Risk Prediction Model for Survival of Wait-List Patients on Axial CF-LVAD: A UNOS Database Analysis

Background: Continuous flow left ventricular assist devices (CF-LVADs) have become a viable option for patients with end-stage heart failure as a bridge to transplantation or the destination therapy. Methods: Adult patients listed for heart transplantation (2010-2015) with an axial CF-LVAD on the wait list were obtained from the UNOS database. The multivariate Cox regression model was used to predict the probability of survival after listing. Patients were divided into derivation (80%) and validation (20%) groups. Receiver operating characteristics curves and area under curves were used to define the strength of the model. Results: Risk factors on multivariate analyses were diabetes type I (hazard ratio [HR], 2.5; P = .018), presence of inotropes (HR, 1.6; P = .005), creatinine at listing (HR, 1.2; P < .001). No significant differences were observed between the derivation and validation groups for any of the variables. The area under the curve at 3, 6, and 12 months on the wait list was 0.69, 0.65, 0.63, respectively in the training set and 0.71, 0.65, 0.60, respectively in the validation set. Survival analyses showed that patients implanted with Heartmate II before listing had a better survival than those who were implanted after being on the wait list (HR, 0.78; P = .048). Conclusion: To our knowledge, this was the first time a risk prediction model was generated for wait-list survival of Heartmate II patients. A significant difference in survival was noted between patients who received their Heartmate II before being put on a wait list vs those who were implanted while on the list.