Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Device Thrombogenicity Emulation: An In Silico Predictor of In Vitro and In Vivo Ventricular Assist Device Thrombogenicity


Ventricular assist devices (VAD), a mainstay of therapy for advanced and end-stage heart failure, remain plagued by device thrombogenicity. Combining advanced in silico and in vitro methods, Device Thrombogenicity Emulation (DTE) is a device design approach for enhancing VAD thromboresistance. Here we tested DTE efficacy in experimental VAD designs. DTE incorporates iterative design modifications with advanced CFD to compute the propensity of large populations of platelets to activate by flow-induced stresses (statistically representing the VAD ‘Thrombogenic Footprint’). The DTE approach was applied to a VAD (MINDTE) design with a favorable thromboresistance profile and compared against a design (MAXDTE) that generated an intentionally poor thromboresistance profile. DTE predictions were confirmed by testing physical prototypes in vitro by measuring VAD thrombogenicity using the modified prothrombinase assay. Chronic in vivo studies in VAD implanted calves, revealed MINDTE calf surviving well with low platelet activation, whereas the MAXDTE animal sustained thromboembolic strokes. DTE predictions were confirmed, correlating with in vitro and in vivo thrombogenicity, supporting utility in guiding device development, potentially reducing the need for animal studies.


With a steadily increasing clinical burden of heart failure (HF), coupled with a persistent heart transplant donor organ shortage worldwide, ventricular assist devices (VADs) have become the standard of care for advanced and end-stage HF patients. In recent years several VADs have been granted regulatory approval for bridge-to-transplant (BTT) and/or destination therapy (DT) indications by the Food and Drug Administration (FDA)1,2,3. All current approved VAD designs generate non-physiological blood flow patterns, imparting supraphysiologic shear stress to circulating platelets, ultimately activating the blood hemostatic response4. As a result, device recipients are prone to post-implant thromboembolic complications, mandating lifelong antithrombotic regimens5,6. The management of these pharmacologic-al regimens remains a major clinical challenge. Thrombotic complications are routinely reported in these FDA approved device recipients7,8,9,10,11,12,13,14. Unfortunately current antithrombotic therapy, recently demonstrated to have limited overall efficacy15,16, may in fact lead to secondary severe complications, e.g., excessive bleeding events17,18,19,20. Device design optimization for reducing shear-induced blood damage, and for avoidance of excessive anti-thrombotic therapy, is essential for fundamentally improving device thromboresistance and overall clinical safety and efficacy.

A device thromboresistance optimization methodology, Device Thrombogenicity Emulation (DTE), was introduced by our group21,22,23. The DTE combines in silico numerical simulations with in vitro measurements by correlating device hemodynamics with platelet activity coagulation markers – before and after iterative design modifications aimed at achieving optimized thromboresistance performance. Its efficacy was previously demonstrated in prosthetic heart valves and VADs studies21,22,23,24,25. In the MicroMed HeartAssist 5 VAD for example, following its thromboresistance optimization by DTE close to a one order of magnitude reduction in platelet activity was achieved (as compared to the predecessor design on which it was based- the DeBakey™ VAD)22. It additionally reduced platelet activity to a level that was far lower than that of a gold standard VAD – the Thoratec HeartMate II (HMII)24 (the first VAD approved by the FDA for destination therapy). The DTE optimization process also achieved platelet activity reduction that was far more effective than conventional antiplatelet drugs regimen therapy, e.g., ASA and Dipyridamole which are routinely prescribed to device recipients16,26. This in silico/in vitro methodology can potentially reduce the research and development (R&D) costs by developing Mechanical Circulatory Support (MCS) devices that are optimized for thromboresistance before proceeding to costly in vivo animal experiments, and prior to the FDA device approval regulatory process.

We tested the hypothesis that predictions of device thrombogenicity derived via the in silico DTE methodology would correlate with both in vitro and in vivo evidence of actual platelet activation and thrombosis (Fig. 1). Here a single design prototype VAD (VADproto; Fig. 2A) was provided by Abbott Labs with which the authors subsequently conducted an in silico DTE baseline analysis. This in silico analysis demonstrated minimal platelet activation and represented an “optimized” baseline design (MINDTE; Fig. 2B). A modification to this baseline design was simulated and fabricated to yield maximal platelet activation (MAXDTE; Fig. 2B, inset). Both prototypes underwent extensive in vitro comparative testing and chronic animal in vivo studies to evaluate and validate the differential device thrombogenicity levels predicted by the DTE methodology following the design modifications.

Figure 1
figure 1

Conceptual schema for presenting the direct enhanced strategy of Device Thrombogenicity Emulation (DTE) methodology.

Figure 2
figure 2

(A) The Thoratec HMII (left) and VADproto (right) VADs shown at the same scale. (B) Interior design schematics of the MINDTE, and (inset) the modified impeller of MAXDTE with a 5 mm segment of impeller blades removed. The red dotted arrow indicates the flow direction. The inlet and outlet diameters were 11.1 and 12.45 mm, respectively. The simulation boundary conditions were described in text boxes and pointed to the designated sections.

Materials and Methods

In silico simulations

A prototype VAD (VADproto) was provided by Abbott Labs (Thoratec Corp., Pleasanton, CA – now Abbott Labs). VADproto is an axial pump, which adapts several pump design concepts from its successful predecessor – HMII (Fig. 2A). VADproto consists of a three-fin inlet stator (Fig. 2B, purple section) followed by a three blades impeller (Fig. 2B, pink section). An aft portion redirects flow outwardly 90° from the axis of the impeller (Fig. 2B, blue section). Hence, blood flow is redirected without the need for inflow and outflow elbow cannulae (Fig. 2B). Unlike the HMII VAD, and similar to current centrifugal type VAD designs, VADproto is proposed to be inserted directly into the apex of the left ventricle and is smaller and lighter than the HMII (Fig. 2A). In order to confirm the DTE approach, a VADproto pump was intentionally modified (5 mm segment of the impeller blade was removed; Fig. 2B, inset) with the intent to create a lower efficiency, high thrombogenicity pump (MAXDTE) to be used as a negative control. The comparative thrombogenic performance of the two variants was used to evaluate the efficacy of the DTE methodology.

ANSYS DesignModeler (ANSYS Inc., Lebanon, NH) was employed for geometry reconstruction in preparation for numerical mesh generation composed of solid and fluid domains that is needed for fluid structure interaction (FSI) simulations (Fig. 2B). The fluid domain of each pump was segmented into two sections – stator and rotor. For rotating mesh preparation, the fluid domain was divided into two symmetric sections according to the cylindrical symmetricity of the shroud27. Straight cylindrical sections with the lengths of five-times hydraulic diameters were added at the inlet and the outlet to allow for the flow to become fully developed (Fig. 2B, green and turquoise blue sections for inlet and outlet, accordingly). Interfaces were defined between these segments. ANSYS Meshing (ANSYS Inc., Lebanon, NH) was utilized for preparing tetrahedral volumetric meshes of the models. Similar mesh algorithms were applied to ensure comparable mesh densities across the devices. Extra attention was paid to interface meshing to ensure matched mesh densities across the interfaces. Rigorous mesh independence studies were conducted and were achieved with mesh densities of approx. 17 million cells for both models.

ANSYS Fluent CFD solver (ANSYS Fluent Inc., Lebanon, NH) was utilized for conducting the FSI simulations. Blood was modeled as a two-phase Newtonian fluid with viscosity and density of 0.0035 kg/m-s and 1,080 kg/m3. The SST k-ω turbulent model was utilized for both studies due to its superior performance for simulating transient turbulent flows and its ability to accurately predict pressure head in VAD simulations24,28,29. Prior to the transient simulations, steady state simulations were conducted to characterize the pumps’ performances, i.e., pre- and post- design modification, with the benchmark operating conditions – 5.3 L/min and 11.4 kPa of cardiac output and pressure head, accordingly that were provided by Abbott Labs. Moving reference frame was employed for modeling the spinning impeller – defined on the impeller segment cell zones with the designated angular velocities, with zero angular velocity applied on the shroud no-slip condition surfaces. Similarly, stationary cell zones were applied on the aft segments with designated angular velocities defined on the aft-impeller surfaces. 9.548 × 10−2 kg/s and 0 Pa of mass flow rate and pressure were applied as the inlet and outlet boundary conditions correspondingly for each device, and series of steady state simulations were conducted by adjusting the impeller speed until the designated pressure head of 11.4 × 103 Pa was reached – 10,600 and 11,600 rpm for the MINDTE and the MAXDTE, correspondingly (Fig. 2B). Two-way coupled discrete phase model (DPM) was employed during the transient simulations22,30,31. Approx. 9.8 × 103 buoyant 3 µm diameter spherical particles (998.2 kg/m3 of density) representing platelets were seeded and released from the upstream of both designs as flow tracers, corresponding to physiological concentration of a single cross-sectional batch of platelets traveling through the VADs. An optimized time step size of 7.53 × 10−5 s was utilized24, and a simulation duration of 0.105 s was determined for both VADs to ensure at least 90% of the platelet population had exited the flow domains24,28. The MINDTE and MAXDTE reached 18.55 and 20.3 rotations, respectively, during the simulations.

The platelet trajectories and their instantaneous stress tensor components were recorded at each time step, and extracted subsequent to each simulation. The stress tensor of each particle was then rendered into a scalar stress value (σ)32.

$$\sigma =\sqrt{\frac{{\tau }_{11}^{2}+{\tau }_{22}^{2}+{\tau }_{33}^{2}-{\tau }_{11}{\tau }_{22}-{\tau }_{11}{\tau }_{33}-{\tau }_{22}{\tau }_{33}+3({\tau }_{13}^{2}+{\tau }_{23}^{2}+{\tau }_{13}^{2})}{3}}$$

The cumulative stresses that may drive platelets beyond their activation threshold were calculated along multiple flow trajectories as follows: the stress loading history in each platelet trajectory is calculated by summation of the instantaneous linear product of the scalar stress value and the exposure time (texp) – termed Stress Accumulation (SA) according to21,23:

$$SA=\sigma \cdot {t}_{exp}={\int }_{{t}_{0}}^{{t}_{exp}}\sigma (t)dt=\sum _{i=1}^{N}{\sigma }_{i}\cdot {\rm{\Delta }}t$$

where σi, i = 1, 2, …, N, is the nodal scalar value extracted from the total stress tensor and Δt is the corresponding time step between successive nodal points.

To statistically represent the distribution of the large ensemble of stress accumulation (SA) values− each reached by a platelet along its flow trajectory, the statistical distribution of this large SA ensemble is collapsed into a probability density function (PDFs) curve − representing the device ‘thrombogenic footprint’22,23,24 that allows for expedient comparison of the thrombogenic potential generated in the varying designs. To compare the statistical distribution of stress accumulations of different platelet populations, while guaranteeing that the percentage activation is independent of variations in the number of seeded particles and spatiotemporal variations, we have interpolated between the smaller and larger populations’ statistical distributions by applying bootstrapping statistics23 to guarantee that PDFs from the different population sizes are compatible and comparable.

Regions of interest (ROI) analyses were conducted based on the cell zone segments – stator, bearing, impeller-shroud gap, blade, and lower-impeller– to evaluate the local thrombogenic potentials affected by specific design modification in these regions (Fig. 3).

Figure 3
figure 3

Regions of interest (ROIs) in the optimized MINDTE and the MAXDTE. Eight ROIs – stator, bearing, impeller-shroud gap, blade, and lower-impeller were defined for each VAD.

In vitro experiments

Abbott Labs provided two VADproto pumps for testing purposes: MINDTE and the MAXDTE were prototyped by Abbott for conducting the in vitro and in vivo experiments. The operating conditions obtained through steady state simulations (i.e., 10,600 and 11,600 rpm for the MINDTE and MAXDTE pump speeds, respectively, and 5.3 L/min of cardiac output) were applied for calibrating the in vitro recirculation flow-loop which consisted of segments of ½′′ and ¼′′ inner diameter (ID) Tygon R3603 tubes. The length of the ¼′′ ID tubes controlled the pressure head across the VADs22,24, and series of reducing connectors, i.e., 1/2′′ – 3/8′′ and 3/8′′ – 1/4′′ (NovoSci, Conroe, TX), were employed to avoid a sudden expansion and reduction of flow with the transition to the flow resistor24.

Citrated blood (120 ml) was obtained from consented healthy adult volunteers (n = 13) who had abstained from antithrombotic medications, e.g., aspirin and ibuprofen, for at least two weeks prior to blood donation. The fresh blood was collected through venipuncture according to a protocol approved by Stony Brook University IRB. Gel-filtered platelets were prepared and diluted to a concentration of 15 × 103 /µl with modified Tyrode’s buffer22,24. For the thrombogenicity measurements the platelet-buffer solution was recirculated in the flow-loop for 30 min with 10 min sampling intervals (t = 0, 10, 20 and 30 min) utilizing the established platelet activity state (PAS) assay that reports thrombin generation rates22,24. PAS assay employs acetylated prothrombin to inactivate the positive feedbacks during shear-induced platelet activation to reach an one-to-one correlation between the shear dosage and the thrombin generation33. The linearly fitted slopes of the time sampled PAS values (normalized by thrombin generation rates achieved by platelets that are fully activated by sonication) represent the platelet activation rate (PAR) during the 30 min recirculation experiments. Experiments for both devices were conducted simultaneously using the same platelet batches to limit the study variability.

Statistical analysis in silico

The SA from the platelet trajectories was collapsed into the probability density function (PDFs) curve, which concept is an integral part of the numerical methods, and is explained above accordingly. in vitro – The PAR for each device was obtained by taking the linear fitting slope of the averaged time sampled PAS values. A secondary individual-experiment-based PAR calculation was achieved by averaging the linear fitting slope from individual experiments. Student’s t-test was utilized to conduct the statistical comparison of the in vitro results between the MINDTE and MAXDTE with a significance level α = 0.05. A presumed sample size of 10, based on our previous studies, was employed before conducting sample size calculation to reach the 95% confidence level. Results are presented as the mean ± standard error of the mean (SEM), unless otherwise stated22,24.

In vivo animal experiments

The bovine calf model has been extensively utilized to evaluate mechanical circulatory support devices34,35,36,37. Chronic animal studies (30 days) were conducted with the two experimental VAD prototypes implanted in young healthy adult calves (approx. 100 kg to congruent to human size) according to an IACUC protocol approved by University of Arizona (the implantation site) and Stony Brook University. The animals received humane care in compliance with the “Principles of Laboratory Animal Care” (NIH Publication No. 85-23, revised 1985). Briefly, following general anesthesia, VADs were implanted via standard left intercostal thoracotomy, the inflow cannula of the pumps were inserted into the left ventricle (LV) and secured, with outflow grafts anastomosed to the ascending aorta. Pumps were initially run at a lower speed of 7500 rpm. After verifying device functional safety, pump speed was increased to the corresponding operating rpm of each VAD prototype. Upon closing the chest, a single driveline for both operation and/or powering of the LVAD was exteriorized. Post-operatively calves were transferred from the OR and placed in sternal recumbency in an ICU stanchion cart and were monitored continuously by veterinary staff until stable, up to 7 days post-op. The veterinary staff determined when a calf could return to standard housing. The VAD implanted was battery powered by an external controller, allowing the calf to be free from the stanchion when necessary.

Routine platelet activity state measurements were conducted on blood samples extracted from the animals using the PAS assay38,39. Blood samples from the animals were collected in 10% ACD-A at pre-implant (before prepping for surgery), and post-operative day (POD) 0, 1, 2, 3, 4, 5, 7, 9, 14, 18, 21, 25, 28, and 30 (termination). Platelet rich plasma (PRP) was collected for the PAS assay. Chemistry panel (CP), Complete Blood Count (CBC), Basic Metabolic Panel (BMP), and Coagulation Panel were examined daily for clinical assessment. Additionally, thromboelastogram (TEG) was used to measure coagulation. The animals were kept with continuous IV heparin infusion, while bridging to warfarin (BID), maintaining ACT >220 s and INR of 1.8–2.8, respectively. 2D and M-mode echocardiography was used to analyze VAD function and cardiac flow. Euthanasia was performed according to the approved protocol at the completion of the experiment or if termination was needed for other clinical reasons. Calves were heavily sedated with Ketamine 3–4 mg/kg IV to ensure recumbency, and were given an IV barbiturate overdose (Beuthanasia-D) 1cc/10lbs IV. Post euthanasia, gross examination of devices was performed, including examination of the implant site and of device zones (intra-device, via internal examination and device sectioning). In addition, histologic evaluation of the device-tissue interface, device zones (intra-device) and detailed scanning electron microscopy of critical device internal zones and subcomponents was also performed.

Use of human participants

Consent was obtained for all the fresh blood donations obtained from healthy adult volunteers according to Stony Brook University Committee on Research Involving Human Subjects approved protocol (CORIHS 2012-4427-R4). All experiments were performed in accordance with relevant guidelines and regulations.

Use of experimental animals

Chronic animal studies were conducted according to an IACUC protocol approved by the implantation site at the Sarver Heart Center, University of Arizona (UAZ IACUC Protocol#10–193), additionally approved by Stony Brook University (IACUC 2012-1992-FAR-USDA). The animals received humane care in compliance with the “Principles of Laboratory Animal Care” (NIH Publication No. 85-23, revised 1985). All experiments were performed in accordance with relevant guidelines and regulations.


In silico simulations

The flow is depicted by two perpendicular cross-sections showing the velocity vectors flow field within MINDTE. Distinct regions of high velocities are formed at the gap clearance between the top of the impeller blade and the shroud (in red) and recirculation zones are formed towards the outflow tract where the blood flow is converted from axial to centrifugal (Fig. 4). The flow trajectories of a cluster of 9.8 × 103 platelets that were released upstream (as described) was then tracked along the flow through the VAD, and used to compute for each individual platelet flow trajectory the resultant stress accumulation (SA) value that may drive the platelet beyond its activation threshold. Several highest SA generating platelet trajectories of platelets flowing through the MINDTE impeller region are shown in Fig. 5, indicating entrapped trajectory patterns along the impeller. The large SA ensemble was then collapsed into a probability density function (PDFs) curve, as explained in Methods.

Figure 4
figure 4

Two perpendicular cross-sections of the instantaneous velocity flow field within the MINDTE from a single freeze frame. High flow velocities were observed at the gap clearance between the impeller and the shroud. In addition, recirculation zones formed towards the outflow tract where the flow is converted from axial to centrifugal (inset).

Figure 5
figure 5

Typical platelet flow trajectories around the rotating impeller are shown (inset)- used to compute the Stress Accumulation (SA). The top ten highest SA generating platelet trajectories in the MINDTE impeller regions.

The global PDFs of MINDTE and MAXDTE (Fig. 6) indicated that: (i) the main mode of MINDTE SA distribution populated the lower SA range as compared to that of the MAXDTE main mode; (ii) the MAXDTE main mode had a wider SA distribution (2–30 dyne·s/cm2) as compared to that of the MINDTE main mode (3–15 dyne·s/cm2); (iii) comparing the SA distribution at the tail regions of the PDFs that are riskier, i.e., prone to activate platelets (SA >50 dyne·s/cm2), a secondary and tertiary modes were found for the MAXDTE, indicating that platelets flowing through the MAXDTE have significantly higher probability to be exposed to higher SA levels than the MINDTE. The secondary and tertiary modes of the MAXDTE populated the highest SA range (SA >150 dyne·s/cm2) and extended further – as compared to that of the MINDTE secondary mode (SA range of 130–160 dyne·s/cm2) (Fig. 6).

Figure 6
figure 6

Global PDF results between the MINDTE (blue) and the MAXDTE (red) VADs. The main mode of the MAXDTE appeared to have wider distribution and located at the higher SA range, compared with the MINDTE main mode. Higher platelet distribution was also found in the tail region (i.e., SA >50 dyne·s/cm2) of the MAXDTE, which represented more platelets traverse through would experience higher levels of shear stress (inset).

A closer look at the ROI comparative PDF results at the stator region (Fig. 7) indicates that the PDF main mode of MAXDTE was shifted towards the higher SA range (to the right), with more spread out and higher probability for larger SA values distribution as compared to the MINDTE. At the bearings region, the MAXDTE main mode was similarly shifted towards the higher SA range, with similar SA distribution. At the impeller-shroud gap region, the MAXDTE had a main bimodal distribution, and the higher larger distribution in the 4–20 dyne·s/cm2 SA range; however, it had a shorter higher SA range tail. Distinct differences were found at the impeller blade region, not only in the main mode of the MAXDTE that populated the higher SA range and a far longer tail region distribution. For the lower impeller region, the MAXDTE shared similar distribution distributions as the original MINDTE; however, the entire distribution was offset toward the higher SA range (as shown in Fig. 7).

Figure 7
figure 7

Regions of interest (ROIs) PDF results of MINDTE (blue) and MAXDTE (red) VADs. The MAXDTE PDFs appeared to have wider SA distribution in every ROIs comparing with the MINDTE PDFs. The MAXDTE PDFs’ main modes also located at the higher SA ranges comparing to the MINDTE, indicated the majority of platelets flowing through such regions in MAXDTE would experience higher shear stress accumulation as compared to the platelets flowing through MINDTE.

In vitro platelet activity state (PAS) measurements

The in vitro measurements of the platelet activity over 30 min. recirculation time through both VADs indicated that the thrombogenicity level of the MAXDTE (expressed as the platelet activity rate (PAR) ─ the slope of the PAS measurements over this circulation time) was approx. 5-fold higher than that of the MINDTE (Fig. 8; PAR = 4 × 10−4 m−1 and 8 × 10−5 m−1 for the MAXDTE and the MINDTE, respectively, n = 13, p ≤ 0.03). A comparative PAR calculation based on individual experiments reveals a 3-fold differences between the MAXDTE and MINDTE (Fig. 8, inset). This corroborated the in silico simulation results where the PDFs (the ‘thrombogenic footprint’) clearly showed a much higher probability of platelets being exposed to elevated SA in the case of the MAXDTE– strongly correlating to the measured higher platelet activation rate.

Figure 8
figure 8

Comparison between the platelet activation rate (PAR) between the MINDTE and the MAXDTE VADs. A 5-fold increase was measured in the MAXDTE (n = 13, p ≤ 0.03). (inset) Experiment-based PAR calculation revealed a 3-fold significant increases in MAXDTE.

Chronic in vivo animal experiments

Distinct differences were observed between the animals that were implanted with the MINDTE and the MAXDTE in the chronic animal studies. By onset of post-implantation day 2 the platelet activity state measurements indicated that both animals experienced an acute platelet activation peak, with similar trends of platelet count found in the two animals (Fig. 9, top). The calf implanted with the MINDTE recuperated and survived well throughout the 30 days experiment. However, the calf implanted with the MAXDTE did poorly, resulting in early termination (POD 2) due to a sudden seizure with accompanying incapacitating stroke from thromboembolism, following which the animal was euthanized for humane reasons (Fig. 9, right- top).

Figure 9
figure 9

Chronic in vivo animal experimental results for the MINDTE (left column) and the MAXDTE (right column). The top row shows the platelet activation state and platelet counts for each implanted device. The bottom row shows the post-mortem explanted devices. White thrombus was observed at the entry of the impeller in MINDTE, and red thrombus was observed widely distributed along the MAXDTE impeller and its aft.

Specifically, the platelet activity state (PAS) of the calf implanted with the MINDTE peaked by POD1, likely due to the insult of surgery and peri-operative inflammation, but subsequently declined by POD4 (Fig. 9, left- top; black) with early healing. By POD5 PAS samples indicated a general trend up for the rest of the study. The activity state of platelets appeared to correlate with the white blood count and lactate dehydrogenase (LDH) levels (not shown) but not with the platelet count (Fig. 9, left- top; purple) or fibrinogen level. At the end of the 30 days chronic study, post-mortem gross and microscopic analysis revealed signs of tissue deposition on the inflow cannula and ingrowth lamination of cells on the outflow graft. There were no sign of thrombosis at the anastomosis region or the outflow graft. Enhanced fibrin deposition (hypercoagulability) was observed on the inner ostium of the inflow cannula, as well as in the disassembled VAD components− within the pump (inlet stator and inlet bearing), and on the outflow graft lining. Thrombus accumulation around the inlet bearing was examined histologically and found to consist of a homogenized, acellular eosinophilic (proteinaceous) matrix material (denatured protein) that is hypothesized to have formed as a result of heat and/or rotational force). Phosphotungstic acid-haematoxylin (PTAH) staining for fibrin confirmed the presence of condensed fibrin laminates within the thrombus material. Distinct contained regions of white thrombus were found at the inlet bearing (ball) of the impeller (Fig. 9, left- bottom).

The post mortem examination of the MAXDTE explanted from the animal revealed evident thrombus in the LV adjacent to the inflow cannula and within the inflow port from the left ventricle that was 40% occluded by a red thrombus (Fig. 9, right- bottom). The outflow graft revealed thrombus in the peri-outflow region. Examination of the disassembled VAD components revealed a large red thrombus with evident adherent thrombus at the front bearing (ball) that was widely distributed from there all along the impeller, as well as around the aft region of the MAXDTE impeller, likely resulting from the presence of recirculation and stagnation flow caused by the design modification (Fig. 9, right- bottom). Whitish older thrombus across the rotor was more adherent and extended from just distal to the bearing across the notch down to the aft portion of the impeller. MRI of the brain of the animal revealed large evident thromboemboli in major vessels surrounding the brain with clear evidence of parenchymal infarction.


The MAXDTE, with the partial impeller blades (5 mm gap) was expected to have a significantly higher thrombogenic potential as compared to the MINDTE design. The extensive in silico numerical simulations and analysis of the thrombogenic potential (via the statistical PDF distribution of the stress accumulation along a large ensembles of individual platelet flow trajectories though each device, that provides the ‘thrombogenic footprint’ of the design), clearly showed the marked differences between the two VAD designs. This was confirmed by comparing both the global PDFs of the devices, as well as by a closer examination and comparison in specific regions of interest (ROI). Physical prototype devices representing MINDTE and MAXDTE were provided and tested in recirculation flow loops under typical VAD operating conditions, wherein the platelet activity generated by each was measured and compared. The in vitro results confirmed the numerical simulations that predicted a much higher thrombogenic potential for the MAXDTE. The platelet activity measurements performed in vitro in the actual pump prototypes operating under clinical conditions in circulation flow loops generated approx. 5-fold significantly higher thrombogenicity (in terms of platelet activity) as compared to the MINDTE. This in vitro validation of the DTE methodology was further corroborated by the subsequent in vivo chronic animal studies. The animal implanted with the MINDTE prototype survived well and thrived throughout the 30 day experiments. While it was expected that the animal implanted with the MAXDTE would fare inferiorly, the much higher thrombogenicity levels generated by the device led to catastrophic thromboembolic complications, leading to stroke, seizure and incapacity, with early termination of the study. The post-mortem analyses of the, e.g., MINDTE, showed that white thrombus were formed at the impeller front bearing region (Fig. 9, left- bottom). This pattern closely matched the location where entrapped high SA platelet trajectories were found in the MINDTE simulations (Fig. 5), and is very similar to the thrombus formation patterns observed in explanted HMII VADs9,10,25,40. The red thrombus found throughout the explanted MAXDTE impeller (Fig. 9, right- bottom) matched the presence of stagnation and recirculation zones caused by the intentional design modification, which additionally led to reduction of the pump efficiency. The depositions of white and red thrombus in MINDTE and MAXDTE, respectively, reveals the distinct flow environments of the two devices. Platelets entrapped at the entry of MINDTE impeller exposed to high shear dosage while spinning along with the higher RPM impeller, thus the high shear related white thrombus was formed7. The red thrombus deposited along the MAXDTE impeller revealed that the fluid flow field was majorly disturbed due to the partial removal of impeller blades. Pump efficacy reduced leads to the formation of recirculation zones and stagnation flow, which leads to low-flow associated thrombus, i.e., red thrombus7.


Isolated platelets were used in the present experiments to examine the direct effect of various device associated shear environments on thrombin generation without the subsequent feedback and platelet aggregation. The platelets were diluted to a concentration of 15,000 µL due to the limited available donor blood volume and the relatively large flow loop volume. These limitations may not directly represent the physiological response of whole blood, which consists of other blood cells and plasma proteins, via which the platelet shear-induced activation is amplified. Although flow cytometry is capable of providing detailed information regarding platelet membrane glycoprotein activity, and has been utilized in several VAD studies41,42, here the PAS assay was utilized due to its ability to provide near-real time information on shear-induced bulk platelet thrombin generation. The DTE predictive capability would have been established more rigorously if a larger study, involving more animals, was conducted. However, such a study would necessitate near insurmountable cost and complexity. Furthermore, the present DTE experiments well demonstrate and underscore the predictive capacity of DTE. While it is recognized that DTE is far removed from the “wet” physiological scenario, such a reductionist approach is capable of predicting outcomes directly from in silico to in vivo.

In silico simulations and in vitro experiments have been widely employed during the R and D phases of VAD development. Previous VAD studies have demonstrated the robust capability of using advanced CFD to examine the intra-device flow conditions and shear environment43,44, leading to predictions of hemolysis29,32 or platelet activation22,24. In vitro experiments were mainly employed to evaluate intra-device flow environments in previous studies45, with some groups beginning to utilize in vitro experiments for examining hemolysis caused by minor design modifications46. However, in contrast, DTE is a pioneering concept in that it couples in silico simulations, in vitro studies and in vivo experiments together, developing a predictive methodology to anticipate in vivo trial outcomes numerically.


The DTE methodology demonstrates the use of advanced numerical simulations combined with experimental techniques for quantifying measures that are directly relevant to thrombosis and coagulation markers in devices, that is predictive of how devices may perform in vivo. The veracity of the DTE in silico simulations predictive capability was validated by in vitro recirculation experiments, and further translated and confirmed by in vivo VAD implantation animal studies. The robust capability of this predictive technology demonstrates its utility as a cost-effective pre-clinical MCS thrombo-optimization approach. DTE offers the potential to transform the conventional approach currently utilized for designing and developing MCS devices. Presently numerous, costly, burdensome in vivo animal studies are required during each development cycle. DTE-enhanced design offers the potential of a more direct approach, with a priori enhanced hemocompatibility (Fig. 1), ultimately reducing both the time and cost of device development. Further, this approach may eventually obviate the need for animal experiments, providing a useful adjunct to enhance regulatory science approaches and approval.

Data Availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. The data of design iterations that were performed during this study are available from Abbott Labs, but restrictions apply to the availability of these data which are not publicly available.


  1. FDA. Thoratec HeartMate II LVAS - P060040/S005, (2010).

  2. FDA. HeartWare Ventricular Assist System - P100047, (2012).

  3. FDA. HeartMate 3 Left Ventricular Assist System (LVAS) - P160054/S008, (2018).

  4. Slepian, M. J. et al. Shear-mediated platelet activation in the free flow: Perspectives on the emerging spectrum of cell mechanobiological mechanisms mediating cardiovascular implant thrombosis. J Biomech 50, 20–25, (2017).

    Article  PubMed  Google Scholar 

  5. Baumann Kreuziger, L. M. Management of anticoagulation and antiplatelet therapy in patients with left ventricular assist devices. J Thromb Thrombolysis 39, 337–344, (2015).

    CAS  Article  PubMed  Google Scholar 

  6. Baumann Kreuziger, L. M., Kim, B. & Wieselthaler, G. M. Antithrombotic therapy for left ventricular assist devices in adults: a systematic review. J Thromb Haemost 13, 946–955, (2015).

    CAS  Article  PubMed  Google Scholar 

  7. Blitz, A. Pump thrombosis-A riddle wrapped in a mystery inside an enigma. Ann Cardiothorac Surg 3, 450–471, (2014).

    Article  PubMed  PubMed Central  Google Scholar 

  8. Kirklin, J. K. et al. Pump thrombosis in the Thoratec HeartMate II device: An update analysis of the INTERMACS Registry. J Heart Lung Transplant 34, 1515–1526, (2015).

    Article  PubMed  Google Scholar 

  9. Meyer, A. L. et al. Thrombus formation in a HeartMate II left ventricular assist device. J Thorac Cardiovasc Surg 135, 203–204, (2008).

    Article  PubMed  Google Scholar 

  10. Mokadam, N. A., Andrus, S. & Ungerleider, A. Thrombus formation in a HeartMate II. Eur J Cardiothorac Surg 39, 414, (2011).

    Article  PubMed  Google Scholar 

  11. Najib, M. Q., Wong, R. K., Pierce, C. N., DeValeria, P. A. & Chaliki, H. P. An unusual presentation of left ventricular assist device thrombus. Eur Heart J Cardiovasc Imaging 13, 532, (2012).

    Article  PubMed  Google Scholar 

  12. Samson, R. et al. Device Thrombosis During Destination Therapy. Am J Med Sci 351, 441–446, (2016).

    Article  PubMed  Google Scholar 

  13. Smedira, N. G. et al. Current risks of HeartMate II pump thrombosis: Non-parametric analysis of Interagency Registry for Mechanically Assisted Circulatory Support data. J Heart Lung Transplant 34, 1527–1534, (2015).

    Article  PubMed  Google Scholar 

  14. Starling, R. C. et al. Unexpected abrupt increase in left ventricular assist device thrombosis. N Engl J Med 370, 33–40, (2014).

    CAS  Article  PubMed  Google Scholar 

  15. Valerio, L. et al. Routine clinical anti-platelet agents have limited efficacy in modulating hypershear-mediated platelet activation associated with mechanical circulatory support. Thromb Res 163, 162–171, (2018).

    CAS  Article  PubMed  Google Scholar 

  16. Valerio, L. et al. Aspirin has limited ability to modulate shear-mediated platelet activation associated with elevated shear stress of ventricular assist devices. Thromb Res 140, 110–117, (2016).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  17. Amer, S., Shah, P. & Hassan, S. Gastrointestinal bleeding with continuous-flow left ventricular assist devices. Clin J Gastroenterol 8, 63–67, (2015).

    Article  PubMed  Google Scholar 

  18. Draper, K. V., Huang, R. J. & Gerson, L. B. GI bleeding in patients with continuous-flow left ventricular assist devices: a systematic review and meta-analysis. Gastrointestinal endoscopy 80, 435–446 e431, (2014).

    ADS  Article  PubMed  Google Scholar 

  19. Eckman, P. M. & John, R. Bleeding and thrombosis in patients with continuous-flow ventricular assist devices. Circulation 125, 3038–3047, (2012).

    Article  PubMed  Google Scholar 

  20. Geisen, U. et al. Non-surgical bleeding in patients with ventricular assist devices could be explained by acquired von Willebrand disease. Eur J Cardiothorac Surg 33, 679–684, (2008).

    Article  PubMed  Google Scholar 

  21. Alemu, Y. & Bluestein, D. Flow-induced platelet activation and damage accumulation in a mechanical heart valve: numerical studies. Artif Organs 31, 677–688, (2007).

    Article  PubMed  Google Scholar 

  22. Girdhar, G. et al. Device thrombogenicity emulation: a novel method for optimizing mechanical circulatory support device thromboresistance. PLoS One 7, e32463, (2012).

    ADS  CAS  Article  PubMed  PubMed Central  Google Scholar 

  23. Xenos, M. et al. Device Thrombogenicity Emulator (DTE)–design optimization methodology for cardiovascular devices: a study in two bileaflet MHV designs. J Biomech 43, 2400–2409, (2010).

    Article  PubMed  PubMed Central  Google Scholar 

  24. Chiu, W. C. et al. Thromboresistance comparison of the HeartMate II ventricular assist device with the device thrombogenicity emulation- optimized HeartAssist 5 VAD. J Biomech Eng 136, 021014, (2014).

    Article  PubMed  Google Scholar 

  25. Chiu, W. C., Slepian, M. J. & Bluestein, D. Thrombus formation patterns in the HeartMate II ventricular assist device: clinical observations can be predicted by numerical simulations. ASAIO J 60, 237–240, (2014).

    Article  PubMed  PubMed Central  Google Scholar 

  26. Sheriff, J. et al. Comparative efficacy of in vitro and in vivo metabolized aspirin in the DeBakey ventricular assist device. J Thromb Thrombolysis 37, 499–506, (2014).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  27. ANSYS. ANSYS Fluent Theory Guide. (2015).

  28. Chiu, W. C. et al. Ventricular Assist Device Implantation Configurations Impact Overall Mechanical Circulatory Support System Thrombogenic Potential. ASAIO J 63, 285–292, (2017).

    Article  PubMed  PubMed Central  Google Scholar 

  29. Fraser, K. H., Zhang, T., Taskin, M. E., Griffith, B. P. & Wu, Z. J. A quantitative comparison of mechanical blood damage parameters in rotary ventricular assist devices: shear stress, exposure time and hemolysis index. J Biomech Eng 134, 081002, (2012).

    Article  PubMed  Google Scholar 

  30. Marom, G. & Bluestein, D. Lagrangian methods for blood damage estimation in cardiovascular devices–How numerical implementation affects the results. Expert Rev Med Devices 13, 113–122, (2016).

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  31. Bluestein, D., Rambod, E. & Gharib, M. Vortex shedding as a mechanism for free emboli formation in mechanical heart valves. J Biomech Eng 122, 125–134 (2000).

    CAS  Article  Google Scholar 

  32. Apel, J., Paul, R., Klaus, S., Siess, T. & Reul, H. Assessment of hemolysis related quantities in a microaxial blood pump by computational fluid dynamics. Artif Organs 25, 341–347 (2001).

    CAS  Article  Google Scholar 

  33. Jesty, J. & Bluestein, D. Acetylated prothrombin as a substrate in the measurement of the procoagulant activity of platelets: elimination of the feedback activation of platelets by thrombin. Anal Biochem 272, 64–70, (1999).

    CAS  Article  PubMed  Google Scholar 

  34. Fumoto, H. et al. In Vivo Acute Performance of the Cleveland Clinic Self-regulating, Continuous-flow Total Artificial Heart. J Heart Lung Transplant, S1053-2498(09)00427-6 [pii] (2009).

  35. Murai, N. et al. Physiological adaptation to a nonpulsatile biventricular assist system. ASAIO J 49, 345–348 (2003).

    Article  Google Scholar 

  36. Fossum, T. W. et al. Chronic survival of calves implanted with the DeBakey ventricular assist device. Artif Organs 23, 802–806 (1999).

    CAS  Article  Google Scholar 

  37. Fossum, T. W. et al. Complications common to ventricular assist device support are rare with 90 days of DeBakey VAD support in calves. ASAIO J 47, 288–292 (2001).

    CAS  Article  Google Scholar 

  38. Jesty, J., Yin, W., Perrotta, P. & Bluestein, D. Platelet activation in a circulating flow loop: Combined effects of shear stress and exposure time. Platelets 14, 143–149, (2003).

    CAS  Article  PubMed  Google Scholar 

  39. Yin, W. et al. Thrombogenic performance of a st. Jude bileaflet mechanical heart valve in a sheep model. ASAIO J 52, 28–33, (2006).

    Article  Google Scholar 

  40. Capoccia, M., Bowles, C. T., Sabashnikov, A. & Simon, A. Recurrent Early Thrombus Formation in HeartMate II Left Ventricular Assist Device. J Investig Med High Impact Case Rep 1, 2324709613490676, (2013).

    Article  PubMed  PubMed Central  Google Scholar 

  41. Slaughter, M. S. et al. Platelet activation in heart failure patients supported by the HeartMate II ventricular assist device. Int J Artif Organs 34, 461–468, (2011).

    CAS  Article  PubMed  Google Scholar 

  42. Houel, R. et al. Platelet activation and aggregation profile in prolonged external ventricular support. J Thorac Cardiovasc Surg 128, 197–202, (2004).

    Article  PubMed  Google Scholar 

  43. Zhang, J., Zhang, P., Fraser, K. H., Griffith, B. P. & Wu, Z. J. Comparison and experimental validation of fluid dynamic numerical models for a clinical ventricular assist device. Artif Organs 37, 380–389, (2013).

    Article  PubMed  PubMed Central  Google Scholar 

  44. Medvitz, R. B., Reddy, V., Deutsch, S., Manning, K. B. & Paterson, E. G. Validation of a CFD methodology for positive displacement LVAD analysis using PIV data. J Biomech Eng 131, 111009, (2009).

    Article  PubMed  PubMed Central  Google Scholar 

  45. Tree, M. et al. In Vitro Examination of the HeartWare CircuLite Ventricular Assist Device in the Fontan Connection. ASAIO J 63, 482–489, (2017).

    Article  PubMed  Google Scholar 

  46. Li, D. et al. Lactic Dehydrogenase in the In Vitro Evaluation of Hemolytic Properties of Ventricular Assist Device. Artif Organs 41, E274–E284, (2017).

    CAS  Article  PubMed  Google Scholar 

Download references


This project was supported by grants from the National Institute of Biomedical Imaging and Bioengineering: Quantum Award Phase I R01, EB008004 (DB) and Quantum Award: Implementation Phase II- 1U01EB012487 (DB).

Author information

Authors and Affiliations



Wei Che Chiu performed the numerical simulations, the in vitro experiments and their analysis, and wrote the initial draft. Phat L. Tran, and Zain Khalpey performed the in vivo experiments and analyzed their results. Eric Lee and Yi-Ren Woo performed design iterations and manufactured the VAD prototypes. Marvin J. Slepian coordinated, participated in and oversaw the in vivo experiments, interpreted their results, and helped in writing the manuscript. Danny Bluestein oversaw and coordinated all aspects of the project, led and participated in all aspects of the numerical simulations and the in vitro experiments, and wrote the manuscript.

Corresponding author

Correspondence to Danny Bluestein.

Ethics declarations

Competing Interests

Authors Eric Lee and Yi-Ren Woo were senior research engineers employed by Thoratec Corporation at the time of this research project (since acquired by Abbott Labs). Authors Wei Che Chiu, Phat L. Tran, Zain Khalpey, Marvin J. Slepian and Danny Bluestein have no competing interests.

Additional information

Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Chiu, W.C., Tran, P.L., Khalpey, Z. et al. Device Thrombogenicity Emulation: An In Silico Predictor of In Vitro and In Vivo Ventricular Assist Device Thrombogenicity. Sci Rep 9, 2946 (2019).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI:

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.


Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing