A Computational Hemodynamics Approach to Left Ventricular Assist Device (LVAD) Optimization Validated in a Large Patient Cohort.

Abstract

With increasing use of left ventricular assist devices (LVAD) it is critical to devise strategies to optimize LVAD speed while controlling mean arterial pressure (MAP) and flow according to patient physiology. The complex interdependency between LVAD speed, MAP, and flow frequently makes optimization difficult under clinical conditions. We propose a method to guide this procedure in silico, narrowing the conditions to test clinically. A computational model of the circulatory network that simulates HF and LVAD support, incorporating LVAD pressure-flow curves was applied retrospectively to anonymized patient hemodynamics data from the University of Washington Medical Center. MAP management on 61 patient-specific computational models with a target of 70 mm Hg, resulting flow for a given LVAD speed was analyzed, and compared to a target output of 5 L/min. Before performing virtual MAP management, 51% had a MAP>70 mm Hg and CO>5 L/min, and 33% had a MAP>70 mm Hg and CO<5 L/min. After changing systemic resistance to meet the MAP target (without adjusting LVAD speed), 84% of cases resulted in CO higher than 5 L/min, with a median CO of 6.79 L/min, using the computational predictive model. Blood pressure management alone is insufficient in meeting both MAP and CO targets, due to the risk of hypervolemia, and requires appropriate LVAD speed optimization to achieve both targets, while preserving right heart health. Such computational tools can narrow down conditions to be tested for each patient, providing significant insight into the pump-patient interplay. LVAD hemodynamic optimization has the potential to reduce complications and improve outcomes.