The development of devices for long-term (> 2 weeks) mechanical circulatory support (MCS) in pediatrics remains a challenge, both technically and commercially. The patient population is small in number but diverse in body size and underlying disease. The Berlin Heart EXCOR (Berlin Heart GmbH, Berlin, Germany) ventricular assist device (VAD) remains the only Food and Drug Administration (FDA)-approved device for chronic support in patients less than 6 years of age. Adverse events related to pump thrombosis, embolic stroke, and bleeding associated with high levels of anticoagulation, occur at higher rates than seen with adult VADs,1,2 especially in the smallest <10 kg patients.
In the United States, the National Institutes of Health has supported Pediatric MCS research and development through the Pediatric Circulatory Support contract program,3,4 in which our group participated, the follow-on PumpKIN program,5 and other grant mechanisms. The Penn State Infant VAD has undergone a series of minor design iterations to improve performance and manufacturability, develop a cannula system, and develop a portable pneumatic driver. The goal of this work has been to provide a safe device, with a rate of neurologic dysfunction at least as low as that in adult devices, and in the process to improve our understanding of device thrombosis in the ovine animal model.
This article describes the preclinical status of the device and the results of recent animal testing. The objectives have been to 1) demonstrate acceptably low thromboembolic risk to support FDA approval, 2) challenge the device by using minimal to no anticoagulation in order to identify any design or manufacturing weaknesses, and 3) improve our understanding of device thrombogenicity in the ovine animal model, using multicomponent measurements of the coagulation system and renal ischemia quantification, in order to better correlate animal results with human results.
Material and Methods
The Penn State Infant VAD is a pulsatile pneumatically actuated pump with a 12–14 ml stroke volume (Figure 1). It can provide left VAD (LVAD), right VAD, or biventricular VAD support in excess of 1 year. The maximum output is 1.6 L/min, appropriate for full left or right sided support in infants up to approximately 10 kg or body surface area of 0.5 m2, and partial support for larger patients.
The pump design is based on the adult Thoratec pneumatic VAD, which was designed at Penn State (Pierce-Donachy VAD) and successfully implanted in more than 5000 patients worldwide. Although continuous flow pumps are the mainstay of adult MCS, there are intrinsic advantages of a paracorporeal pneumatic pump for this population (infants, hospitalized, not destination therapy). These include physiologic pulsatile flow, simple preload-sensitive automatic control of pump flow, lower shear stresses, and cannulation flexibility in providing left, right, or biventricular support, in the setting of congenital anatomic variability.
Our efforts have focused on minimizing thrombus formation and thromboembolization, which is technically challenging in an infant-sized pump because of the low Reynolds number. The pump uses monostrut Delrin disk valves adapted from a Björk-Shiley design, which provide a large effective orifice area. This results in high inflow velocities in the pump chamber that produce a sustained rotational flow and effective washing of the sac surface,6–9 without recirculation regions where thrombus can form. These valves, which are now manufactured in our labs, are of the same design and materials that have been used in the successful Thoratec adult pneumatic VAD, Arrow LionHeart VAD, and the Penn State Electric Total Artificial Heart. There have been no documented valve failures with these devices. The pediatric size valves are designed with custom flanges that fit precisely in the pump without steps or gaps, allowing for minimal biological deposit at the valve junctions.
The blood pump uses a seamless, highly smooth blood sac, fabricated from segmented polyether urethane urea. The basic pump design has remained stable with incremental improvements in the valve, sac fabrication, and inspection. The cannulae are 6 mm inner diameter expanded polytetrafluoroethylene (ePTFE) which are modified with helical reinforcement, overcoated, and can be cut to length by the surgeon. The important pump-cannulae connections are designed to minimize steps and gaps.
Preoperative, Surgical and Postoperative Protocols
All animal studies were approved by the Institutional Animal Care and Use Committee of the Penn State College of Medicine. The animals were housed in facilities accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International, and veterinary care was provided in accordance with the eighth edition of The Guide for the Care and Use of Laboratory Animals.10
A detailed explanation of the preoperative, surgical, and postoperative animal care has been described previously.11 After an initial learning curve in which protocols for anesthesia, ventilation, and analgesia were developed, healthy survival was routinely attained. Study duration was originally 4 weeks and then extended to 8 weeks in 2013. Dorset crossbred lambs weighing 18–29 kg (median 24.1 kg) were used during this study phase. Females were used predominantly because of their docile temperament, cleanliness, and better acclimation to the stanchions compared with males.
The implantation was via left thoracotomy with cannulation of the left ventricle and descending aorta without cardiopulmonary bypass. Partial aortic clamping ensuring palpable downstream pressure was first applied to perform outlet cannula anastomosis. A tapered dilator was then inserted into the ventricle to ensure smooth and quick insertion of the inlet cannula, which was secured in place via a presutured felt flange. After carefully connecting both cannulas, filled with saline and blood, to the saline filled pump, the pump was visually inspected for air bubbles. If found, the bubbles were removed through a needle hole in the outlet graft, which was subsequently sutured. The pump was started and implanted subcutaneously in the left flank. The sham animals received anesthesia, thoracotomy, and partial aortic clamping in the duration of the average total anesthesia and anastomosis times from the previous LVAD studies, but did not receive a pump.
Postoperative care, analgesia, and antibiotic therapy have been described previously.11,12 Standard hematology and serum chemistries were measured biweekly beginning 2 weeks preoperatively, through the first 2 weeks postoperatively, and at least weekly thereafter. Global coagulation was measured by thromboelastography (TEG 5000 Thrombelastograph Hemostasis Analyzer System; Haemonetics Corp., Braintree, MA). Activated partial thromboplastin time (aPTT) levels were measured on a CA1500 coagulation analyzer (Dade Behring, Newark, DE) using functional clot-based assays. Heparin levels were measured on the CA1500 using a chromogenic anti-Xa assay. The accuracy of the anti-Xa assay for ovine blood was confirmed by in vitro heparin titration. Platelet aggregation was measured using a whole blood aggregometer (Chrono-Log Model 560CA, Havertown, PA), using adenosine diphosphate (ADP) (0.2, 0.6, 2.5, and 10 µM) and collagen (0.2, 0.6, 2.5, and 10 µg/dL) agonists. Plasma-free hemoglobin samples were collected with Ethylenediaminetetraacetic acid (EDTA) and/or lithium heparin (LiHp) vacutainers and measured according to the Shinowara method.13
Statistical t-test comparing two-samples assuming unequal variances (p < 0.05) was used to compare means of preopertaive and postoperative (>2 weeks) data for all the LVAD animals combined or LVAD and Sham data.
The VAD was operated in automatic full-to-empty detect mode, utilizing the air flow fill detection system. The end-diastolic delay was set to 30 msecs, which was found in previous in vitro studies to minimize hemolysis related to inlet valve closing.14 The diastolic drive pressure (vacuum) was set to deliver a mean VAD flow rate of 1.0–1.3 L/min.
During surgery before aortic anastomosis, unfractionated heparin (UFH) was administered as an initial intravenous (IV) bolus at 300 units/kg and then followed with an initial continuous infusion at a rate of 100 units/kg/hour. The heparin infusion rate was adjusted every 15 minutes to maintain activated clotting time (ACT) between 400 and 600 seconds until the pump was started, at which time heparin was discontinued. Unfractionated heparin was then restarted if necessary after hemostasis was established, typically 6–8 hours postoperatively.
The anticoagulation regimen during the previously reported animal testing conducted with the Penn State Infant VAD12 targeted 1.5–2 times normal aPTT (40–54 seconds). It was then reduced to achieve a target TEG R-time of 2x normal, with anti-Xa monitoring, requiring significantly lower levels of heparin. The goal of this study was to further reduce the heparin dose in order to provide a challenging test for the device and uncover any potential areas of thrombosis. More recently, we have eliminated the use of heparin except to maintain normal R-time during the first 2–3 weeks of the postsurgery acute hypercoagulable phase, similar to the approach of Copeland et al.15
Therefore, during this series of testing, one group of animals was tested targeting two times normal TEG R time (2R LVAD group—six animals), whereas the second group of animals was tested targeting normal TEG R times (1R LVAD group—six animals). No heparin was given after approximately 1–3 weeks in the 1R group. The same anticoagulation regimen was maintained in the two groups of sham animal studies (2R Sham—three animals, and 1R Sham group—two animals).
Platelet inhibitors were not given at any point throughout the study. Aspirin is ineffective in sheep.16 Clopidogrel has been shown to be effective in inhibiting platelet aggregation in the ovine,17 but was not used. Ovine platelets have been shown to be similar to human platelets in shear induced clotting time,18 but less aggregable by ADP and collagen agonists,19 and less adhesive than human platelets to biomaterials in vitro.20 Therefore, platelet inhibition was not justified.
Necropsy and Pump Explant Analysis
At 30–60 days postoperatively, animals were sacrificed and a complete gross necropsy exam (except the brain) was conducted by a single American College of Veterinary Pathologists diplomate pathologist (T.K.C.). Ten minutes after administration of 10,000 units of heparin, the pump with the exposed inlet and outlet cannulae was rinsed with saline under pressure through the inlet port and cut outlet port. The pump was subsequently fixed in formalin, and rerinsed with saline solution approximately 1 hour later. The fully anesthetized animal was humanely euthanized immediately after stopping the pump.
Tissue samples for histology were collected from kidneys, lungs, and any additional tissues that appeared grossly abnormal. Kidneys were individually perfused with normal saline through the renal artery before removal. The size and number of any grossly evident renal infarcts were documented. The ischemic surface area was calculated, and the degree of damage relative to the total kidney surface area was estimated assuming an ellipsoidal shape of the kidney (6, 3, 2 cm). The aorta and carotid, pulmonary, renal, and iliac arteries, as well as the jugular veins, were dissected and examined for gross thromboembolism.
A detailed pump examination was conducted within 2 days of explantation. Standardized macroscopic and microscopic photographs were taken. The location, size, and color of any biological deposits were recorded with respect to pump orientation.
A total of 24 studies (device and surgical sham) have been conducted since the last published report.12 All studies were conducted with the current pump system. Seventeen studies were electively terminated after approximately 30 or 60 days. Five animals died during surgery or within a few hours after surgery, and two on the first postoperative day due to problems unrelated to the LVAD. Table 1 in the online supplement lists all the complications and the reasons for early termination (see Table 1, Supplemental Digital Content, http://links.lww.com/ASAIO/A372).
Among 17 completed studies, 12 animals underwent device implantation and five were surgical sham animals. A complete list of animals in each study group, device or sham at two levels of coagulation represented by the targeted TEG R time, and their study duration is presented in Figure 2. Table 2 in the online supplement provides additional information regarding the variations in study duration (see Table 2, Supplemental Digital Content, http://links.lww.com/ASAIO/A373).
The mean and standard deviation pump flow obtained from hourly data measurements of individual animals throughout the course of the study are presented in Figure 3. The corresponding average beat rate (not shown) ranged from 78.1 ± 5.7 beats/min (BPM) to 93.1 ± 8 BPM. All animals were under automatic full-to-empty control at all times. Average diastolic drive pressure ranged from –1.7 ± 3.3 mm Hg to 27.5 ± 3.2 mm Hg. Diastolic drive pressure was often set to positive levels in order to limit the pump rate as the healthy animals with normal ventricular pressure quickly filled the pump.
The results in this section pertain to the animals represented in Figure 2 (long-term surviving device and sham animals). The graphs represent the average and standard deviation of the specified parameter for the represented group throughout the course of the experiment.
Figure 4 shows the heparin infusion rates and measured markers of coagulation (anti-Xa, aPTT and TEG R time) for 1R and 2R LVAD animals. The rise in heparin dose during the last week of study (2R) stemmed from increased anticoagulation for one animal that had developed a chronic systemic infection (ID 4754) requiring larger dosages of heparin to reach 2R TEG times. In the 1R group, heparin rates were highest in the first few postoperative days, but rates decreased over the first month, and none of these animals received any heparin in the second month.
A direct measure of heparin activity represented by anti-Xa (IU/ml) is presented in Figure 4B. The minimum detectable value of anti-Xa for our lab was originally 0.01 IU/ml and later <0.1 IU/ml; values <0.1 IU/ml are presented as 0.05 IU/ml. As shown, both groups had barely detectable heparin activity throughout the course of the study, much lower than the recommended therapeutic range in humans under the Edmonton Guideline (0.35–0.50 IU/ml).
Activated partial thromboplastin time is presented in Figure 4C. The levels remained consistently within the preoperative normal range during the postoperative study duration. There was no statistical difference (p < 0.05, t-test) between the groups or between the aPTT for the preoperative and postoperative data (> 2 weeks).
Figure 4D shows the TEG R times of the two cohorts, with the 2R (twice normal) group maintained above the normal range, and the 1R group maintained in the normal range. ACT levels peaked during surgery in response to the initial heparin bolus. Afterwards, ACT levels remained in the upper normal range (data not shown).
There is a pronounced and consistent increase in fibrinogen in the first postoperative week in both the animals who underwent thoracotomy for LVAD implantation and those who only underwent thoracotomy (sham group), Figure 5. This transiently hypercoagulable state corresponds to the heparin requirement in the 1R animals.
There was no statistical difference in plasma-free hemoglobin (plHb) measurements between preoperative (6.1 ± 4.2 mg/dl) and postoperative, >2 weeks (5.6 ± 4.0 mg/dl), samples collected with EDTA vacutainers (three animals). There was no statistical difference in plHb measurements between preoperative (3.3 ± 3.2 mg/dl) and postoperative, >2 weeks (2.1 ± 1.1 mg/dl), samples collected with LiHp samples (seven animals).
During two concurrent studies (one in preoperative and one in postoperative phase), plHb samples collected with EDTA vacutainers suddenly became elevated in both animals, although serum samples (nonanticoagulated tubes) showed no evidence of plHb (no hemolysis). The concurrent measurements with 82 EDTA and LiHp samples from five animals (including two shams) showed a problem with EDTA vacutainers. Therefore, LiHp vacutainers were used for further plHb measurements. Platelet aggregation results will be presented in a future publication.
Clinical Outcomes and Necropsy Findings
As summarized above, the LVAD studies were conducted with low levels of heparin resulting in anticoagulation parameters that were normal or near-normal. We used three instruments to discern the occurrence of thromboembolism: clinical course including physical exam, extensive blood work, and organ evaluation upon necropsy.
There were no clinically evident strokes or evidence of end organ dysfunction in any of the animals in either group. Neurologically, no animals showed any focal or generalized deficits or dysfunction. Hepatic serial enzyme and bilirubin levels remained normal. There were no statistical differences between preoperative and postoperative levels of blood urea nitrogen and creatinine, indicating no clinically evident renal damage.
The degree of renal ischemic lesions expressed as a percentage of the cortical surface area of the kidneys for each study animal is presented in Figure 6. There was no statistically significant difference (t-test, p < 0.05 criteria) between LVAD and sham renal scores. Most of the grossly evident renal infarcts were superficial and covered less than 3.5% of total cortical surface area. Additionally, we have investigated the use of urinary biomarkers of renal ischemia,21 and these results are being published separately.
The lungs in all 12 device-implanted long-term animals were grossly normal (pink and spongy) except for five animals that had various degrees of grossly consolidated right cranial lung lobe lesions consistent with chronic enzootic pneumonia due to Mycoplasma ovipneumoniae. This pathogen causes subclinical chronic pneumonia in sheep and is relatively common in commercially available sheep herds. None of the animals had evidence of clinical disease. One animal showed a solid green irregular mass in the cranioventral portion of the right caudal lung lobe, typical of caseous lymphadenitis due to Corynebacterium pseudotuberculosis. A few animals had a thromboembolus in either left or right pulmonary artery at the caudal lung region. Considering the location of the thromboemboli, this is consistent with an origin from the jugular catheter implantation sites and could not originate from the pump. There was no evidence of thromboembolism in any other organ of any of the study animals. There were no pump exchanges.
Pump Explant Analysis
One animal (ID 4754) had a chronic systemic infection (Pseudomonas aeruginosa, a significant and difficult to treat nosocomial pathogen originating from the jugular catheter) and showed significant biological deposits covering all blood contact surfaces (pump and cannulae). The deposits were well-adhered, and there were no renal infarcts or evidence of embolization. This was an unplanned severe test of the device which may be relevant to the device performance in patients with infection. That animal is excluded from the summary in this section.
We found that most thrombus formations were associated with the pump assembly, surgical installation, or surface finish problems. There was no discernible difference in the magnitude of deposits between 1R and 2R animals.
The left ventricular apical tip (Figure 7A) was generally well adhered and patent. In ID 2123, the graft material in the inlet cannula tip was not fully sealed (bonded) and partially (40%) occluded the lumen (image not shown). The aortic anastomoses (Figure 7B) were clean and patent in all cases. The ePTFE grafts (Figure 7C) showed sporadic gelatinous well-adhered deposits, most commonly distal to small kinks.
Four studies had no and three studies had only 1–2 (<1 mm) macroscopic adhesions in blood sacs (Figure 7D). Four sacs had several thin well-adhered crescent-shaped thrombus formations, mostly in the upper roll region of the sac. A change was made to reduce the sac thickness variation in the interport region to improve the consistency of the sac flexing in that region. In the final three studies after that change, there was only a single (<1 mm) size deposit in one blood sac.
The connector–graft junctions and the valve–connector junctions (Figure 7, E and F) were generally clean or had a thin fibrin ring. In two animals, there was a gap during pump assembly between the outlet valve and the outlet connector, resulting in a thin thrombus, and in one case the inlet graft was cut short and was partially pulled off the connector. Most frequently, deposits were found on the valve–sac junctions (Figure 7G), slightly more on the outlet side than on the inlet side. The valves utilized in this pump have a flange geometry that mates precisely with the blood sac, but a seam is unavoidable. The valve disks (Figure 7H) were clean with occasional indents from the strut contact. In one case, the disk edge wore due to two unpolished machine marks on the major strut and in one case due to inadequate disk-to-strut clearance.
The magnitude of renal infarcts did not correlate with the number and amount of biological deposits found in pump and cannulae. The “cleanest” pump (ID 403) had renal infarcts over approximately 1% of surface area, whereas the pump with the most biological deposits (ID 4754) had no macroscopic infarcts at all.
A major challenge in the development of small, infant-sized blood pumps for chronic use has been reducing thromboembolism. Fluid flow in small pumps is characterized by low Reynolds number and less turbulent mixing at the walls. Recirculation and stasis behind the valve leaflets of the EXCOR pump is presumably a factor in thrombus formation. Our approach has been to use tilting disk valves with an open flow path and an established history in adult VADs, while focusing on pump flow patterns, surface finish, and minimizing gaps at component junctions.
Our anticoagulation regimen was minimal. Unfractionated heparin was titrated to achieve a normal and twice normal TEG R time (Groups 1R and 2R). Both groups were essentially normocoagulable, with undetectable anti-Xa levels and normal aPTT. By comparison, the Edmonton guidelines22 for the Berlin Heart EXCOR calls for UFH titrated to anti-Xa 0.35–0.5 IU/ml and aPTT 1.5–2.5 times baseline, plus platelet inhibitors. Interestingly, a recent modification of the human protocol by the Stanford group23 demonstrated that increasing platelet inhibition reduced the incidence of stroke, pointing to the role of platelets in device-related stroke. Our decision to not inhibit platelets further adds to the thrombogenic challenge of our animal protocol.
The difference in heparin target levels between the 1R and 2R groups was evolutionary rather than designed to test heparin response. The anticoagulation levels in both groups are so low as to be undetectable by anti-Xa, aPTT, or ACT. Only TEG was sensitive enough to measure any differences. Hence, both groups may be considered nonanticoagulated by most clinical definitions.
A secondary goal of this work has been to develop more rigorous (and broadly applicable) animal testing protocols, since these studies are expensive and the number of test subjects is small. We have instituted multicomponent anticoagulation and platelet monitoring. The use of TEG as a measure of whole blood coagulation is likely to be more comparable between species than plasma-based assays reflecting only portions of the coagulation pathway. TEG is also sensitive and has guided our heparin protocol, as heparin levels have been reduced and finally eliminated. ACT and aPTT were found to be unreliable at low heparin levels.
All animals (including shams) showed a consistent acute phase response post surgery, reflected in high fibrinogen and TEG maximum amplitude levels, peaking at approximately 7 days postoperatively. This hypercoagulable phase was treated with heparin, to maintain normal coagulability in the 1R animals. Platelet inhibition and/or corticosteroids24 would be indicated in humans during this phase but was not used in these studies. We have previously measured the effect of clopidogrel in sheep19 but chose to avoid its use in order to further challenge the device.
We have also developed quantitative measures of subclinical renal ischemia—an important measure of thromboembolism in MCS animal testing. These include careful measurements of the total cortical surface area of renal infarcts, with comparison to surgical shams. We have also measured urinary biomarkers of renal ischemia, which provides a sensitive and temporal marker of renal ischemia.21
A number of minor improvements were made in the manufacturing process as a result of the animal testing which might not have been identified if we had used higher levels of heparin and/or platelet inhibitors. There were no blood sac or valve failures during in vivo or in vitro testing. In vitro durability testing of five systems tested at ~100 BPM has demonstrated reliable performance for 2811 cumulative days (tests were electively stopped after 379–675 days).
The inclusion of surgical shams as a control group was found to be a valuable method to separate possible device effects from surgery and IV anticoagulation effects. Of interest was the finding that the highest magnitude renal infarction (approximately 3% of total cortical surface area) was in a sham animal.
It should be noted that the presence of adherent thrombi and renal infarcts are presented here in the context of a relatively extreme test of the limits of the VAD and with a goal of developing more thorough, quantitative measures that may improve the sensitivity of animal studies in MCS. The levels we report here are well below the acceptable limits for human clinical use, based on the comparison here to surgical shams and prior experience in testing adult VADs in animals. In addition, clinical and blood markers of end organ function did not differ from surgical shams in all cases (except in one animal that developed chronic systemic infection unrelated to the LVAD), and there was no clinical evidence of thromboembolism in any animals.
These results represent a limited number of studies conducted in an animal model. Arguably, the organ of most interest, the brain, is likely unfairly protected in our model, because the outlet cannula is grafted to the descending aorta, thus directing any potential thromboemboli away from the brain. The brain was not examined for subclinical strokes, based on previous work by our group and others25 where cerebral ischemic lesions are rarely found in animals even when occurring in the kidneys. It remains possible that our many surveillance methods to detect thromboembolism do not have sufficient sensitivity.
The ovine model is generally accepted as the best species for cardiovascular device testing. However, hematologic differences include the reduced platelet adhesion as noted, but with a tendency toward hypercoagulability due to higher platelet count and diminished thrombolysis, relative to humans. The relative importance of these mechanisms, and appropriate anticoagulation and antiplatelet approaches, requires further exploration and was a motivation for the multicomponent testing in this study.
The number of animals was limited to 12 LVAD and five sham primarily because of the cost of 24 hour per day care, rather than a statistical power analysis. However, even in the renal infarct data (one data point per animal), we estimate a power of 0.7 to detect a difference of 2 standard deviations at the 0.05 significance level.
The choice of the 18–29 kg weight range (approximately twice the weight of the expected human recipients) was a compromise to avoid using unweaned lambs, improve survival, provide adequate blood volume for sampling, and allow full implantation of the pump. The LVAD provided partial support of approximately half of the total output. Although the sensitivity to detect thromboembolism and systemic effects is likely to decrease as the animal size increases relative to the pump, the similarity of the postoperative to preoperative data, and LVAD to sham data, suggests that similar conclusions would be reached if tested in smaller animals.
The pumps were all operated in similar, nominal flow ranges, with adequate pump filling pressure provided by healthy left ventricles. Future studies will investigate the effect of low pump rate on device-related thromboembolism.
The results of this study support the conclusion that the Penn State Infant VAD has a low rate of thromboembolism comparable to surgical shams, using a challenging protocol in which heparin is used but at doses that result in effectively no anticoagulation. The low heparin protocol has allowed the identification of minor design improvements. Finally, extensive monitoring of coagulation parameters has improved our understanding of temporal changes in coagulation, and quantification of renal ischemic lesions has improved our ability to compare animal results among study groups, especially surgical shams.
The authors acknowledge and thank Christopher Siedlecki, PhD, Keefe Manning, PhD, Steven Deutsch, PhD, Gerson Rosenberg, PhD, Deloris Shepps, RFNA, Nathan Shanaman, CST, Ronald Wilson, VMD, MS, Joy L. Ellwanger, BS, CVT, Lori Davis, BAS, AA, CVT, Erin Mattern, BS, RLAT, and numerous animal care technicians who cared for the animals 24 hours/day for >1000 days.