Abstract P335: Predictors of Cognitive Resilience in Those With and Without High White Matter Hyperintensity Burden: A Post-Hoc Analysis of Systolic Blood Pressure Intervention Trial Memory & Cognition in Decreased Hypertension (SPRINT MIND)

Abstract

Introduction: Little is known if predictors of cognitive resilience (CR) differ between adults with or without large white matter hyperintensity burdens. In the current analysis of cross-sectional baseline SPRINT MIND data, we a) determined bivariate associations of sociodemographic, vascular-metabolic, and neurodegenerative biomarkers with CR and b) delineated variables that maximized predictive fit of CR among participants with the highest and lowest white matter hyperintensity volumes (WMHv). Methods: We included SPRINT MIND participants with baseline available plasma biomarker concentrations & Montreal Cognitive Assessment (MoCA) scores. We then created two analytic samples: those with the highest tertile of WMHv (N = 186) and those with the lowest tertile (N=134). We defined CR two ways: as a MoCA score a) >25 or b) >22. We determined the association between sociodemographic, vascular-metabolic, and serum neurodegenerative variables, as well as WMHv with CR using logistic regressions to generate unadjusted odds ratios (OR) & 95% confidence intervals (CI). Multivariate lasso regressions were then used to determine which variables optimized predictive fit of CR, and R 2 values were calculated. Results: In participants with the lowest tertile of WMHv , the median age was 66 years, 45.5% were female and 63% were Non-Hispanic white. In participants with the highest tertile of WMHv, the median age was 73 years, 45.7% were female and 66% were Non-Hispanic white. . CR defined as MoCA scores > 25 & >22 was present in 37.3% and 64.9% in those with the lowest tertile of WMHv and 28.5% & 56.5% in those with the highest tertile, respectively. Higher education and self-reported race generally had the strongest ORs in all analyses (Table 1 excerpt). Only age and education level were included in all four predictive models after lasso shrinkage. The R 2 values for all lasso regressions were <0.25. Conclusion: Education level may be associated with CR irrespective of WMHv. However, a large variation of CR remains unaccounted for.