Prediction of Survival After Fully Magnetically Levitated LVAD

Quick Takes

  • A practical, six-component, pre-implant HeartMate 3 risk score accurately predicts 1- and 2-year survival after LVAD implant.
  • This tool can help with patient shared decision making when also considering potentially lower survival without LVAD on continued medical therapy.

Study Questions:

For patients with a HeartMate 3 (HM3) left ventricular assist device (LVAD), can a pre-implant risk score be developed and validated to predict 1- and 2-year survival after LVAD implant?


A total of 2,200 patients implanted with an HM3 LVAD were included in this analysis, with 515 patients from the HM3 randomized arm of the MOMENTUM 3 trial (2014–2016) and 1,685 patients in the subsequent Continued Access Protocol study cohort (2016–2018). These patients included both bridge-to-transplant and destination therapy LVADs. All patients were followed for 2 years or until death, heart transplantation, permanent HM3 removal or deactivation, or withdrawal. These 2,200 patients were randomized in a 70:30 ratio to a derivation (n = 1,540) or validation cohort (n = 660).

The outcomes of interest were 1- and 2- year survival post-implant of the HM3 LVAD. Univariate analyses were conducted on pre-implant clinical and testing variables to screen for potential predictors. Univariate predictors, along with prespecified demographic variables, were considered for inclusion in a multivariable model. Using a stepwise selection, a multivariable model was created to incorporate into the HM3 risk score (HM3RS). The HM3RS was then applied to the validation cohort to evaluate score performance. Additional analyses were conducted with the combined derivation and validation cohorts. This combined cohort was divided into tertiles based on HM3RS values to assess differences in survival. A Meta-Analysis Global Group in Chronic Heart Failure (MAGGIC) score was also calculated for each patient to estimate a 1-year survival on continued medical therapy. This was compared to 1-year survival with LVAD therapy.


The derivation and validation cohorts were similar except for more African American patients (26.2% vs. 31.6%, p = 0.011) and more frequent moderate to severe tricuspid valve regurgitation (27.6% vs. 33.2%, p = 0.009) in the validation cohort. At 2-year follow-up, 271 patients (17.6%) in the derivation cohort and 120 patients (18.2%) in the validation cohort died. Several variables were identified in the univariate analysis as a possible predictor of mortality. Of these, age, prior valve procedure or coronary artery bypass surgery, lower serum sodium, higher blood urea nitrogen, left ventricular end-diastolic diameter (LVEDD) <5.5 cm, and right atrial/pulmonary capillary wedge pressure ratio (RAP/PCWP) >0.6 remained as significant predictors of mortality in the multivariable model.

When applying the HM3RS to the validation cohort, the area under the curve (AUC) for the time-dependent receiver operating characteristic (ROC) curves at 1 and 2 years were 0.76 (95% confidence interval [CI], 0.70-0.81) and 0.71 (95% CI, 0.66-0.77), respectively. After dividing patients into quintiles based on HM3RS, correlation between predicted and observed survival was high with Pearson correlation coefficients of 0.986 and 0.994 at 1 and 2 years, respectively. With the combined derivation and validation cohort, patients were divided into tertiles based on HM3RS. There was an approximate twofold increase in observed mortality risk between tertiles. Predicted mortality at 1 year without HM3 (MAGGIC score) and with HM3 (HM3RS) were compared, with lower mortality with HM3 predicted for 97% of patients.


The authors concluded that a six-component HM3RS was derived and accurately predicts 1- and 2-year survival after LVAD implant.


LVAD technology continues to improve, leading to better survival and less hemocompatibility-related adverse events. With this advance in technology, newer risk prediction models for contemporary LVADs like the HM3 are needed. While risk prediction scores in general have limitations, they can better inform clinicians and patients when deciding on treatment choices. The HM3RS from this study appears accurate in predicting mortality in patients with advanced heart failure after LVAD implant, with acceptable ROC AUC values. Importantly, the variables of this risk score are readily obtainable in clinical practice and make the tool more likely to be used. The HM3RS may have a future role in patient selection and shared decision making for LVAD therapy, but it will be interesting to see how this risk score will perform in an external cohort outside of the MOMENTUM trial portfolio and in a real-world setting. Use of this risk score in other contexts such as directing optimizing of patients prior to LVAD by targeting modifiable variables or as a tool for heart transplantation decisions are still unclear and may be areas of future investigation.

Clinical Topics: Cardiac Surgery, Heart Failure and Cardiomyopathies, Invasive Cardiovascular Angiography and Intervention, Prevention, Cardiac Surgery and Arrhythmias, Cardiac Surgery and Heart Failure, Acute Heart Failure, Heart Transplant, Mechanical Circulatory Support, Interventions and Structural Heart Disease

Keywords: Cardiac Surgical Procedures, Coronary Artery Bypass, Heart Failure, Heart Transplantation, Heart-Assist Devices, Pulmonary Wedge Pressure, Risk Factors, Secondary Prevention, Tricuspid Valve Insufficiency, Survival

< Back to Listings