SIGNS OF EARLY COGNITIVE DECLINE WITHIN CONNECTED SPEECH: EVIDENCE FROM THE WISCONSIN REGISTRY FOR ALZHEIMER'S PREVENTION (WRAP)

Abstract

Alzheimer's disease (AD) is primarily characterized by early memory impairment; however, many patients show deficits in language across all stages of AD. Typical language testing is insensitive to early language problems that may manifest in discourse. We examined connected speech longitudinally in a group of late-middle-aged adults at risk for AD. We hypothesized that those with early Mild Cognitive Impairment (eMCI) would show decline in connected speech over time compared to cognitively healthy (CH) adults. First, we submitted connected speech measures from a picture description task to exploratory and confirmatory factor analyses (EFA, CFA) from a subgroup of cognitively healthy WRAP participants (n=399; mean age=60±7, 67% female). Next, we examined resulting factor scores as outcomes in a subset of participants with speech samples obtained at 2 time points (n=219), using a linear mixed effects model (LME), with fixed effects of time point and baseline cognitive status (CH/eMCI), and a random effect of intercept nested within subjects. Secondary analyses used logistic regression to investigate the effect of speech factors on cognitive status at latest visit. The EFA/CFA factor structure (Fig 1, Table 1) met goodness-of-fit criteria. Participant characteristics for longitudinal analyses are presented in Table 2. Cognitive status at time 1 was a significant predictor of Semantic (p =.007) and Syntax (p <.001) factors, with a significant interaction of time and cognitive status for the Semantic factor (p =.04) (Table 3, Fig.2). Syntax was a significant predictor of eMCI (p <.0001), as was change in Fluency (p=.01) (Table 4). This is the first study to demonstrate a confirmed factor structure of connected speech measures in a prospective AD-risk-enriched cohort. Evident are baseline differences between eMCI and CH in syntax complexity. Baseline syntax and change in Fluency were significant predictors of eMCI at the latest visit. Unexpectedly, participants with eMCI performed higher on the Semantic factor at baseline; however, they declined more steeply than the CH group over time. Future analyses will continue to examine longitudinal relationships among speech factors and subtypes of cognitive impairment (memory, executive function, language), which may contribute to early identification of and intervention for AD. Conceptual model submitted to confirmatory factor analysis using the cross-validation subsample(n=149). Note: CFA met the following satisfactory fit indices: χ2=35.63, CFI = .99, NNFI = .99, RMSEA=.04, 90% confidence interval=.00, .08, GFI=.98, AGFI=.97, SRMR = .07. Definitions: SUID is Semantic Unit Idea Density (number of information units/total words). Type-token ratio: number of unique words/total words. Pronoun index: total pronouns divided by nouns + pronouns. Maze Index: total number of revisions, repetitions, and filled pauses/total words. Verb Index: total number of verbs/utterances. Grammatical complexity: The number of grammatical relations that mark syntactic embeddings divided by the total number of grammatical relations. Density: ratio of the propositions corresponding to verbs, adjectives, adverbs, prepositions and conjunctions to the total number of words (excluding repetitions and fillers). Boxplots of Connected Speech Factor by Visit and Consensus Diagnosis at Speech Time 1 (unadjusted means).