Smartphone-based long-term delayed memory performance is associated with the Preclinical Alzheimer's Cognitive Composite and CSF levels of β-amyloid.

Abstract

Remote and unsupervised cognitive testing using smartphones may improve sensitivity to detect preclinical Alzheimer's disease (AD). In particular, cognitive testing at-home enables the evaluation of the potentially more sensitive long-term delayed memory retrieval, which is commonly not feasible in a clinical setting. Here, we investigated the association between remote and unsupervised smartphone-based cognitive assessments and in-clinic cognitive measurement in older participants (CDR = 0, MMSE = 29.2, age = 77). The sample (non-demented at age 70) was derived from the Swedish population-based Gothenburg H70 Birth Cohort Studies. In a subsample of participants, the retrospective correlation with cerebrospinal fluid (CSF) biomarkers analyzed at age 70 was further explored. To this point, 177 H70-participants have consented and were enrolled into the smartphone sub-study. They completed an "Object-in-Room Recall Test" (ORR) remotely and unsupervised using own mobile devices, and the recall session was initiated 24 hours after the initial learning phase. Available previously assessed CSF measures included the Aβ42/Aβ40-ratio, t-tau, p-tau, neurofilament light (NFL), and neurogranin, available in-clinic cognitive tests included the Rey Auditory Verbal Learning Test (RAVLT), Semantic Fluency, Digit-Symbol-Coding, and Mini Mental State Examination. The above-mentioned cognitive tests were used to derive the Preclinical Alzheimer's Cognitive Composite (PACC5), a composite designed to be sensitive to cognitive change in preclinical AD. All associations were examined using Pearson correlation coefficient. Preliminary analyses in a subset of participants yielded relatively strong positive correlations between PACC5 (R = .50, p<.001) and the ORR Long-term Delayed Retrieval. The strongest correlation for an individual test was observed between the RAVLT Delayed Retrieval and the ORR Long-term Delayed Retrieval (R = .52, p<.001). Regarding the CSF biomarkers, Aβ42/Aβ40-ratio at age 70 was positively correlated with the ORR Long-term Delayed Retrieval at age 77 (R = .43, p<.01), while measures of t-tau, p-tau, NFL, and neurogranin at age 70 were not. Our findings demonstrate the potential of smartphone-based cognitive assessment to detect features of preclinical AD in a population-based sample of older adults. Future analyses will address the relationship between these novel cognitive measures in relation to longitudinal CSF- and neuroimaging-based biomarkers and clinical progression.