The risk of intraventricular thrombosis due to changes in inflow cannula cannulation angles implanted at the left ventricular apex was investigated. Based on the modelled scenario, this study determined whether the common cannula alignment towards the mitral valve is best practice and evaluated potential consequences when deviations occurred. We used a numerical approach going beyond the existing computational models considering ventricular wall movement, a mitral valve geometry and the surrounding cardiovascular system by means of a lumped parameter network. Results based on the blood residence time, rate of left ventricular washout, kinetic energy and pulsatility indices elucidated that inflow cannula alignment towards the mitral valve resulted in potentially more favourable flow dynamics than severe angulation towards either the septum or free wall.
From clinical observations of acute angles between the HeartMate II pump body and inflow cannula, it was difficult to determine the exact position of the inflow cannula relative to internal left ventricular features from the presented literature [10, 14, 19]. Nevertheless, this study assumed that an acute inflow cannula to pump angle might have the highest probability of being directed towards the septum. The increased risk of pump thrombosis due to this angulation could be correlated with the 25 s case, whereby the total blood residence time was 0.8% higher than inflow cannula angulation towards the mitral valve. However, angulation towards the free wall was found to be worse than septal alignment as blood residence times were up to 5% longer than mitral valve alignment. Considering the small differences, the clinical observation of increased pump thrombosis may not be solely attributed to the inflow cannula angle regarding the total blood residence time; however, it may have a cumulative effect with other factors. These include the biochemical elements of thrombus formation such as platelet activation by tissue factor and the role of the coagulation cascade [20]. Further aspects are the cannula design and insertion length as well as clinically relevant parameters like excessive LVAD speed, LV size and lack of aortic valve opening [16].
Longer blood residence times were consistently observed around the left ventricular outflow tract, but with cannula angulation towards the free wall, the LV anterior and posterior basal regions had increased volumes of prolonged residence times, highlighting the importance of assessing the entire left ventricular volume and not only in a single plane or section [8, 21]. Despite arguably small differences in blood residence times (~ 5%) and the rate of washout (~ 1%), it provided an interesting perception of the effects of cannulation angles. Even the slightest improvement to intraventricular flow could be advantageous considering that small thrombi may be conducive for the formation of larger structures due to a cumulative effect [3, 22].
Orientating the inflow cannula towards the free wall may increase the probability of inducing shorter flow paths between the left atrium and inflow cannula tip. The idea to increase the flow path between the mitral valve and inflow cannula tip, to preserve the natural clockwise vortex, has the potential to reduce the risk of thrombosis [6, 7]. Even though alignment towards the septum was more favourable than the free wall, if the inflow cannula long axis intersected the septum, then lower pulsatility indices were observed at the left ventricular outflow tract along with higher blood residence times. Reduced pulsatility indices for when the inflow cannula is angled towards the septum has been previously reported in a PIV study [23]. Based on the results, it is recommended to avoid alignment of the inflow cannula towards the septum, which would also avoid inflow cannula occlusion. The maximum angulation limit towards the septum could potentially be parallel to the septal wall, which may also prevent septal shift at higher pump speeds and reduce the likeliness of left ventricular suction, similar to suggestions by Chivukula et al. despite modelling differences [16].
Chivukula et al. indicated that when the inflow cannula was angled closer to alignment with the LV long axis, flow from the mitral valve to the inflow cannula was more direct [16]. However, this may not be typical for patients with dilated cardiomyopathy: the mitral jet is often directed more towards the free wall [24]. Furthermore, it was speculated that the direct flow between the mitral valve and inflow cannula inlet was favourable, contrary to the recommendations presented here. Due to the evaluation method of injected particles using a Lagrangian approach, the direct flow from the mitral valve to the inflow cannula does not entirely explain the risk of intraventricular thrombosis. Perhaps due to the smaller simulated left ventricular geometry, the effects may not be as pronounced in capturing recirculation zones, hence similar conclusions.
Attenuation of vortex structures due to cannulation orientation has similarly been reported [5]. The previous in vitro study suggested that when an inflow cannula was aligned towards the left ventricular outflow tract, more distinct vortices were formed, corroborating with the presented results in our study. Comparing the velocity vectors between 25 s and 30f in Fig. 1 at end-systole, a large uniform clockwise fluid rotation was present when the inflow cannula was angled towards the septum and mitral valve. In contrast, when the inflow cannula was angled towards the free wall, multiple smaller vortices were present.
The pulsatility index contour plot (Fig. 5) indicated that 25 s, 20 s and 30f had low indices around the left ventricular outflow tract, which can be related to the blood residence time visualisation in Fig. 2. Interestingly, the pulsatility index contour plot of the mitral jet can also be associated with the rate of ventricular washout (Fig. 3): the high pulsatility regions due to the mitral jet can be seen to align with the inflow cannula, especially with angulation towards the free wall (20f and 30f); therefore, potentially explaining the fast but detrimental clearance of incoming blood. The calculated pulsatility indices correspond with the results obtained in the in vitro study [3] with values between 0 and 6 throughout the left ventricle, values below 1 at the left ventricular outflow tract and values > 5 in most parts of the left ventricle. It has to be noted, that pulsatility index is a spatially resolved value and, if the interest lies in obtaining a spatially averaged parameter, the blood residence time should be used instead.
The most significant finding in this study was in agreement with Laumen et al. [5]. There, it was suggested that the ideal inflow cannula placement would be parallel to the septal wall. It was thought that an inflow cannula position pointed towards the septal wall would result in a large uniform vortex, which was found in the presented study. Nevertheless, the alignment of inflow cannulae is not solely related to cannula angulation at the apex, but also the site of cannulation and patient-specific left ventricular geometry.
The results elucidated the importance of cannulation angles, emphasising the significance to minimise early clearance of incoming blood during diastole. However, consideration of endocardial contact and suction should not be overlooked. It is expected that no ideal inflow cannula angle, length or site exists for all patients, but instead will be dependent on the patient condition, such as the degree of left ventricular dilation.
Limitations
Limitations of the model have been previously described, which included the assumption of smooth endocardial walls; ventricular movement in a balloon-like manner based on reduced twist [25]; and neglected internal left ventricular features [26]. Laminar flow was assumed, however the flow might be in the transitional/turbulent regimen, necessitating turbulence modelling. The effect of this simplification on the thrombus risk metrics should be evaluated. The assumed heart rate of 60 bpm is low for VAD patients and may not represent typical patients and only one patient-specific geometry was investigated. Thus, it must be remembered that this paper only presents one potential scenario and the results must be interpreted with caution. All the evaluated metrics indirectly predict the risk of thrombosis and are only an indicator of blood stagnation. A quantifiable relation of these haemodynamic parameters (blood residence time, washout, pulsatility index) towards thrombus formation if difficult to determine as spatially resolved experimental data on thrombus formation in the LV during LVAD support is scarce to non-existent. Each metric has its advantages and disadvantages. Blood residence time, ventricular washout and kinetic energy are spatially averaged metrics allowing for a quick comparison, however neglecting differences within the LV. On the other hand, pulsatility index and the ventricular flow field can be resolved spatially (and if required temporally), however a quantitative comparison for several cases might become difficult. Thus, rather a combination of these metrics should be considered when interpreting the results and we do not believe that there is one single value that can serve as a surrogate marker for flow-induced thrombus formation.
With the awareness of limitations intrinsic to numerical studies, the direct clinical significance is difficult to predict. However, further insights provided into intraventricular flow characteristics due to changes in cannulation angles may provide additional aspects of consideration when designing and implanting LVADs. Current clinical principles for inflow cannula alignment is to angle the inflow cannula parallel to the septum and towards the left ventricular centre, which is in agreement with the suggestions presented in this study [27]. However, the presented findings may be limited to the simulated patient condition; therefore, other haemodynamic and spatial combinations should be pursued. Since no aortic valve opening was simulated and atypical to clinical preference, it is hypothesised that superior ventricular washout could be obtained in the basal region if aortic valve opening is promoted; however, parallel septal alignment is expected to be superior based on the avoidance of early blood clearance and higher kinetic energies at the LV apex.
In order to leverage clinical significance, future applications of this numerical model should focus on a higher number of patient-specific geometries. One case study, as shown within this work, is not able to quantify the role of flow alterations and provide statistically significant results with a clinical implication. Investigations of many geometries require much higher computational costs, a problem that needs to be tackled as the next step. We believe that reduction of model complexity under simultaneous considerations of the thrombus risk metrics presented herein, should be performed. In other words, one should gradually simplify the model and compare the changes in the outcomes such as pulsatility index, blood residence time and intraventricular vortices. Possible simplifications could be steady-state simulations at some time points instead simulations of the full cardiac cycle, a rigid wall assumption instead of a fluid–structure interaction simulation or the disregard of the lumped parameter network model.
The results from our study can be used as a reference case to see, if model simplifications violate the prediction of numerical models focussing on thrombus risk in the left ventricle during LVAD support.