Premium
P1–255: Ethnicity‐specific pathways between biomarkers and cognition
Author(s) -
Royall Donald,
Palmer Raymond
Publication year - 2013
Publication title -
alzheimer's and dementia
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 6.713
H-Index - 118
eISSN - 1552-5279
pISSN - 1552-5260
DOI - 10.1016/j.jalz.2013.05.480
Subject(s) - structural equation modeling , dementia , biomarker , mediation , cognition , ethnic group , psychology , gerontology , clinical psychology , developmental psychology , medicine , oncology , disease , biology , genetics , psychiatry , mathematics , statistics , sociology , anthropology , political science , law
Background: Structural equation models (SEM) can distinguish dementiarelevant variance in cognitive performance (i.e., "d" for dementia). However, it is an empiric question whether the association between any putative biomarker and cognition is mediated through d, independently of d, or through d’s residual in Spearman’s g (i.e., "g’ "). Only associations mediated via d will be dementing. We examine these issues in Hispanic and nonHispanic participants of the Texas Alzheimer’s Research and Care Consortium (TARCC). Methods: Data from n 1⁄4 2016 TARCC participants (n 1⁄4920 AD, n1⁄4277 MCI, and n1⁄4819 controls) were used to construct d, as previously described (Royall, Palmer & O’Bryant, 2012). Individual serum biomarkers were regressed onto d, g’ and each of their indicators. This effectively results in a " Multiple Indicators Multiple Causes" (MIMIC) model containing a set of nested mediation paths between the biomarker and each indicator (Figure 1). All indicator variables were adjusted for age, gender, and education. Additionally, serum biomarkers were adjusted for batch effects. The entire model was stratified by ethnicity (Hispanic n 1⁄4 537; Non-Hispanic N 1⁄4 1479). The models have acceptable fit (i.e., Figure 1: c 2 1⁄4 163.2:32, p < 0.001; CMIN/DF 1⁄4 5.1; CFI 1⁄4 0.981; RMSEA 1⁄4 0.045). Results: IL-10’s associations with cognition were entirely mediated through d in non-Hispanics. In contrast, Fatty Acid Binding Protein (FABP)’s associations were entirely mediated by g’. Vascular cell adhesion protein 1 (VCAM-1) was associated with cognition only through d, and only in Hispanics. R egulated and normal T cell expressed and secreted protein (RANTES) was directly associated with Boston Naming (r 1⁄4 0.127, p 1⁄4 0.009) in Hispanics and had no associations in nonHispanics. Conclusions: Using MIMIC models in an SEM framework, we can examine the pathways that link individual biomarkers to cognition. Only associations mediated by d are likely to be disabling, and therefore "dementing". Associations mediated either directly or through g’ may have no impact on Instrumental Activities of Daily Living (IADL). These assessments appear to vary by ethnicity, but might also vary with stage of illness.