Implantable rotary left ventricular assist devices (LVADs) provide mechanical circulatory support to patients with end-stage heart failure.1 Right ventricular failure can develop following LVAD support, with the incidence ranging up to 40%.2,3 Treatment for right ventricular failure may include pharmaceutical therapies or temporary extracorporeal support devices. However, for instances where long-term support is required, a right ventricular assist device (RVAD) may be indicated. Currently, there are no durable implantable RVADs that are commercially available, and instead LVADs have been used “off-label” as an RVAD to provide simultaneous biventricular (BiVAD) support.4 To compensate for the lower-pressure pulmonary circulation during RVAD use, the LVAD is typically modified by either reducing pump speed or by banding the outflow graft.4–6
Given the severity of heart failure associated with VAD support, end-organ perfusion is primarily provided by VAD outflow, making VAD outflow an important metric in managing patient treatment.7 The mean VAD outflow can be used to set patient perfusion levels, while the outflow waveforms can be used to detect aortic valve opening, cardiac arrhythmia, ventricular suction, and estimate ventricular contractility.8–10 Managing VAD outflow is particularly important in BiVAD patients due to the increased difficulty in maintaining circulatory balance as exacerbated by the limited capacity of the body to regulate blood flow under such severe pathophysiology.11 Ideally, VAD outflow should be measured using flow sensors, and although VAD flow sensors are commercially available (though not presently FDA-approved) via the aVAD and HeartAssist 5 (ReliantHeart Inc., Houston, TX), their clinical accuracy is unclear and these devices have not seen widespread clinical adoption.12
Instead, LVAD flow estimation remains the most common method of flow interrogation in VAD patients. Devices such as the HeartMate II, HeartMate 3 (Abbott Laboratories, Abbott Park, IL), and HeartWare HVAD (Medtronic Inc., Dublin, Ireland) all use intrinsic pump parameters (operating power and VAD speed) and extrinsic parameters (patient hematocrit [HCT]) to estimate the VAD outflow using lookup tables and equations programmed into the VAD controller.13,14 Previous studies have evaluated the accuracy of some clinically available LVAD flow estimators and quantified the level of accuracy over a wide variety of operating scenarios.13,15,16 However, the use of an LVAD “off-label” as for BiVAD support has now been performed in numerous centers,17–19 and the accuracy of the flow estimator for this operating mode has not been extensively characterized.
This study aims to quantify the accuracy of the HeartWare HVAD flow estimator when used “off-label” as an RVAD. In clinical and animal studies, it can be difficult to independently and consistently control many of the variables contributing to HVAD flow estimation. Therefore, for direct comparison between operating methods, the HVAD flow estimator was evaluated in a mock circulatory loop (MCL) using the HVAD in standard LVAD support as well as using dual HVADs for BiVAD support with RVAD reduced speed and banded outflow graft modifications. Each support method was evaluated under multiple variations in MCL parameters, pump speed, and fluid viscosity. Accuracy of the HVAD flow estimator was determined by comparing the mean and instantaneous flow waveforms between estimated and measured VAD outflow over a wide variety of simulated operating (speed and HCT/viscosity) and cardiovascular parameters (heart rate, vascular resistance, and volume).
Mock Circulatory Loop
All experiments were conducted using an MCL, presented in detail in previous works.20 Briefly, the MCL is a five-element Windkessel model of the pulmonary and systemic circulations. Four independent polyvinyl chloride Windkessel cylinders simulate the lumped arterial and venous compliance, while four smaller polyvinyl chloride cylinders simulate the atria and ventricles. Ventricular contraction occurs via electro-pneumatic regulators (ITV1030-31N2BL5-X88; SMC Pneumatics, Tokyo, Japan), which adjust pneumatic force based on measured end-diastolic volume, thereby mimicking the Frank–Starling mechanism.21 Mechanical umbrella valves ensure forward flow through the loop by serving as atrioventricular and semilunar valves (UM350.001-151.01; Minivalve Inc., Oldenzaal, the Netherlands). Lumped systemic and pulmonary vascular resistances (SVR and PVR) are autoregulated to a user-defined value by pneumatically controlled socket valves (VMP025.03X.71, AKO, Alb Klein Ohio LLC, Plain City, OH). The MCL working fluid was a water and glycerol solution mixed to different viscosities at 23°C using a cone and plate viscometer (DV2T with CPA-40Z attachment; Brookfield Engineering, Middleborough, MA) representing different HCTs, as described in the Experiments section of this article. Initial heart rate was 80 bpm, and systolic duty was maintained at 35% for all experiments.
To evaluate the accuracy of the estimator for LVAD support, severe left heart failure was simulated as mean aortic pressure of 64 mm Hg, cardiac output of 3.4 L/min, left atrial pressure of 15.6 mm Hg, and right atrial pressure of 7.5 mm Hg. The HVAD was then added to support the ventricle with the inflow cannulation site at the left ventricle and the outflow cannulation site at the aorta. A biventricular heart failure condition was also simulated consisting of a mean aortic pressure of 61 mm Hg, cardiac output of 3.1 L/min, left atrial pressure of 8.7 mm Hg, and right atrial pressure of 10.6 mm Hg.22 Biventricular heart failure was then supported by an LVAD (HVAD) set to a constant speed of 3,000 rpm (chosen as the middle of the operating range), and RVAD (HVAD) in two configurations—RVAD operated at reduced speed and RVAD operated at LVAD speeds with the outflow banded to create a 20-mm long × 5-mm diameter flow path using a 3D-printed inline restriction. This resulted in two BiVAD support methods; BiVAD with reduced speed RVAD (BiVAD-RS) and BiVAD with a banded outflow RVAD (BiVAD-B) (Figure 1).
The MCL control and data acquisition were achieved with a dSPACE 1202 MicroLabBox (dSPACE GmbH, Paderborn, Germany) at a rate of 2 kHz and downsampled to 200 Hz for postprocessing. A Transonic clamp-on ultrasonic flow sensor measured LVAD flow during LVAD support and RVAD flow during BiVAD support (10PXL probe; TS410 meter, Transonic Inc., Ithaca, NY). Meanwhile, em-tec clamp-on flow sensors measured systemic and pulmonary flow rates (em-tec BioProTT 3/8″ × 1/8″ for systemic flow; 1″ × 1/8″ for pulmonary flow; SonoTT DigiFlow board; em-tec GmbH; Finning; Germany). The flow sensors were calibrated to each test viscosity using a viscosity insensitive paddle flowmeter (VCB-4.5-B-ES-LL, AW-LAKE Company, WI) to within ± 0.2 L/min. All pressures were measured by TruWave disposable pressure sensors (Edwards Lifesciences, CA). The HVAD flow data were directly recorded to dSPACE at approximately 20 Hz using a customized communications protocol (see supplemental material for further details http://links.lww.com/ASAIO/A527).
Each support method (LVAD, BiVAD-RS, and BiVAD-B) was evaluated under a variety of independent perturbations to MCL parameters to determine their effect on HVAD flow estimator accuracy, as outlined in Table 1. For LVAD and BiVAD-B support, the pattern of experimentation was to initially set the MCL parameters (SVR, PVR, heart rate—HR, or central venous pressure—CVP, affecting MCL ventricular contractility) to their default values (Table 1). Next, the target pump was initialized (LVAD for LVAD support or RVAD for BiVAD support) to 2,000 rpm and 60 seconds of data was recorded. The speed of the target pump was then continually increased by 500 rpm every 60 seconds up to 4,000 rpm or until ventricular suction occurred in either ventricle (ventricular volume = 0 ml). Following this, one of the MCL variables was changed, and the experiments were repeated, recording data for all VAD speeds and for all changes in MCL parameters. The same pattern of experimentation was used for the BiVAD-RS support configuration; however, due to the reduced speed operation, the pump was initially started at 2,000 rpm, the same initial speed as the other support methods, and increased in 200 rpm increments up to 2,800 rpm or until suction was detected. The experimental protocol was then repeated for each support configuration at four different viscosities (2.1, 2.4, 3.6, and 4.9 cP, ± 0.1 cP) representing typical HCTs (20%, 30%, 40%, and 50%) as described in previous studies to determine the effect of viscosity on estimator accuracy.13,16,23 Each experimental data point was acquired only once due to the highly repeatable nature of the MCL and the low variation found experimentally in the HVAD flow estimator.20
Table 1. -
Variations in Mock Circulatory Loop Parameters for Evaluating HVAD Flow Estimator Accuracy
||← 100 →
||← 1,300 →
||← 80 →
|CVP (mm Hg)
||← 10 →
CVP, central venous pressure; HR, heart rate; HVAD, HeartWare left ventricular assist device; PVR, pulmonary vascular resistance; SVR, systemic vascular resistance.
All data were analyzed in MATLAB 2018b (The MathWorks, Natick, MA). The measured and estimated VAD flow waveforms were recorded directly to dSPACE using the custom communications protocol. For each data point (pump operating at a given speed for a given condition), the flow rate was averaged over five heartbeats to produce a mean value. The mean flow error was then determined by subtracting steady-state measured VAD flow from steady-state estimated VAD flow.
Meanwhile, features of instantaneous flow were calculated by a function, which analyzed the last 20 seconds (the steady state) of a data point and identified the peak and trough of the VAD flow for each heartbeat. The identified peaks and troughs were then averaged to give the average measured and estimated peaks and troughs for each speed and condition combination. The error was calculated for each average peak and trough by subtracting the average measured peak and trough values from the average estimated peak and trough values. Finally, average flow pulsatility error was calculated by subtracting the averaged peak error from the averaged trough error.
Kolmogorov–Smirnov tests were used to test for normality, and all data were found not to be normally distributed (p < 0.01). MATLAB’s fitglm performed multiple regression to determine the effect of SVR, PVR, HR, CVP, and HCT (as simulated by viscosity) on the accuracy of the HVAD flow estimator for each support method (α = 0.05). A Kruskal–Wallis test was used to determine if there was a significant difference in estimator error across each of the three support methods for all data points (via MATLAB’s kruskal–wallis and multcompare functions). Linear regression was performed on the pulsatility error to determine the correlation between the instantaneous measured and estimated flow pulsatility.
Mean Flow Error
Unless otherwise specified, data are presented as median ± interquartile range (IQR). N = 540 data points were collected, of which N = 442 remained after removing suction event data points (N = 153 for LVAD, N = 150 for BiVAD-RS, and N = 139 for BiVAD-B). The analysis produced descriptive statistics for each support method at each HCT (Figure 2). General observations show that LVAD support at HCT 20% had the widest error range −1.2 to 1 L/min (0.4 ± 0.3 L/min), while LVAD support at HCT 40% had the lowest error range −0.4 L/min to −0.1 L/min (−0.2 ± 0.1 L/min). The IQR of each support method was lower at higher HCT values (40% and 50%), compared with their respective lower HCT values (20% and 30%), resulting in a less variable error. Across all HCTs, LVAD demonstrated the largest error range and IQR (−1.2 to 1 L/min and 0.1 ± 0.3 L/min), with BiVAD-RS demonstrating the second-highest range (−1 to 0.6 L/min) and IQR (−0.1 ± 0.2 L/min), and finally, BiVAD-B had the lowest range (−0.6 to 0.8 L/min, IQR = 0 ± 0.2 L/min). Further analysis of the correlation between measured and estimated VAD flow for each of the support methods demonstrated that LVAD had the highest correlation coefficient (r2 = 0.96), followed by BiVAD-B (r2 = 0.95) and BiVAD-RS (r2 = 0.89) (Figure 3). Statistical testing showed no significant difference in error between LVAD and BiVAD-B (p = 0.46), but significant differences between LVAD and BiVAD-RS (p < 0.001) and between BiVAD-RS and BiVAD-B (p = 0.001). This suggests that the mean HVAD flow error is statistically similar when operated as an LVAD or banded RVAD, but different when operated at reduced speed.
Multiple regression analysis showed that for both LVAD and BiVAD-B support, HCT and VAD speed were significant predictors of error (p < 0.001) (Table 2). Meanwhile, for BiVAD-RS, HCT and PVR were significant predictors of error (p < 0.001), while speed was not (p = 0.08), likely due to the low range of VAD speeds tested for BiVAD-RS (only 2,000–2,800 rpm). Despite the limited range of speeds tested during BiVAD-RS, the flow ranges (4.2–8.1 L/min) were similar to those of LVAD (1.2–8.1 L/min) and BiVAD-B (3.2–9.0 L/min) support, with a notable difference that LVAD support produced much lower flow rates during the high SVR experiments. It is also interesting to note that SVR was nearly a significant factor in BiVAD-B error (p = 0.08) and that VAD speed was nearly significant in error prediction during BiVAD-RS (p = 0.08).
Table 2. -
Significant Predictors of Mean Flow Error for LVAD, BiVAD-RS, and BiVAD-B
BiVAD-B, banded biventricular assist device; BiVAD-RS, reduced speed biventricular assist device; CVP, central venous pressure; HCT, hematocrit; HR, heart rate; LVAD, left ventricular assist device; SVR and PVR, systemic and pulmonary vascular resistance.
A strong correlation was observed between measured and estimated LVAD flow pulsatility (r2 = 0.98) with median error and IQR of 0.7 ± 0.3 L/min and 0.0–2.7 L/min, respectively (Figures 4 and 5). In comparison, although a similar median error was observed for LVAD support with BiVAD-RS and BiVAD-B (both 0.7 L/min), there was a smaller IQR (±0.3 and ±0.2 L/min, respectively) and correlation (r2 = 0.85 and 0.60, respectively). The smaller IQR range may be explained by the smaller variations of pressure head with right ventricular support.
This study evaluated the HVAD flow estimator under different support methods at different HCTs and various circulatory conditions in a MCL. No statistically significant difference in HVAD flow error was found between LVAD support and BiVAD-B support over a wide variety of circulatory conditions, suggesting that the HVAD flow estimator accuracy is similar when the pump is used as an LVAD or banded RVAD, likely due to similar pump pressure-flow characteristics under these two support methods. All support methods demonstrated a statistically significant difference in HVAD error based on simulated HCT, with higher HCTs producing lower error. A statistically significant difference in the errors produced for each support method across all HCTs was found, with medians of 0.1 ± 0.3, −0.1 ± 0.2, and 0 ± 0.2 L/min for LVAD, BiVAD-B, and BiVAD-RS, respectively. Although statistically significant, given the small difference in mean flow errors between support methods across all HCTs (±0.1 L/min differences in IQR), it is unlikely that the differences are clinically significant for most patient cases, especially when interpreted and combined with clinical expertise. The lowest error for all of the experimental methods was found for an LVAD at an HCT of 40% (range = 0.5 L/min; IQR = −0.2 ± 0.1 L/min), followed closely by BiVAD-B at HCT 20% (range = 0.6 L/min; IQR = 0 ± 0.1) and BiVAD-RS at HCT 50% (range = 0.6 L/min; IQR = −0.3 ± 0.1 L/min).
Estimated pump flow pulsatility was consistently lower than that of the measured pulsatility, possibly due to differences in the internal filtering from the HVAD flow estimator (measured and estimated flow rates were not additionally filtered at the dSPACE during recording). Estimated and measured pump flow pulsatility of LVAD support demonstrated a strong correlation (r2 = 0.98), with a median error of 0.7 ± 0.3 L/min IQR, similar to the level of accuracy displayed by the mean LVAD flow estimator. Comparing the LVAD pump flow pulsatility results of this study to Granegger et al.,24 who reported an accuracy of −0.27 ± 0.2 L/min, a larger error was observed. However, this may be explained by the greater ranges of pump flow pulsatility tested in this study in contrast to Granegger et al. (0.5–11.5 L/min in the current study vs. 3.5–6.0 L/min in Granegger et al.). When considering the flow pulsatility results of RVAD support, the error ranges were smaller compared with LVAD support (0.7 ± 0.3 and 0.7 ± 0.2 for BiVAD-RS and BiVAD-B, respectively); this may be due to the smaller variations of pressure head from right ventricular pressure. Furthermore, with BiVAD-B, the correlation of flow pulsatility (r2 = 0.60) was notably smaller than LVAD (r2 = 0.98) and BiVAD-RS (r2 = 0.85) support, and therefore, clinical interpretation should be performed with caution.
This study demonstrated that PVR was a statistically significant factor in HVAD flow estimator error for BiVAD-RS (p < 0.001), possibly as PVR is a significant contributor to pulsatility transmitted through the RVAD in BiVAD-RS support. This would corroborate work by Reyes et al.,13 who previously reported a decreased correlation of the HVAD flow estimator with reduced pulsatility for LVAD support. Reduced accuracy with reduced pulsatility may be due to the internal filtering of the flow estimator algorithm, which attenuates the waveform peaks and troughs. This may be exacerbated by the MCL used in this study, which has a peak change in ventricular pressure (dP/dt), which is on the lower end of the range (300–400 mm Hg/s) reported for patients with heart failure at rest (300–1,000 mm Hg/s).25
VAD flow management is an important aspect of patient treatment regimens.1,15,26 While LVAD support for end-stage heart failure is well established, BiVAD support remains challenging, with INTERMACS data for BiVAD patients demonstrating 1-year survival rates of around 55%, having stagnated over the last 5 years.19 Previous case studies for BiVAD patients have anecdotally reported a lack of accuracy of the HVAD flow estimator for RVAD operated at reduced speed.5 This study further illuminates those studies, showing a good estimation of mean VAD flow rate, but a reduced estimation of instantaneous VAD flow rate for a reduced speed RVAD. Regardless, it is evident that an accurate flow estimator would be of great value for helping clinicians to achieve flow balance during BiVAD support due to the greater challenge compared with LVAD support alone.11
The results of the mean flow error from these experiments are in a similar range, though slightly higher than those reported by Reyes et al.13 and Giridharan et al.,16 who presented in vitro centrifugal LVAD flow estimator accuracies as described in Table 3. The error ranges reported in this study (Table 3) may again be higher than those reported due to the broader range of cardiovascular conditions simulated here. Some error may also stem from the variability between studies in matching viscosity to HCT.
Table 3. -
Reported Error in L/min for the HVAD Flow Estimator When Used as a LVAD at Different Hematocrits
|Reyes et al.13
||0.2 ± 0.2
||0.2 ± 0.2
||0.4 ± 0.3
||0.2 ± 0.1
|Giridharan et al.16
||0 ± 0.5
||0 ± 0.2
||0 ± 0.5
||0.4 ± 0.3
||−0.2 ± 0.2
||−0.2 ± 0.1
||0.1 ± 0.1
HCT, hematocrit; HVAD, HeartWare ventricular assist device; LVAD, left ventricular assist device.
A major limitation of in vitro evaluation of the HVAD flow estimator is the use of fluid viscosity as a corollary to HCT. The relationship between viscosity and HCT is complicated as HCT does not take into account the variability in blood plasma viscosity present in critical care patients. While the glycerol solutions used in this study have been widely accepted as a substitute for blood during in vitro testing and pump analysis, some differences in correlation with HCT may arise compared to clinical scenarios due to the non-Newtonian nature of blood, particularly when exposed to high shear conditions as associated with high speeds of a VAD impeller. Therefore, further studies should be conducted with whole blood to verify the results presented here and in other reported studies. Furthermore, the accuracy of the flow estimator was not investigated during suction events due to the MCL’s incapacity to model the VAD inlet obstruction, which occurs during suction. While future work may be required to evaluate the accuracy of the estimator during suction events, the capacity for the flow estimator to detect suction events has been evaluated extensively elsewhere.9,27,28
Another limitation of this study is the fact that it was conducted at 23°C, room temperature, instead of 37°C as would be found in the body. Although the viscosities investigated in this study were temperature compensated, it is possible that the accuracy of the HVAD flow estimator would be improved at physiologic temperatures due to the potential differences in pump operating current, and therefore power, at higher temperatures; however, this is only speculation.
The left atrial pressure values produced during simulated biventricular heart failure (8.7 mm Hg) are lower than the ranges reported in the literature (11–40 mm Hg).22 The low left atrial pressure in the MCL is a direct result of the current end-diastolic pressure–volume relationship in the MCL’s left ventricle, which is not physiologic due to the high-volume capacity of the ventricle. Despite low baseline values, left atrial pressure and other hemodynamics variables were within physiologic ranges (left atrial pressure: 8–25 mm Hg) during nominal BiVAD support (speeds between 2,000 and 3,000 rpm).
Another limitation with this study is the accuracy of the flow sensors (10PXL probe; TS410 meter, Transonic Inc.) used for the reference LVAD or RVAD flow, which is reported at ±4%. While the inherent error is quite small, this would equate to an error of 0.08–0.40 L·min−1 for flow rates of 2.0–10.0 L min−1. The calibration errors were reported much lower than this at approximately ± 0.2 L/min; however, this was for mean flow and could be worse for instantaneous changes.
A further potential limitation of the study is the use of a single HVAD and HVAD controller during experiments (N = 1), due to the limited availability of HVAD pumps at the authors’ institution. Anecdotally, experiments were completed under LVAD support at HCT 40% looking at potential hysteresis from increasing, decreasing, and randomizing VAD speed. There was no significant difference in error caused by the order in which the VAD speed was changed. This suggests that there is no hysteresis in the VAD flow estimator algorithm and that the error is consistent regardless of whether VAD speed is increasing or decreasing.
This study demonstrated a statistically significant difference in HVAD mean flow estimator error when used as an LVAD or banded RVAD compared with an RVAD operated at reduced speed in an MCL. The study showed that HCT was a statistically significant factor in HVAD flow error accuracy, suggesting that flow estimator accuracy improves at higher HCTs. Although further studies with whole human blood are required to validate the results, given the likely nonsignificant clinical differences in mean error between support methods, the outcomes from this study support the use of the HVAD flow estimator for BiVAD applications regardless of support method. Nonetheless, caution is advised when interpreting the instantaneous waveform for RVAD support methods. Studies such as this will allow clinicians to quantify and incorporate the known error into their decision-making process, thereby potentially improving patient management and outcomes.
1. Slaughter MS, Pagani FD, Rogers JG, et al. Clinical management of continuous-flow left ventricular assist devices in advanced heart failure. J Heart Lung Transplant. 29:S1–S39, 2010
2. Patel ND, Weiss ES, Schaffer J, et al. Right heart dysfunction after left ventricular assist device implantation: a comparison of the pulsatile HeartMate I and axial-flow HeartMate II devices. Ann Thorac Surg. 86:832–840; discussion 832, 2008
3. Lampert BC, Teuteberg JJ. Right ventricular failure after left ventricular assist devices. J Heart Lung Transpl. 34:1123–1130, 2015
4. Hetzer R, Krabatsch T, Stepanenko A, Hennig E, Potapov EV. Long-term biventricular support with the heartware implantable continuous flow pump. J Heart Lung Transplant. 29:822–824, 2010
5. Strueber M, Meyer AL, Malehsa D, Haverich A. Successful use of the HeartWare HVAD rotary blood pump for biventricular support. J Thorac Cardiovasc Surg. 140:936–937, 2010
6. Krabatsch T, Potapov E, Stepanenko A, et al. Biventricular circulatory support with two miniaturized implantable assist devices. Circulation. 12411 supplS179–S186, 2011
7. Schmidt T, Bjarnason-Wehrens B, Mommertz S, et al. Changes in total cardiac output and oxygen extraction during exercise in patients supported with an HVAD left ventricular assist device. Artif Organs. 42:686–694, 2018
8. Clifford R, Robson D, Gross C, et al. Beat-to-beat detection of aortic valve opening in HeartWare left ventricular assist device patients. Artif Organs. 43:458–466, 2019
9. Moscato F, Granegger M, Edelmayer M, Zimpfer D, Schima H. Continuous monitoring of cardiac rhythms in left ventricular assist device patients. Artif Organs. 38:191–198, 2014
10. Salamonsen RF, Lim E, Gaddum N, et al. Theoretical foundations of a Starling-like controller for rotary blood pumps. Artif Organs. 36:787–796, 2012
11. Saito S, Sakaguchi T, Miyagawa S, et al. Biventricular support using implantable continuous-flow ventricular assist devices. J Hear Lung Transplant. 30:475–478, 2011
12. Banerjee D, Dutt D, Duclos S, Sallam K, Wheeler M, Ha R. Simultaneous ramp right heart catheterization and echocardiography in a ReliantHeart left ventricular assist device. World J Cardiol. 9:55–59, 2017
13. Reyes C, Voskoboynikov N, Chorpenning K, et al. Accuracy of the HVAD pump flow estimation algorithm. ASAIO J. 62:15–19, 2016
14. Farrar DJ, Bourque K, Dague CP, Cotter CJ, Poirier VL. Design features, developmental status, and experimental results with the Heartmate III centrifugal left ventricular assist system with a magnetically levitated rotor. ASAIO J. 53:310–315, 2007
15. Slaughter MS, Bartoli CR, Sobieski MA, et al. Intraoperative evaluation of the HeartMate II flow estimator. J Hear Lung Transplant. 28:39–43, 2009
16. Giridharan GA, Koenig SC, Soucy KG, et al. Left ventricular volume unloading with axial and centrifugal rotary blood pumps. ASAIO J. 61:292–300, 2015
17. Kirklin JK, Naftel DC, Pagani FD, et al. Seventh INTERMACS annual report: 15,000 patients and counting. J Hear Lung Transplant. 34:1495–1504, 2015
18. Kirklin JK, Pagani FD, Kormos RL, et al. Eighth annual INTERMACS report: Special focus on framing the impact of adverse events. J Hear Lung Transplant. 36:1080–1086, 2017
19. Kormos RL, Cowger J, Pagani FD, et al. The society of thoracic surgeons intermacs database annual report: Evolving indications, outcomes, and scientific partnerships. J Hear Lung Transplant. 38:114–126, 2019
20. Gregory SD, Pauls JP, Wu EL, et al. An advanced mock circulation loop for in vitro
cardiovascular device evaluation. Artif Organs. 44:E238–E250, 2020
21. Gregory SD, Stevens M, Timms D, Pearcy M. Replication of the Frank-Starling response in a mock circulation loop, in Proceedings of the Annual International Conference of the IEEE Engineering in Medicine and Biology Society, EMBS, 2011, pp. 6825–6828
22. Muthiah K, Humphreys DT, Robson D, et al. Longitudinal structural, functional, and cellular myocardial alterations with chronic centrifugal continuous-flow left ventricular assist device support. J Hear Lung Transplant. 36:722–731, 2017
23. Guyton AC, Hall JE. Textbook of Medical Physiology. Guyton AC, Hall JE. 11th ed. Pennsylvania: Elsevier Saunders, 2006
24. Granegger M, Moscato F, Casas F, Wieselthaler G, Schima H. Development of a pump flow estimator for rotary blood pumps to enhance monitoring of ventricular function. Artif Organs. 36:691–699, 2012
25. Kolias TJ, Aaronson KD, Armstrong WF. Doppler-derived dP/dt and -dP/dt predict survival in congestive heart failure. J Am Coll Cardiol. 36:1594–1599, 2000
26. Kirklin JK, Pagani FD, Goldstein DJ, et al. American Association for Thoracic Surgery/International Society for Heart and Lung Transplantation guidelines on selected topics in mechanical circulatory support. J Hear Lung Transplant. 159:865–896, 2020
27. Ferreira A, Boston JR, Antaki JF. A control system for rotary blood pumps based on suction detection. IEEE Trans Biomed Eng. 56:656–665, 2009
28. Vollkron M, Voitl P, Ta J, Wieselthaler G, Schima H. Suction events during left ventricular support and ventricular arrhythmias. J Hear Lung Transplant. 26:819–825, 2007