Dynamic Risk Estimation of Adverse Events in Ambulatory LVAD Patients: A MOMENTUM 3 Analysis

Area of Science:

• Cardiovascular Medicine and Surgery
• Biomedical Engineering
• Clinical Risk Prediction

Background:

• Adverse events related to hemocompatibility are common after left ventricular assist device (LVAD) implantation.
• Predicting these adverse events in LVAD patients remains a significant clinical challenge.

Purpose of the Study:

• To develop and validate a dynamic risk modeling approach using a multistate model.
• To predict the risk of gastrointestinal bleeding (GIB), stroke, or death in ambulatory LVAD patients.

Main Methods:

• Secondary analysis of the MOMENTUM 3 trial involving HeartMate 3 LVAD recipients followed for up to 2 years post-hospital discharge.
• A multistate model incorporating 145 initial variables was developed using multivariate logistic regression.
• Model performance was evaluated using the area under the curve (AUC) in a validation cohort, and a risk stratification tool was created.

Main Results:

• The final multistate model included 39 variables to predict GIB (16 variables), stroke (10 variables), and death (19 variables).
• At 2 years post-implantation, the incidence rates were 25.6% for GIB, 6.0% for stroke, and 12.3% for death.
• The model demonstrated strong predictive performance with AUCs of 0.70 for GIB, 0.69 for stroke, and 0.86 for death in predicting 30-day event risks.

Conclusions:

• The developed multistate model effectively informs 30-day risk assessment in ambulatory LVAD recipients.
• The model allows for dynamic risk recalculation as new patient-specific data become available.
• This tool facilitates accurate risk stratification, potentially guiding clinical decision-making for adverse event prevention.