Multi‐constituent simulation of thrombus formation at LVAD inlet cannula connection: Importance of Virchow’s triad

Abstract

As pump thrombosis is reduced in current‐generation ventricular assist devices (VAD), adverse events such as bleeding or stroke remain at unacceptable rates. Thrombosis around the VAD inlet cannula (IC) has been highlighted as a possible source of stroke events. Recent computational fluid dynamics (CFD) studies have attempted to characterize the thrombosis risk of different IC‐ventricle configurations. However, purely CFD simulations relate thrombosis risk to ad hoc criteria based on flow characteristics, with little consideration of biochemical factors. This study investigates the genesis of IC thrombosis including two elements of the Virchow's triad: endothelial injury and hypercoagulability. To this end a multi‐scale thrombosis simulation that includes platelet activity and coagulation reactions was performed. Our results show significant thrombin formation in stagnation regions (|u| < 0.005 m/s) close to the IC wall. In addition, high shear‐mediated platelet activation was observed over the leading‐edge tip of the cannula. The current study reveals the importance of biochemical factors to the genesis of thrombosis at the ventricular‐cannula junction in a perioperative state. This study is a first step toward the long‐term objective of including clinically relevant pharmacological kinetics such as heparin or aspirin in simulations of inflow cannula thrombosis.

Uncertainty quantification of a thrombosis model considering the clotting assay PFA‐100®

“…The uncertainties related to the microfluidic assay geometry were not considered, as they are not as important as in physiological vessel hemodynamics. Due to the computational cost of performing an analysis considering all the model parameters (64 core‐hours per simulation), we have selected a subset of 15 influential parameters based on our experience gained from previous thrombosis simulations in both macroscopic and microscopic configurations 33,41,48,49 . The parameters are listed in Table 2 with their definition and distribution values.…”

“…Due to the computational cost of performing an analysis considering all the model parameters (64 core-hours per simulation), we have selected a subset of 15 influential parameters based on our experience gained from previous thrombosis simulations in both macroscopic and microscopic configurations. 33,41,48,49 The parameters are listed in Table 2 with their definition and distribution values. The literature does not show evidence of particular type of uncertainty distributions for these parameters.…”

“…Third, a first triage of parameters was perform selecting only 15 parameters to perform the sensitivity study. Even though this selection was informed by our previous thrombosis studies 33,41,48,49 it remains somewhat arbitrary. Furthermore, the parameters excluded in this first triage step may have non-negligible parameter interactions with the 15 parameters selected.…”

Effects of the atrium on intraventricular flow patterns during mechanical circulatory support

“…Two established measures were used in this study to evaluate the risk of thromboembolism, due to the fact that platelet activation and consequently thrombosis is a multi-factorial phenomenon. The first one is related to regions with low-velocity 10 , 29 – 31 and wall shear stress (WSS) 32 – 34 and highlights the local regions with a high risk of platelet deposition and the near-wall thrombus growth. While the second one is based on the residence time (RT) and shear stresses history (SSH) of each particle 14 , 18 , 35 and evaluate the risk of platelet activation within the whole LV.…”

Toward modeling thrombosis and thromboembolism in laminar and turbulent flow regimes

“…The need to better understand the processes of T&TE in blood‐contacting devices makes in‐silico models attractive, as a much richer understanding of the localized environment is possible including stresses exerted on a blood clot, local concentrations of procoagulant species, and high‐resolution flow field data, though challenges remain including a disconnect between physiological and numerical timescales, experimental validation and incorporating pathologies among others 7 . Modern in‐silico T&TE models are increasingly computationally affordable, and are being used in device‐specific simulations to inform thrombus risk, 8 to investigate thrombolysis in strokes, 9 and to predict venous thrombosis 10 . Fogelson and Guy 11 proposed a single‐scale model based on conservation equations for resting and activated platelets, a generic activator chemical taking the place of the coagulation cascade, and viscoelastic constitutive equations for a cohesive‐link tensor and a cohesive‐link density.…”