PARTICIPANT AND INFORMANT PREDICTION AND IDENTIFICATION OF COGNITIVE IMPAIRMENT IN PRE-CLINICAL ALZHEIMER’S DISEASE: IMPLICATIONS OF INFORMANT RELATIONSHIP

Abstract

Alzheimer's disease (AD) trials require enrollment of a participant and an informant. In preclinical AD trials, participants are cognitively unimpaired, raising questions of whether informants or participants provide more insight into participant cognition and whether the additional benefit of informant information differs by informant type. We used data from the Alzheimer's Disease Cooperative Study (ADCS) Prevention Instrument Project, a 48-month longitudinal study of cognitively unimpaired participants aged 75 years and older, to compare participant and informant ability to predict the cognitive performance of the participant. We used the Cognitive Function Instrument (CFI) to represent assessment of the participant's cognition and an altered version of the modified ADCS Preclinical Alzheimer's Cognitive Composite (mADCS-PACC) as an objective measure of cognition (Table 1). Random forests stratified by cognitive status and informant type (including mMMSE, CDR-SB, ethnicity, sex, cardiovascular disease, age, and education), were used to assess how well participants and informants predict cross-sectional and future cognitive status. When using baseline CFI scores to predict mADCS-PACC, participants performed better than their informants. In cross-sectional analyses, informant and participant CFI performed similarly. Dyad type impacted this relationship. Among spousal dyads, informant baseline CFI better predicted mADCS-PACC scores in years three and four. Among non-spousal dyads, participant baseline CFI better predicted future scores. In the fourth year, among spousal dyads in the lowest mADCS-PACC quartile, the observed variable importance (oVI) associated with participant baseline CFI was 0.13 (95%CI: 0.11, 0.14), compared to 0.19 (95%CI: 0.18, 0.21) for informant baseline CFI. Among non-spousal dyads, the oVI for participant baseline CFI was 0.37 (95%CI: 0.35, 0.41) compared to 0.07 (95%CI: 0.06, 0.11) for informant CFI. In cross-sectional models, oVI was 0.08 (95% CI: 0.07, 0.09) for participant and 0.21 (95% CI: 0.19, 0.23) for informant CFI among spousal dyads and 0.17 (95% CI: 0.16, 0.18) for participant and 0.10 (95% CI: 0.09, 0.11) for informant CFI among non-spousal dyads.