Inflammatory biomarkers profiles and cognition among older adults

Abstract

Inflammation plays a major role in cognitive aging. Most studies on peripheral inflammation and cognitive aging focused on selected major inflammatory biomarkers. However, inflammatory markers are regulated and influenced by each other, and it is therefore important to consider a more comprehensive panel of markers to better capture diverse immune pathways and characterize the overall inflammatory profile of individuals. We explored 23 circulating inflammatory biomarkers using data from 1,743 participants without dementia (≥ 65 years-old) from the community-based, multiethnic Washington Heights Inwood Columbia Aging Project. Using principal component analysis (PCA), we developed six inflammatory profiles (PC-1 to PC-6) based on these 23 biomarkers and tested the association of resulting inflammatory profile with cognitive decline, over up to 12 years of follow-up. PC-1 described a pro-inflammatory profile characterized by high positive loadings for pro-inflammatory biomarkers. A higher PC-1 score was associated with lower baseline cognitive performances. No association of this profile with cognitive decline was observed in longitudinal analysis. However, PC-5 characterized by high PDGF-AA and RANTES was associated with a faster cognitive decline. Among older adults, a circulating pro-inflammatory immune profile is associated with lower baseline cognitive performance, and some specific pro-inflammatory cytokines might be associated with faster cognitive decline.

Keywords: Cognition; Cognitive decline; Cohort; Epidemiology; Peripheral inflammation.

Fig. 1

Associations of 23 inflammatory biomarkers with cognition, estimated by linear mixed models, WHICAP (n = 1743). The figure displays the effect estimates for the associations of continuous inflammatory biomarkers with cognitive z-scores at baseline (Panel A) and cognitive decline (Panel B) over the 12-year follow-up. The trajectories of change in cognition were estimated using linear mixed models across up to 6 repeated neurocognitive examinations. Models consider a linear function of time, with corresponding random effect, and include an intercept representing the cognitive z-score at baseline (and corresponding random effect), inflammatory markers (continuous, log-transformed and standardized), covariates (age, gender, race/ethnicity, status for ɛ4 allele of the apolipoprotein E gene, and indicator for first cognitive assessment), and their interactions with time. The estimate for baseline association (Panel A) is the coefficient for the inflammatory markers variable term; and the estimate for cognitive decline (Panel B) is the coefficient for the biomarker-by-time interaction term. Effect estimates (β coefficients and 95% confidence intervals) are reported for 1-SD increase in log-transformed biomarker value. Significant associations are indicated by an asterisk: * for p-value < 0.05, ** for p-value < 0.01, and *** for p-value < 0.05 after False Discovery Rate correction for multiple testing.