Comparing the Effectiveness of an Axial and a Centrifugal Left Ventricular Assist Device in Ventricular Unloading

Abstract

Centrifugal (CFG) and axial flow (AX) left ventricular assist devices have different hydrodynamic properties that may impact the effectiveness of left ventricular unloading. We sought to determine whether patients implanted with the HeartWare HVAD (CFG) and HeartMate II (AX) had a similar degree of hemodynamic support by comparing parameters measured using echocardiography and right heart catheterization. Using our prospectively collected database, we identified 268 patients implanted with the AX and 93 with the CFG. Demographic characteristics were similar between groups. AX patients had a significantly lower INTERMACS score. Baseline ventricular dimension, mitral regurgitation, right ventricular systolic pressure, right atrial pressure, mean pulmonary artery pressure, cardiac output, and pulmonary vascular resistance were similar. Wedge pressure was higher, and left ventricular ejection fraction was lower at baseline in the AX. After implantation, there was a greater reduction of right atrial pressure, pulmonary capillary wedge pressure, mean pulmonary artery pressure, and left ventricular internal diameter during diastole in the AX cohort. After implantation, cardiac output by Fick calculation showed a greater improvement in the AX group. These results demonstrate that both AX and CFG devices resulted in left ventricular unloading; however, AX devices may offer advantages in the magnitude of left ventricular unloading, which could have implications in myocardial recovery or reduction in pulmonary vascular resistance before transplantation.

Left ventricular assist devices (LVADs) are becoming increasingly used as a treatment for advanced heart failure as a bridge to transplant, destination therapy, or bridge to recovery. There are two continuous flow LVADs in clinical use today, the HeartMate II (HMII, Thoratec Corporation, Pleasanton, CA) and the HVAD (HW, HeartWare International Inc., Framingham, MA). The HeartMate II is an axial flow device, whereas the HVAD is a centrifugal flow device. The mechanical and hydrodynamic characteristics of these devices have been extensively reviewed in the literature.^1,2 In comparison with axial flow pumps, centrifugal design pumps are generally more sensitive to changes in the pressure differential across the pump. These differences can potentially impact the adequacy of mechanical circulatory support. Although there have been numerous reports comparing survival and the prevalence of adverse events between these pumps, there are little data specifically assessing their clinical performance in regard to ventricular volume and pressure unloading. Using our granular University of Michigan Mechanical Circulatory Support Registry, we sought to analyze the effectiveness of continuous flow ventricular assist device implantation on parameters of left ventricular unloading and to determine whether there were important differences between HMII and HW pumps.

Methods

All patients undergoing primary device implant who received an axial flow HMII or centrifugal flow HeartWare HVAD (HW) for any indication at the University of Michigan from May 2004 to April 2015 were eligible for inclusion in the study. Demographic and preoperative data were gathered either from a prospectively collected database or retrospectively from chart review of operative, echocardiogram, and right heart catheterization (RHC) reports.

Postoperative data were analyzed between 3 and 12 months after implantation, with a target of 6 months, if available. Patients who died or were transplanted before 3 months of implantation were excluded from the study. Left ventricular assist device speeds were identified from routine follow-up visits. If postimplantation parameters were not available within the 3–12 month window, that patient was excluded from both the baseline and the postimplantation calculations for that parameter.

Comparison of demographics, baseline characteristics, postimplantation data, and mean improvement between the two groups was performed using a two-sample t-test. The differences from baseline to postimplantation within LVAD groups were analyzed using paired two sample for means t-tests. A 95% confidence interval was calculated for each value. A p value of less than 0.05 was considered significant. Full IRB approval with waiver of informed consent was maintained throughout the duration of the study.

Results

In total, 268 HMII and 93 HW devices have been placed at our institution during the study interval. There were no intergroup differences in the number of patients excluded because of transplant or death before the minimum 3 month follow-up time point, as outlined in Table 1. After exclusion, 240 HMII and 84 HW patients were included in the analysis. Baseline demographics are shown in Table 2, with no differences in age, sex, body mass index, proportion of patients with ischemic heart disease, bridge to transplant versus destination therapy, and a similar percentage who underwent redo sternotomy. There were also no differences in the proportion of patients requiring mechanical right ventricular support or tricuspid valve repair at time of LVAD implantation. The INTERMACS score was lower in the HMII group than in the HW group. RHC and echo data were similar between groups at baseline (Table 3). Exceptions to this were wedge pressure, which was higher in the HMII group, and left ventricular ejection fraction, which was lower in the HMII group.

Table 1.: Total Number of LVAD Patients with Exclusions and Total Patients

Table 2.: Baseline Demographics

Table 3.: Baseline Right Heart Catheterization and Echocardiogram Parameters

Postimplantation values are shown in Table 4. The average pump speed was 9,180 rpm for the HMII and 2,700 rpm for the HW. Mean follow-up time within the HMII patients was 5.65 months (±0.279) for RHC parameters and 5.24 months (±0.211) for echo. Mean follow-up time for the HW patients was 5.53 months (±0.611) for RHC and 5.19 months (±0.310) for echo, with no significant difference in follow-up interval. The HMII group demonstrated improvement in all studied parameters postimplantation except right atrial pressure, which remained similar to baseline. The HW group demonstrated significant improvement in all studied parameters postimplantation except for Fick calculated cardiac output, which remained similar to baseline, and right atrial pressure, which increased significantly.

Table 4.: Comparison of Preimplantation and Postimplantation Parameters within Groups

The postimplant variables of the groups were then compared with each other, shown in Table 5 and mean differences shown in Figure 1. The HMII group had greater reduction in mean pulmonary artery pressure, wedge pressure, and left ventricular internal diameter in diastole. Fick estimated cardiac output increased more in HMII patients. All other parameters were similar between the two groups postimplantation. There was no difference in the degree of aortic valve opening.

Table 5.: Comparison of Postimplantation Parameters Between Groups

Figure 1.: Differences in right heart catheterization values from preimplant to postimplant.

Discussion

Continuous flow LVADs have transformed the treatment of patients with end-stage heart failure. Implantation is now standard of care for patients at imminent risk of death, either as a bridge to heart transplant or as destination therapy. These pumps operate at a fixed rotational speed, set by the clinician to optimize the adequacy of support. As physiologic conditions change, such as increased intravascular volume or during exercise, adaptive response in flow and pressure generated by these pumps is entirely dependent upon their intrinsic hydrodynamic properties. The HeartWare HVAD and the Thoratec HeartMate II, the two pumps currently approved for use in the United States, have substantial differences in their pumping mechanisms, which can impact how they respond under these dynamic physiologic conditions. Centrifugal design pumps may be more sensitive to inflow and outflow pressure than their axial design counterparts. This has been demonstrated by in vitro models, where preload and afterload conditions are adjusted.³ These differences in mechanistic properties between the centrifugal and the axial pumps used in this analysis theoretically could impact the degree of ventricular unloading and thus have clinical implications on heart failure support. In a small in vivo animal study, Giridharan et al.⁴ evaluated the two designs in a chronic heart failure bovine model and demonstrated no significant difference in unloading between the two devices over a short interval of support (Figure 2).

Figure 2.: Differences in echo parameters from preimplant to postimplant.

To date, all clinical comparative studies of continuous flow LVADs with respect to ventricular unloading have been limited in sample size and have reported little difference between axial and centrifugal pumps. Hosseini et al.⁵ analyzed postoperative echocardiograms to assess for differences in LV geometry. Their data demonstrated that both the axial flow and the centrifugal pumps produced significant ventricular unloading and showed a trend toward better unloading with the axial design, but did not reach statistical significance. Topkara et al.⁶ compared axial and centrifugal pumps when used as a bridge to transplant. They found no differences in hemodynamic parameters at the time of transplant; however, they did not report the effectiveness of unloading when compared with preimplant conditions.

Our study was substantially larger than previous reports and incorporated matched echo and right heart catheterization results. We verified that both pumps are extremely effective at improving cardiac output and reducing intracardiac filling pressures and pulmonary vascular resistance, and can be implanted in critically ill patients with very low perioperative mortality. We confirmed our hypothesis that the HeartMate II, which is less responsive to inflow and outflow conditions, was more effective at ventricular unloading, reducing left ventricular dimensions, pressures, and pulmonary vascular resistance when compared with the HeartWare HVAD.

Perhaps, the improved unloading observed in the HMII group was related to a more conservative approach to speed setting. There were fewer patients implanted with the HW pump, and the smaller experience could have resulted in clinical decisions to err on providing lower pump speeds. However, there has not been a significant change in pump speed setting over time with the HMII group to suggest that more aggressive speed setting occurs with additional experience. Aortic valve opening is often used to help set pump speeds both inside and outside of the operating room. There were no differences in the degree of aortic valve opening on follow-up echocardiograms in the two cohorts of patients, suggesting that similar approaches to LVAD speed were used. The HW speeds in our center’s experience were nearly identical to the speeds reported in the prospective multicenter clinical trial of the HVAD pump.⁷ Regardless of the pump technology employed, pump speeds are set on the basis of the echocardiographic and hemodynamic data available. We feel that the findings here represent real world data about pump performance by a high volume and experienced VAD program.

The HMII has an intrinsic pathway to detect ventricular suction and respond with transient speed reduction, which allows ventricular filling and prevents persistent LV collapse. Suction can be hazardous, as it is associated with right ventricular geometric changes, tricuspid regurgitation, and ventricular arrhythmias. There is no similar algorithm in the HW device. It is certainly conceivable that clinicians are concerned about ventricular suction and intentionally set pump speeds lower to provide assurances that suction will not occur. It is unknown whether a suction detection and response algorithm could result in improved ventricular unloading in the centrifugal pump.

There were some baseline differences between the two cohorts manifested by lower INTERMACS score and higher wedge pressures in the HMII patients. Perhaps, these differences were related to the investigational nature of the HW pump in the United States, specifically with respect to the destination therapy population of patients. Clinical trials have restrictive inclusion and exclusion criteria, which potentially could have contributed to some of the differing characteristics of the cohorts. One could postulate that the group of HMII patients had worse function at baseline and therefore have more room for improvement. However, we observed a significantly greater improvement in their ventricular dimensions and pressures when compared with the HW, despite their lower baseline function.

Are the differences we identified clinically relevant? We have learned that the frequency of adverse events like gastrointestinal bleeding, pump thrombosis, stroke, and right ventricular failure is of paramount importance for growth of mechanical circulatory support and expansion of ventricular assist device indications. We did not specifically address these adverse events, as they have been extensively reported in the context of randomized clinical trials and INTERMACS registry reports.^8,9 We sought to categorically focus on the hypothesis of ventricular unloading, which is physiologically intuitive but has not been clinically reported. Although the absolute difference in unloading is small, it is possible that even modest improvements in unloading could affect the likelihood of myocardial recovery and subsequent device explantation, of particular benefit in the small bridge-to-recovery population. In addition, satisfactory unloading is essential for prospective transplant candidates with pulmonary hypertension to improve their pulmonary pressures and candidacy for transplant. Although there were no overt clinical differences in our experience with these pumps, it is important to understand the impact of intrinsic hydrodynamic properties and flow adaptation to varying changes in physiologic conditions. As future pump designs are considered, suction response algorithms or even pressure and volume mediated pump speed adjustment capabilities could potentially address the limitations of rotary pumps that currently operate at fixed rotational speeds.