Validation of Intrinsic Left Ventricular Assist Device Data Tracking Algorithm for Early Recognition of Centrifugal Flow Pump Thrombosis

Abstract

Advanced stage heart failure patients can benefit from the unloading effects of an implantable left ventricular assist device. Despite best clinical practice, LVADs are associated with adverse events, such as pump thrombosis (PT). An adaptive algorithm alerting when an individual’s appropriate levels in pump power uptake are exceeded, such as in the case of PT, can improve therapy of patients implanted with a centrifugal LVAD. We retrospectively studied 75 patients implanted with a centrifugal LVAD in a single center. A previously optimized adaptive pump power-tracking algorithm was compared to clinical best practice and clinically available constant threshold algorithms. Algorithm performances were analyzed in a PT group (n = 16 patients with 30 PT events) and a thoroughly selected control group (n = 59 patients, 34.7 patient years of LVAD data). Comparison of the adaptive power-tracking algorithm with the best performing constant threshold algorithm resulted in sensitivity of 83.3% vs. 86.7% and specificity of 98.9% vs. 95.3%, respectively. The power-tracking algorithm produced one false positive detection every 11.6 patient years and early warnings with a median of 3.6 days prior to PT diagnosis. In conclusion, a retrospective single-center validation study with real-world patient data demonstrated advantageous application of a power-tracking algorithm into LVAD systems and clinical practice.