Abstract

INTRODUCTION: Adverse long-term kidney outcomes with left ventricular assist devices (LVADs) are common but little understood. Creatinine-based estimated glomerular filtration rate (eGFR) is used clinically in this setting; however, it is of uncertain validity and tends to change nonlinearly after LVAD implantation with an early peak followed by decline. We investigated nonlinear eGFR trajectories over 90 days following LVAD implantation and the relationships with patient factors and outcomes. METHODS: We used 2 complementary cohorts of LVAD recipients, a single-center cohort (<i>n</i> = 190) and a US national cohort (<i>n</i> = 10794). Segmented linear regression with 2 segments was compared to other trajectory models. Relationships of trajectory parameters with patient factors and clinical outcomes were assessed. RESULTS: Segmented linear regression resulted in the best fit, yielding these parameters: early and late trajectory slopes, breakpoint time, and breakpoint eGFR. Patient factors were associated with trajectory parameters. For late trajectory slope, female sex was associated with change of -1.11 (-1.54 to -0.66) ml/min per 1.73 m²/30 d and older age (/10 yrs) of -0.80 (-0.94 to -0.66) ml/min per 1.73 m²/30 d in the national cohort. All 4 trajectory parameters were associated with survival in unadjusted models, and breakpoint eGFR and late trajectory slope associations remained significant following adjustment. CONCLUSION: Kidney function trajectory following LVAD implantation can be modeled with linear early and late phases separated by a variable breakpoint, yielding metrics associated with survival and preimplantation patient and hemodynamic factors. These trajectory metrics may help investigations of causes of adverse kidney outcomes and interventions in LVAD recipients.