Abstract 19250: Importance of Considering Competing Risks in Patients With a Left Ventricular Assist Device (LVAD)
Introduction: Whereas a competing risk analysis has discriminated causes of death after heart transplantation, a similar analysis of outcomes after LVAD implant is not well established. Patients (pts) with LVADs are susceptible to competing risks, as outcomes of death, transplant, and explant are possible. Traditional survival analyses, commonly used in LVAD research, may incorrectly estimate the probability of an outcome and the relative influence of covariates.
Methods: We analyzed all LVAD pts at our institution from 2008 to present. Traditional Kaplan-Meier (KM) and cox-proportional hazards (CPH) analysis was performed for the outcome of all-cause mortality using covariates considered relative contraindications to heart transplant. Pts were censored at competing events of transplant or explant. Analyses were repeated with a competing risks regression (CRR) model based on the cumulative incidence function (CIF).
Results: A total of 256 pts were included, 177 as bridge to transplant (BTT) and 79 as destination therapy (DT). At a median follow up of 252 days, 63 pts died, 101 were transplanted, 9 were explanted, and 83 remained alive on support. While there was no difference in survival in DT vs. BTT pts using CPH (HR 1.4 [0.8-2.3], p=0.22), DT pts were more likely to die with an LVAD in CRR analysis (HR 2.6 [1.6-4.2], p=0.0001). Mortality was similar for pts age ≥65 vs. <65 using CPH (HR 1.3 [0.7-2.2], p=0.4), while CRR revealed older pts had a greater incidence of death on an LVAD (HR 2.1 [1.3-3.3], p=0.004). KM curves over-estimated risk of death when compared to the CIF (Figure). Meaningful differences in mortality were also observed for covariates such as the composite of diabetes, chronic kidney disease, or lung disease using CPH vs. CRR (HR 1.4 [0.8-2.6], p=0.2 vs. HR 2.0 [1.1-3.5], p=0.017).
Conclusions: Competing risks methodology is more appropriate for an accurate assessment of risk and may significantly alter the current interpretation of outcomes data in LVAD pts.
Author Disclosures: R.M. Wehbe: None. A.J. Rodriguez: None. A.S. Anderson: None. K. Ghafourian: None. E.E. Vorovich: None. J.E. Wilcox: None. C.W. Yancy: None. D.T. Pham: None. J.D. Rich: Consultant/Advisory Board; Modest; Heartware, St Jude.
- © 2016 by American Heart Association, Inc.