A Fully Magnetically Levitated Left Ventricular Assist Device — Final Report

Trial Design

In this randomized trial, we compared a centrifugal-flow left ventricular assist device with an axial-flow device in patients with advanced-stage heart failure. Details of the trial design have been published previously.^9-11 The protocol was approved by each institutional review board at 69 participating centers in the United States. The trial was sponsored by Abbott, which provided the devices, selected the sites, and analyzed the data (as indicated in the Supplementary Appendix), with verification performed by an independent statistician. The authors vouch for the completeness and accuracy of all the data and analyses and for fidelity of the trial to the protocol, which is available at NEJM.org.

Trial Population

We enrolled patients with advanced-stage heart failure who were deemed to be candidates for therapy with a left ventricular assist device, irrespective of whether the intended goal was to provide support as a bridge to transplantation or as destination therapy. Patients were excluded from the trial if biventricular circulatory support was expected to be necessary or if irreversible end-organ dysfunction or active infection was present. Written informed consent was obtained from all patients or their authorized representatives. Detailed inclusion and exclusion criteria are provided in the Supplementary Appendix.

Randomization and Data Collection

Patients were randomly assigned in a 1:1 ratio to receive either the centrifugal-flow pump or the axial-flow pump. Randomization was performed with the use of permuted blocks and stratification according to study center and was implemented with the use of an electronic data-capture system (eClinicalOS, Merge Healthcare). The investigators and patients were aware of the treatment assignments. Data on end points and adverse events were collected after implantation at 1 day, 1 week, discharge, 1 month, 3 months, 6 months, and then every 6 months until 2 years of follow-up had been completed.

Left Ventricular Assist Systems

The two left ventricular assist devices used were the HeartMate 3 fully magnetically levitated centrifugal continuous-flow pump and the HeartMate II mechanical-bearing axial continuous-flow pump. All the investigators were experienced in the use of the control axial-flow pump and underwent training before their first implantation of a centrifugal-flow pump. The recommended antithrombotic treatment in each group included aspirin at a dose of 81 to 325 mg daily and warfarin (target range for the international normalized ratio, 2.0 to 3.0).

End Points

The composite primary end point was survival at 2 years free of disabling stroke or reoperation to replace or remove a malfunctioning device. In the analysis of the primary end point, disabling stroke was defined by a modified Rankin score of greater than 3 (scores range from 0 to 6, with higher scores indicating greater disability). The principal secondary end point, which the trial was separately powered to assess, was pump replacement at 2 years after implantation. Other secondary end points included actuarial survival, rehospitalization, functional status, and quality of life. Subgroup analyses were prespecified for age, sex, race, intended goal of therapy (bridge to transplantation or destination therapy), and severity of illness at baseline.

The rates of major adverse events such as stroke, bleeding, right heart failure, and infection were also evaluated (definitions are provided in the Supplementary Appendix). An independent clinical-events committee, the members of which were unaware of the treatment assignments, adjudicated the causes of death and bleeding, infection, neurologic dysfunction, suspected device thrombosis, and hemolysis. A 6-minute walk test administered by a trained technician and New York Heart Association (NYHA) classification were used to evaluate functional status. Quality of life was assessed with the European Quality of Life–5 Dimensions (EQ-5D) 5-Level questionnaire (EQ-5D-5L), the EQ-5D visual-analogue scale (EQ-5D VAS), and the Kansas City Cardiomyopathy Questionnaire (KCCQ).

Statistical Analysis

The sample size for the full cohort of patients who underwent randomization was calculated to test whether the centrifugal-flow left ventricular assist device was superior to the axial-flow left ventricular assist device with respect to the percentage of patients undergoing pump replacement at 2 years. We determined that, with 7% of patients in the axial-flow pump group and 3% of patients in the centrifugal-flow pump group expected to have undergone pump replacement at 2 years, 1028 patients would be required to show superiority with a power of 80% and an alpha of 0.05 (two-sided) (details are provided in the Supplementary Appendix).

The primary end point was analyzed in the intention-to-treat population, which included all patients who underwent randomization. Patients were considered to have had treatment failure with respect to the primary end point if any of the following events occurred: withdrawal from the trial before implantation (i.e., the patient underwent randomization but did not undergo implantation of the assigned device), withdrawal from the trial after implantation, death, disabling stroke, any pump replacement, urgent transplantation for device malfunction, or pump explantation or permanent deactivation for a reason other than myocardial recovery. If a patient had more than one of these events, the event that occurred first was noted as the treatment-failure event in the analysis of the primary end point. Transplantation for reasons other than device malfunction and pump explantation or permanent deactivation for myocardial recovery were not considered treatment-failure events.

Noninferiority with regard to the primary end point would be shown if the lower 95% confidence boundary of the difference between the treatment groups (centrifugal-flow pump minus axial-flow pump) in the percentage of patients remaining alive and free of disabling stroke or reoperation to replace or remove a malfunctioning device at 2 years was numerically greater than −10 percentage points (i.e., closer to or greater than zero) with a two-tailed P value of less than 0.05, calculated by the Farrington–Manning risk-difference approach. If noninferiority was met, then superiority was evaluated with the z test of proportions, with the use of the normal approximation to the binomial distribution. Relative risks and 95% confidence intervals for the primary end point and its component events are shown, and actuarial event-free survival was calculated by the Kaplan–Meier method.

All secondary end points were analyzed in the per-protocol population, which included all patients who underwent randomization and received the assigned device. The principal secondary end point was analyzed with Fisher’s exact test. Poisson regression was used to compare the rates of major adverse events between the two treatment groups as numbers of events per patient-year. The rate differences are described as relative risks with the corresponding 95% confidence intervals. The incidence of rehospitalization in discharged patients was evaluated with the Andersen–Gill model. Overall survival was analyzed by the Kaplan–Meier method, with data censored for nonfatal outcomes, such as transplantation. Hazard ratios were calculated from Cox proportional-hazards models. Longitudinal changes in functional status and quality of life were analyzed by means of linear mixed-effects modeling.

The reported P values for the primary and principal secondary end point are two-tailed. The 95% confidence intervals have not been adjusted for multiplicity, and therefore inferences drawn from these intervals may not be reproducible. Statistical analyses were performed with the use of SAS software, version 9.4 or higher (SAS Institute).

Patients and Device Implantation

Table 1. Table 1. Baseline Characteristics of Patients in the Intention-to-Treat Population.

From September 2014 through August 2016, a total of 1028 patients underwent randomization (516 patients to the centrifugal-flow pump group and 512 patients to the axial-flow pump group) (Fig. S2 in the Supplementary Appendix). The baseline characteristics of the patients are shown in Table 1, and in Table S1 in the Supplementary Appendix.

One patient who had been assigned to receive the centrifugal-flow pump and 7 who had been assigned to receive the axial-flow pump did not undergo implantation per protocol. The remaining patients (per-protocol population) included 515 who underwent implantation of a centrifugal-flow pump and 505 who underwent implantation of an axial-flow pump. A total of 126 surgeons performed 1020 implantations at 69 sites.

End Points

Table 2. Table 2. Primary and Principal Secondary End Points.

All patients were followed for 2 years, and no end-point data were missing. In the analysis of the primary end point, a larger percentage of patients in the centrifugal-flow pump group than in the axial-flow pump group remained alive and free of disabling stroke or reoperation to replace or remove a malfunctioning device at 2 years (397 patients [76.9%] vs. 332 patients [64.8%]). The noninferiority criterion (absolute between-group difference, 12.1 percentage points; 95% confidence interval [CI], 6.0 to 18.2; P<0.001) and superiority criterion (relative risk, 0.84; 95% CI, 0.78 to 0.91; P<0.001) were met. Details of the primary end-point components according to the first treatment-failure event that occurred are shown in Table 2, and in Table S2 in the Supplementary Appendix.

Figure 1. Figure 1. Kaplan–Meier Estimates of the Primary End Point in the Intention-to-Treat Population.

The intention-to-treat population included all patients who underwent randomization. The primary end point was a composite of survival free of disabling stroke or reoperation to replace or remove a malfunctioning device at 24 months after implantation. Disabling stroke was defined by a modified Rankin score of greater than 3 (scores range from 0 to 6, with higher scores indicating more severe disability). Kaplan–Meier estimates for the primary end point at 6, 12, and 24 months are shown. The 95% confidence intervals have not been adjusted for multiplicity, and therefore inferences drawn from these intervals may not be reproducible.

In a worst-case sensitivity analysis, patients in the axial-flow pump group who withdrew from the trial before or after implantation or who underwent urgent heart transplantation for device malfunction were considered to have had treatment success, whereas such patients in the centrifugal-flow pump group were still considered to have had treatment failure (Table S3 in the Supplementary Appendix). The result of this analysis was similar to that of the primary end-point analysis (relative risk, 0.91; 95% CI, 0.84 to 0.98). The Kaplan–Meier estimates of actuarial event-free survival at 2 years (primary end point) in the intention-to-treat population were 74.7% in the centrifugal-flow pump group and 60.6% in the axial-flow pump group (Figure 1).

The principal secondary end point of pump replacement at 2 years occurred less frequently in the centrifugal-flow pump group than in the axial-flow pump group (12 patients [2.3%] vs. 57 patients [11.3%]; relative risk, 0.21; 95% CI, 0.11 to 0.38; P<0.001) (Table 2). A list of the specific reasons for pump replacement in each group is provided in Table S4 in the Supplementary Appendix.

Clinical Course

Table 3. Table 3. Postdischarge End Points among Patients Discharged while Receiving Left Ventricular Assist Device Support (Per-Protocol Population).

Discharge from the hospital with the device in place occurred in 485 of 515 patients (94.2%) in the centrifugal-flow pump group and 471 of 505 (93.3%) in the axial-flow pump group. The median length of stay during the hospitalization for implantation was 19 days in the centrifugal-flow pump group and 17 days in the axial-flow pump group. The median number of days of rehospitalization, days spent outside the hospital receiving left ventricular assist device support, and all-cause rehospitalization rates are shown in Table 3.

There were 98 deaths in the centrifugal-flow pump group and 103 in the axial-flow pump group after implantation, with no significant difference in overall survival between the groups (Fig. S3 in the Supplementary Appendix). The most common causes of death among the patients with either device were right heart failure, stroke, and infection (Table S5 in the Supplementary Appendix). Competing-risk curves that reflect the cumulative percentages of patients in each group who had an outcome of ongoing device support, transplantation, device explantation or deactivation, or death through 2 years are shown in Figure S4 in the Supplementary Appendix.

Safety Outcomes

Figure 2. Figure 2. Principal Safety Outcomes in the Per-Protocol Population.

Relative risks of key adverse events, calculated on the basis of the number of events per patient-year in the centrifugal-flow pump group as compared with the axial-flow pump group, are shown. The per-protocol population included all patients who underwent randomization and received the assigned device. The relative risk of an adverse event favors the centrifugal-flow pump when the upper boundary of the 95% confidence interval is less than 1.0. Neither pump is favored when the 95% confidence interval spans the line of unity. The 95% confidence intervals have not been adjusted for multiplicity, and therefore inferences drawn from these intervals may not be reproducible. Narratives for suspected pump-thrombosis events in the centrifugal-flow pump group are provided in the Supplementary Appendix. P values for relative risk are derived from Poisson regression. Other neurologic events included transient ischemic attack, seizure, encephalopathy, and neurologic events other than stroke.

Between-group comparisons of event rates for hemocompatibility-related adverse events (including pump thrombosis, stroke, and bleeding) and other major events are shown in Figure 2. Instantaneous hazard risk curves are shown in Figure S5 in the Supplementary Appendix. There were 70 cases of suspected or confirmed pump thrombosis with the axial-flow pump and 7 cases with the centrifugal-flow pump; narratives of these events with the centrifugal-flow pump are provided in the Supplementary Appendix. Data on other major adverse events are shown in Figure S6 and Tables S6 and S7 in the Supplementary Appendix. Ventricular arrhythmias were less frequent with the centrifugal-flow pump than with the axial-flow device.

There were no significant differences between the groups with respect to clinical factors associated with stroke risk, including history of atrial fibrillation, history of stroke, maintenance of mean arterial blood pressure at or below 90 mm Hg, or receipt of antithrombotic therapy during the trial. Lactate dehydrogenase levels were lower in the centrifugal-flow pump group than in the axial-flow pump group. Details of these analyses and data regarding hepatic and renal function are provided in Table 1, and in Tables S8 through S10 and Figure S7 in the Supplementary Appendix.

Functional Status and Quality of Life

For both groups, there were sustained improvements from baseline in the 6-minute walk test, NYHA functional class, KCCQ overall summary score, EQ-5D-5L total score, and the EQ-5D VAS (Fig. S8 in the Supplementary Appendix). There were no significant differences in functional-status test results or quality-of-life measures between the treatment groups. Sensitivity analyses accounting for missing values favored the centrifugal-flow pump group over the axial-flow pump group with regard to the NYHA classification and the EQ-5D VAS.

Subgroup Analyses

No interaction between the groups was observed for the prespecified subgroups of age, sex, race or ethnic group, intended goal of pump support (bridge to transplantation or destination therapy), or Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) profile with regard to the primary end point. Details are provided in Figure S9 in the Supplementary Appendix.

In this final report of the MOMENTUM 3 trial, we showed that the HeartMate 3 fully magnetically levitated centrifugal-flow pump was superior to the HeartMate II axial-flow left ventricular assist system with respect to survival free of disabling stroke or reoperation to replace or remove a malfunctioning device. In addition to reducing the need for pump replacement (mostly for pump thrombosis), the centrifugal-flow pump was associated with a lower incidence of either ischemic or hemorrhagic strokes of any severity and fewer bleeding events. We calculate that for every 10 patients who received the centrifugal-flow pump rather than the axial-flow pump, 2.2 pump-thrombosis events, 2 strokes, and 6.8 bleeding events (3.6 gastrointestinal bleeds) were averted over a 2-year period. These findings were accompanied by fewer days spent in the hospital at 2 years among patients who received the magnetically levitated left ventricular assist device.

Reports of lower stroke rates with the centrifugal pump that were noted in earlier analyses of this trial were interpreted cautiously, since only the rates of nondisabling strokes were favorably influenced.^11-14 In the population in the current analysis, which was 3 times as large, we found lower rates of both hemorrhagic and ischemic strokes and a lower rate of disabling strokes. Stroke rates were not influenced by differences in baseline rates of atrial fibrillation, anticoagulation regimens, or blood pressure control in our current or previous analyses.¹⁴

The magnetically levitated centrifugal pump does not degrade high-molecular-weight multimers of Von Willebrand factor to the extent observed with other devices.¹⁵ This finding is consistent with the lower lactate dehydrogenase levels that were found in patients with this pump, which suggests that less hemolysis was occurring. However, in previous analyses of the MOMENTUM 3 trial, no discernible benefit with regard to the incidence of mucosal bleeding was noted.^10,11 In this final analysis involving the full trial population, we noted that the centrifugal pump was associated with lower rates of bleeding, including gastrointestinal bleeding, which is linked with the unique physiological features of continuous-flow left ventricular assist devices.^16,17 However, a significant residual risk of bleeding persists, which suggests the need to investigate whether a reduction in the exposure of patients with these pumps to antiplatelet or anticoagulation therapy would be beneficial.^2,18

In our trial, infections affecting the drive line exit site and in other locations occurred frequently, with no significant differences between left ventricular assist devices. Studies have made note of aberrant T-cell activation and compromised cellular immunity in patients with older pumps.^19,20 Efforts to eliminate the drive line by developing internally powered durable heart pumps are likely to be an important advance in this regard. However, vigilance in preventing generalized infections will remain important.^21,22 Late-onset right heart failure, a leading cause of death, also needs further study.^23,24 Outflow-graft twist occlusion, a unique adverse event with the centrifugal pump, has been noted previously.^25-27 Recognition of this insidious complication is important, since it compromises pump flow and may mimic pump thrombosis. A surgical clip to fix the pump outflow graft to the connector is now available to help prevent outflow-graft rotation.

In summary, in this final analysis from the MOMENTUM 3 trial, the centrifugal-flow HeartMate 3 left ventricular assist device was associated with a less frequent need for pump replacement than the axial-flow HeartMate II left ventricular assist device and was superior to the axial-flow pump with respect to survival free of disabling stroke or reoperation to replace or remove a malfunctioning device.