Derivation and Validation of the EUROMACS Left Ventricular Assist Device Score for Long-Term Outcome - The EUROMACS-LVAD-Score

Purpose Durable left ventricular assist devices (LVAD) have been validated in randomized trials for bridge to transplantation and destination therapy in patients with advanced heart failure. To guide treatment outside of selected study populations, a reliable tool for risk assessment is needed, especially for contemporary devices. The aim of this project is to derivate and validate a European risk score for assessment of long-term outcome after LVAD implantation in advanced heart failure patients. Methods The EUROMACS Registry, a multi-centre, European database of consecutive patients implanted with a LVAD was used for this project. From the database, patients with primarily LVAD implantation and available follow-up information were selected. The primary endpoint was all-cause mortality at different time points; patients were censored at the event of LVAD explantation for recovery or heart transplantation. Results 2881 patients were identified from the registry. Mean age was 53.76 ±12.29 years and 81.8% of the patients were male. Mean length of follow-up was 455.35 ±479.79 days. The rate of the primary endpoint was 42.1%. Multiple variables (such as age, mitral regurgitation, use of vasopressors, functional status) were independently associated with the primary endpoint. For the risk score, a non-random split will be used to create a derivation and a validation cohort. Hereafter, the least absolute shrinkage and selection operator method will be used to select variables best suited for the predictive score. The score accuracy will then be tested in the validation cohort. Results will be presented at the ISHLT annual meeting 2019. Conclusion The EUROMACS- Score can serve as a viable tool to guide therapeutic decisions in advanced heart failure patients. Importantly, this score will be based on the largest European database using contemporary devices and should be able to minimize regional bias.