Pre-clinical evaluation of an adult extracorporeal carbon dioxide removal system with active mixing for pediatric respiratory support

Device description

The Hemolung RAS (ALung Technologies Inc., Pittsburgh, PA, USA) consists of a cylindrical cartridge containing a cylindrical bundle of hollow fiber membranes (HFMs) arranged lengthwise and potted annularly between the outside housing and a rotating, hollow, cylindrical core. A cross-sectional schematic of the Hemolung device is shown in Figure 1. The rotating core centrifugally pumps blood from an upper central inlet channel through distribution channels in the upper end of the rotating core that run from the center of the core to the outer surface. From the distribution channels, the blood flows downward, across the bundle of hollow fibers in the space between the rotating cylinder and outer housing wall (see the expanded area in Fig. 1), exiting through an outlet port at the bottom of the outer housing.

The rotation of the cylindrical core also serves to generate secondary flow patterns adjacent to the fibers which enhance gas transfer efficiency by breaking down diffusional boundary layers. The thickness of the gap between the core and the bundle of hollow fiber membranes, as well as the radius and optimal rotational speed of the core were chosen to ideally generate Taylor vortices in the gap. The actual formation of Taylor vortices is exceedingly difficult to confirm for this design, however gas exchange has been verified to improve by more than 60% due to rotation of the core as shown in Figure 2. The shear stress generated on the outer surface of the core at maximum rotational speeds used for therapy (1400 rpm) is estimated to be approximately 100 dyne/cm², assuming laminar flow, and conservatively, up to 1000 dynes/cm² if laminar flow is not assumed. This range is well below the accepted level of 1500 dyne/cm² for red cell lysis (5). In clinical feasibility studies of the Hemolung device in adults, there were no reports of clinically significant hemolysis.

The hollow fiber membranes are microporous polypropylene coated with an ultrathin (1 micron) nonporous layer of siloxane to prevent wetting. The siloxane layer is further coated with covalently-bound heparin to prevent thrombus formation. The total fiber surface area of the Hemolung is 0.59 m², which is more than 90% of the total circuit surface area to which the blood is exposed. Sweep gas (pure O₂ or room air) is pulled through fibers under negative pressure with a vacuum pump to prevent air from crossing into the blood in the event of a fiber break.

The core rotates freely on a fixed shaft with a bearing flushed by infused 0.9% NaCl sterile saline (30 ml/h). The saline flush flows from inside the bearing space below the blood compartment across the seal and into the blood compartment. The saline flush both prevents blood from leaking into the rotor compartment, and helps to prevent stagnation of blood in the blood compartment space below the spinning core, adjacent to the seal. Magnets in the rotating core are coupled to driving-magnets on the Hemolung controller. Circuit blood flow is controlled by setting the core rotation rate on the controller interface. Sweep gas flow rate is also controlled directly on the controller interface. Gas pathway equipment necessary for measurement of the rate of CO₂ removal is integrated into the controller, and the CO₂ removal rate is displayed in real time on the user interface. In the pediatric configuration being investigated, the circuit inlet and outlet tubing (each 6 ft (1.82 m) long, ¼” (6.35 mm) inner diameter) are connected to an 8 Fr or 10 Fr one-piece femoral venous cannula (96830-008/96830-010; Medtronic, Minneapolis, MN, USA) and an 8 Fr one-piece arterial cannula (77008; Medtronic, Minneapolis, MN, USA), respectively. The Hemolung cartridge has a measured priming volume of 144 mL, and total circuit priming volume is 267 mL.

In vitro testing

In vivo testing

Study termination and necropsy

At the end of each study, a final round of blood samples were drawn. A heparin bolus was administered (100 U/kg) prior to euthanasia. The Hemolung circuit was disconnected and flushed with heparinized saline and inspected for thrombus deposition. A complete necropsy was conducted with ECC cannula in place. Indwelling vessels were opened and inspected for gross damage or thrombus related to cannula placement.

Statistical analysis

Measurements associated with hematology and organ function on Day 7 of chronic in vivo studies were compared to baseline levels using a two-tailed Student’s t-test in Microsoft Excel. Equality of variance (i.e., hetero- or homoscedasticity) was determined with an f-test to justify use of parametric statistical comparisons.

In vitro testing

The pump performance curves for the Hemolung cartridge are shown in Figure 3A. The pump curves represent the ability of the Hemolung to generate a pressure head against an afterload for a given pump speed and circuit flow. Test conditions and the resulting pump curves roughly simulate veno-venous attachment since net head across inflow and outflow cannula during in vitro testing was negligible. The Hemolung RAS achieved fluid flow rates of 0.3 l/min at 1000 rpm with a pressure head of approximately 100 mmHg and with both cannula (i.e., 8 or 10 Fr inflow and 8 Fr outflow). Maximum blood flow was 0.82 l/min at 2000 rpm against a pressure head of 320 mmHg.

In vitro gas exchange rates in blood with and without pediatric cannulation are shown in Figure 3B together with necessary O₂ delivery to fully replenish metabolic consumption versus body weight in kilograms (W_kg) for pediatric applications (125 mL O₂/BSA, where BSA[m²] = [4 * W_kg + 7]/[90 + W_kg] (9,10)). The plots are overlaid such that W_kg aligns with relevant circuit blood flow rates (125-150 ml/min per kg) (11). Both normalized CO₂ removal and O₂ delivery increased approximately linearly versus blood flow with and without circuit cannula. The maximum transfer rates with pediatric cannulation were 108 mL CO₂/min and 56 mL O₂/min at 0.83 l/min of blood flow and a pump speed of 2000 rpm. Without cannula, the highest transfer rates were 154 mL CO₂/min (Q_b = 2.0 l/min; 1825 rpm), and 125 mL O₂/min (Q_b = 2.3 l/min; 2000 rpm). The CO₂ removal measurement at 2.3 l/min was omitted since a steady state reading could not be reached at that flow due to the sampling time restrictions associated with the single-pass test loop. Blood at the device outlet remained 100% saturated at the controller’s maximum motor speed (2000 rpm; Q_b = 2.3 l/min) with inlet sO₂ = 63%, indicating the O₂ rated flow exceeds 2.3 l/min.

Acute in vivo

The Hemolung RAS was successfully connected to two animals (S008-13, S009-13) for acute in vivo testing. Maximum blood flow rates available to the Hemolung circuit (i.e., the blood flow rate delivered to the right jugular) were 0.33 l/min at 1000 rpm and 0.47 l/min at 1500 rpm in each animal. Maximum normalized CO₂ removal and O₂ transfer rates were 65.0 ± 8.9 mL CO₂/min and 22.2 ± 2.8 mL O₂/min at Q_b = 0.47 l/min as shown in Figure 4A. Both CO₂ and O₂ transfer rates varied linearly with blood flow at constant sweep gas flow of 8.0 l/min. Circuit blood flow rates in each of the two studies did not match exactly due to the variation in blood vessel diameter for each animal, however gas transfer repeatability between cartridges can be inferred from the overlapping datasets.

Exchange rates with varying blood flow and fixed sweep gas flow are shown in Figure 4B as a fraction of gas transfer at maximum sweep gas flow (8.0 l/min). Oxygen transfer was generally insensitive to increasing sweep gas flow as anticipated, achieving 88% of the O₂ delivery at 2.0 l/min compared to 8.0 l/min. CO₂ removal increased significantly between sweep gas flows of 2.0 and 4.0 l/min (54%), then appeared to plateau as flow approached 8.0 l/min (<10% increase from Q_g = 4.0 to 8.0 l/min). The diminishing return in gas exchange for increasing sweep gas flow is reflective of the high blood flow to sweep gas flow ratio (Q_b:Q_g) and indicates that further increasing sweep gas flow would not significantly affect CO₂ removal.

Chronic in vivo

All 12 animals were successfully connected to the Hemolung RAS. Seven days of treatment was achieved in seven animals, with five studies ending prematurely. None of the five early terminations were a result of the Hemolung RAS, but were due to complications associated with central venous catheters or monitoring lines. A summary of all animal outcomes is listed in Table I. The first animal (S021-11) died unexpectedly after 19 h of treatment due to cranial hematoma and ischemic brain injury. The cannulation in the first animal differed from subsequent animals in that the drainage and perfusion cannula were placed in independent jugulars (i.e., left and right). We believe this arrangement restricted blood flow away from the head, resulting in cranial edema and heightened intracranial pressure, which was the primary cause of the ischemic injury that ultimately lead to study termination. The cannulation strategy was subsequently modified such that both drainage and perfusion cannula were placed in the right external jugular. In two cases (S022-11 and S006-13), the animal was electively euthanized after 5 h and 86 h of treatment due to cannula dislodgement (drainage and perfusion, respectively) from the REJ. Two studies were terminated due to respiratory related issues. Animal S023-11 died suddenly approximately 19 h after initiation of therapy due to pulmonary embolism. A necropsy conducted immediately following the event revealed that the venous monitoring line had migrated 3 cm to 4 cm above the SVC-right atrial junction and was surrounded by thrombus. The pump outflow cannula, which was closest to the thrombus was clean and properly positioned inside the jugular lumen. During surgery, prior to inserting the Hemolung perfusion cannula, a lack of backflow observed in the REJ suggests the thrombus may have developed prior to cannulation. Animal S010-13 experienced respiratory distress shortly following surgery and ventilator extubation, indicated by arterial hypercapnia and hypoxemia. Oxygen was provided by mask intermittently for ~2 h until blood gases recovered, at which time the animal was alert and stable. At hour 84, the animal became agitated with shortness of breath prior to collapsing, at which time the animal was euthanized per IACUC protocol requirements. The Hemolung cartridge, circuitry, and indwelling vessel were clear of thrombus or blockage. Upon opening the right chest a large consolidated clot (6-8 cm axial plane diameter) was found within the chest, superior to the heart and lungs (dorsal to the bronchial region) extending 8-10 cm up the trachea. A small perforation (2-3 mm) was found in the LEJ immediately distal to the tip of the indwelling maintenance line. Margins of the perforation were rough, and advanced clotting in the surrounding tissue was observed. The tip of the maintenance line was cut to a point for ease of insertion prior to placement during surgery, possibly causing the puncture during placement. Gradual aggregation of a consolidated clot in the chest due to repeated agitation from animal movement in the days following surgery culminating in respiratory distress would be consistent with our observations.

The daily averaged blood flow rates are shown in Figure 5A. Circuit flow was maintained at or above target flows (280-300 ml/min) for the duration of most studies. In two animals (S005-13 and S007-13), the jugular became fully exsanguinated at target flows resulting in suction of the vessel onto the drainage cannula, so pump speeds were reduced to achieve 240-260 ml/min (Day 1 averages), where no flow limitations were observed. In these animals, daily average blood flow decreased by 2% to 4%. Cranial edema was observed post-operatively (within 3 h) in three animals (S021-11, S041-12, and S010-13), possibly an adverse reaction to surgery or a result of increased flow resistance in the cranial venous return due to maintenance line and Hemolung circuitry in the jugulars. The Hemolung cartridge pump speed was increased to raise circuit flows in the latter two animals, increasing fluid drainage from the head. Edema waned over several hours in these animals, at which time pump speeds were reduced.

Normalized CO₂ removal rates were averaged daily and are shown in Figure 5B. CO₂ removal rate was anticipated to be similar to the levels achieved in adult sheep with the Hemolung RAS (a minimum of 50 ml/min at a venous PCO₂ = 45 mmHg) (12). Daily venous PCO₂ averages are listed in Table II together with the overall average acid/bases status of venous blood. The overall average reduction in blood PCO₂ due to circuit flow (inlet minus outlet) was 24.5 ± 1.9 mmHg, as shown in Figure 6. Normalized CO₂ removal rates greater than 50 ml/min were maintained for the full duration of all but one study (S005-13), in which low CO₂ removal rates correspond with reduced blood flow rates. Blood drawn from the Hemolung circuit perfusion line remained at 100% oxygen saturation for the duration of all studies, indicating maximal oxygenation of circuit blood flow. Arterial blood gases and acid/base status of each animal was monitored, the daily averages of which are provided in Table III. Arterial pH and oxygenation levels remained within normal physiologic ranges throughout testing. Arterial pCO₂ and HCO₃^- each remained at the lower end of normal ranges for the majority of testing, likely due to the CO₂ removed by the external circuit in addition to that eliminated by the healthy lungs. Base excess was consistently below the typical reference range, which is not surprising given the lower levels of CO₂ and HCO₃^-, with the exception of a spike on Day 3 that was consistent in most animals. This observation is common in chronic sheep studies, and is generally related to diet as animals acclimate and resume more normal eating habits, which can slightly alkalize circulating blood.

The hematology data from collected blood samples are summarized in Table IV. Necropsy and cartridge inspection conducted at the conclusion of each seven-day study revealed minor thrombi concentrated at the fiber knitting in the devices as shown in Figure 7A, but no significant thrombus formation was found in the cartridge or extracorporeal circuit in any of the animals. Some coagulation was observed in one cartridge (S005-13) near the top and bottom of the fiber potting (Fig. 7B). The accumulation was potentially due to flow stasis in those areas, which was likely related to the below-target average blood flow rates. Fibers in proximity of the cartridge outlet were clear of aggregation (Fig. 7C). No evidence of thrombus formation or vessel injury was found after REJ cannula removal in any animal.