GDF15 and its association with cognitive performance over time in a longitudinal study of middle-aged urban adults

Abstract

Serum GDF15 levels are correlated with multiple neurodegenerative diseases. Few studies have tested this marker's association with middle-aged cognitive performance over time, and whether race affects this association is unknown. We examined associations of initial serum GDF15 concentrations with longitudinal cognitive performance, spanning domains of global mental status, visual and verbal memory, attention, fluency, and executive function in a sub-sample of adults participating in the Healthy Aging in Neighborhoods of Diversity across the Life Span (HANDLS) study (n = 776, Age_v1:30-66y, 45.6 % male, 57.0 % African American, 43.0 % below poverty). This analysis consisted of mixed-effects regression models applied to the total selected sample, while also stratifying the analyses by race in the main analyses and further stratifying by sex, age group and poverty status in an exploratory analysis. Our main findings, which passed multiple testing and covariate-adjustment, indicated that GDF15 was associated with poorer baseline performance on several cognitive tests, including animal fluency [overall sample: (Model 1: γ₀ ± SE: -0.664 ± 0.208, P < 0.001; Model 2, γ₀ ± SE: -0.498 ± 0.217, P < 0.05)]. Among White adults, GDF15 was linked to poorer performance on a brief test of attention (Model 1: γ₀ ± SE: -0.426 ± 0.126, P < 0.001; Model 2, γ₀ ± SE: -0.281 ± 0.139, P < 0.05); and Trailmaking test, part B (Model 1: γ₀ ± SE: +0.129 ± 0.040, P < 0.001; Model 2, γ₀ ± SE: +0.089 ± 0.041, P < 0.05), the latter being also linked to higher GDF15 among individuals living below poverty. Among women, GDF15 was associated with poor global mental status (Normalized MMSE: Model 1: γ₀ ± SE: -2.617 ± 0.746, P < 0.001; Model 2: γ₀ ± SE: -1.729 ± 0.709, P < 0.05). GDF15 was not associated with decline on any of the 11 cognitive test scores considered in ∼ 4 years of follow-up. In sum, we detected cross-sectional associations between GDF15 and cognition, although GDF15 did not predict rate of change in cognitive performance over time among a sample of middle-aged adults. More longitudinal studies are needed to address the clinical utility of this biomarker for early cognitive defects.

Keywords: Aging; Cognitive decline; GDF15; Health disparities; Race.

Figure 1.

Participant flowchart Abbreviations: HANDLS=Healthy Aging in Neighborhoods of Diversity across the Life Span; k=# of observations/participant; v₁=Visit 1; v₂=Visit 2.

Figures 2.. Predictive margins of GDF15 vs. cognitive performance over time, Animal Fluency, overall

^a GDF15v₁ values are Log_e transformed and z-scored. Levels of exposure are −1: mean – 1SD; 0: at mean; +1: mean+1SD. 1 SD of baseline Log_e(GDF15) is estimated at 0.70; Mean=6.55. All test scores presented in these figures are coded in the direction of higher score → better performance. Abbreviations: AF=Animal Fluency; GDF15_v1=Plasma GDF15 levels, Log_e transformed, z-scored at v₁.

Figure 3.. Summary of mixed-effect linear regression models, reduced model (i.e. Model 1), overall

Abbreviations: AA=African American; AF=Animal Fluency; BC=Baseline cognitive performance; BTA=Brief Test of Attention; BVRT=Benton Visual Retention Test; CC=Cognitive change; CDT=Clock Drawing Test; CVLT-DFR=California Verbal Learning Test-Delayed Free Recall; CVLT-List A=California Verbal Learning Test-List A; DS-B=Digits Span-Backward; DS-F=Digits Span-Forward; FC=Follow-up cognition; GDF15_v1=Plasma GDF15 levels, Log_e transformed, z-scored at v₁; TRAILS A=Trailmaking Test, Part A; TRAILS B=Trailmaking Test, part B. ^a 1 SD of baseline Log_e(GDF15) is estimated at 0.70; Mean=6.55.. BVRT, TRAILS A and B are coded in the direction of higher score → poorer performance. All other test scores are in the direction of higher score → better performance. ^bCognitive tests were: 1. Normalized MMSE; 2.CVLT-List A; 3.CVLT-DFR;4.BVRT;5.BTA;6.AF;7.DS-F;8.DSB;9.CDT;10.TRAILS A;11.TRAILS B.