Abstract P3011: Empirically derived biomarker patterns in the REasons for Geographic and Racial Differences in Stroke Study

Abstract

Background: Contemporary molecular epidemiologic approaches often consider individual or small groups of biomarkers and their relationship to cardiovascular disease risk factors or events. Factor analysis can group many biomarkers that represent the physiology and dysfunction of multiple organ systems into a small number of distinct patterns. These patterns may better reflect the complexity of human physiology and advance understanding of cardiovascular disease onset. We sought to empirically identify such patterns using factor analysis, and hypothesized that biomarkers would cluster into distinct patterns. Methods: REGARDS recruited 30,239 Black and White adults aged ≥45 years from the continuous US in 2003-2007 with a second visit in 2013-2016. Specialized biomarkers were measured in specimens collected at baseline from 4,400 participants who attended both visits; lipids, C-reactive protein (CRP), cystatin C, urine albumin/creatinine ratio, complete blood count, blood urea nitrogen (BUN), and serum albumin were measured in all or most participants. Factor analysis was used to derive patterns in a random split sample. Scree plot guided selection of factors used an eigenvalue cut point of 1. Validity was confirmed across race, sex, and age groups (<55, 55 to <65, 65 to <75, and ≥75 years) A congruence test coefficient >0.5 was considered acceptable. Final factor patterns were derived in the entire analytical sample using a 3-factor solution and named based on biologically known pathways. Results: Of the 4,400 participants, 31% were excluded who were missing any biomarker leaving an analytical sample of 2,723. The congruence coefficient between the factors was >0.78 for all subgroup comparisons. As is shown in the Figure , the “thrombo-inflammatory” pattern (pattern 1) had higher levels of inflammation, coagulation, and leptin; the “renal-inflammatory” pattern (pattern 2) had high levels of inflammation, renal dysfunction, and also high NT-proBNP and proenkephalin; the “dyslipidemic-inflammatory” pattern (pattern 3) had high levels of inflammation and triglycerides and low levels of HDL and adiponectin. Discussion: Three observed biomarker patterns had strong congruence across race, sex, and age groups. Future research will determine if these patterns are associated with cardiovascular risk factors and events.