[P4–179]: ASPECTS OF SUBJECTIVE COGNITIVE DECLINE THAT PREDICT INCIDENT MCI IN THE MAYO CLINC STUDY OF AGING

Previous evidence suggests subjective cognitive decline (SCD) may be an indicator of underlying Alzheimer's disease. Whether certain aspects of SCD provide more optimal case finding than others is still unclear. Therefore, we aimed to investigate which way of assessing subjective cognitive decline (SCD) best predicted incident MCI in cognitively normal (CN) participants from the general population. A total of 1167 CN individuals, aged 70–95, from the Mayo Clinic Study of Aging were included based on the availability of 2 SCD scales (5 questions adapted from the Blessed memory test and the Everyday Cognition scale [ECog]) and at least 1 subsequent follow-up. The ECog includes 2 global (multi-domain) scores and scores for four separate SCD domains (memory, language, visuospatial functioning and executive functioning). It is administered to the participant and the informant. Cox proportional hazard models with age as time variable were used to predict incident MCI in models adjusted for objective memory performance, depression, anxiety, sex, APOEe4 carriership and physical comorbidities. Predictors included the above-mentioned scales and concern (yes/no) about cognitive abilities. SCD was very commonly endorsed (any SCD on Blessed 860 [73.7%] and ECog 1071 [91.8%]), but severe SCD was rare. Any SCD predicted incident MCI using the Blessed scale, but not the ECog. In continuous models, all ECog domains and both global ECog scores predicted incident MCI with a small advantage for global SCD (ECog global HR 2.18 [1.36–3.50]) and memory SCD (ECog memory HR 1.99 [1.45–2.74]). Overall, informant-based SCD performed slightly better than self-perceived SCD (iECog global 2.78 [1.77–4.36]). In combined models concern (HR 1.60 [1.10–2.32]) and memory SCD (HR 1.96 [1.39–2.76]) were independent predictors of incident MCI, while global SCD only remained predictive in unconcerned participants (HR 2.43 [1.30–4.52] vs. 1.38 [0.61–3.14]). SCD is common and highly depends on the questions asked. Predictive value can be improved by including questions regarding concern and severity of SCD, especially in the memory domain. Other SCD domains also contribute in those who are not concerned about their cognitive abilities. Importantly, endorsement of SCD by an informant may be of greater value than self-perceived SCD.