracorporeal lung assist system

MV

mechanical ventilation

PAF

platelet activating factor

PfHb

plasma free hemoglobin

POD

postoperative day

WBC

white blood cells