Clàudia Porta-Mas, Gonzalo Sánchez-Benavides, Elisa Mallick, Johannes Tröger, Nicklas Linz, Alexandra König, Andreea Radoi, Carlota Medina, Alba Cañas-Martínez, Anna Brugulat-Serrat, Lidia Canals-Gispert, Isabel Pérez-Gutiérrez, Marc Suárez-Calvet, Juan Domingo Gispert & Oriol Grau-Rivera
* Poster presented at the 17th Clinical Trials on Alzheimer’s Disease (CTAD), Madrid (Spain)
Abstract
Background: Automated remote assessments are promising tools for low-burden collection of cognitive outcomes in regionally diverse populations. Although digital literacy (use of touchscreens and internet-connected devices) is increasing, there are still some vulnerable groups in which such technologies cannot be reliably applied. Phone-based cognitive assessments delivered by automatic software systems that provide transcription of speech-based cognitive tests and evaluation of performance could be a feasible alternative to screen individuals at risk of cognitive decline regardless their digital literacy. In the present study, we explored the associations between automatically phone-delivered memory assessments, including memory process scores, and the intensity of Subjective Cognitive Decline (SCD) in a sample of cognitively unimpaired (CU) individuals.
Methods: We analyzed baseline data from the first 63 participants (mean[SD]age:66.4[6.6];39 women) from the B-AARC cohort included in the PROSPECT-AD study. B-AARC is a prospective research cohort composed of individuals with SCD (CU or Mild Cognitive Impairment [MCI]) that have sought or aim to seek for medical help. Only CU individuals were included in the present analyses. PROSPECT-AD1 is a multi-cohort European longitudinal study aiming to develop algorithms to identify speech biomarkers for Alzheimer’s disease (AD). Speech productions during cognitive tasks and free speech is collected via the Mili platform (ki-elements) over the phone. In the present study, the following automated scores obtained from 4 learning trials from the 15-word list Auditory Verbal Learning Test (AVLT) were analyzed: total learning (sum of the words recalled in the first 4 trials), primacy and recency indexes (percent of the first and last 5 words), learning slope and constancy (words in the last trial that have appeared at least once in previous trials). The intensity of SCD was measured in a face-to-face visit with the Cognitive Function Index (CFI) for both the participants (self-CFI) and the study partners (partner-CFI). We ran a set of exploratory linear models using memory outcomes as dependent variables and self- and partner-CFI as predictors. In further analysis we adjusted the models by covariates (age, education, and sex).
Results: We found that higher partner-CFI scores (beta=-0.91;p=0.027), but not self-CFI (beta=-0.14;p=0.73), were associated to lower AVLT total learning scores. Partner-CFI scores were also marginally associated to recency index (beta=0.68;p=0.088), with individuals with higher partner related complaints being more prone to recall the most recent words from the list. No associations were found for primacy index, learning slope and constancy. In adjusted models, the association between partner-CFI and total learning score mostly disappears (beta=-0.63;p=0.15), but the association with recency index remains unchanged (beta=0.73;p=0.086).
Conclusions: Automated memory assessments delivered by phone provide several metrics that seem related to the intensity of cognitive complaints in individuals without evidence of objective cognitive impairment. Although preliminary, our results suggest that this approach is feasible and able to capture memory phenotypes related with SCD intensity reported in AD, such as diminished learning and increased reliance on working memory rather than episodic memory. As the study and data collection continues, we will follow up with more advanced analyses on a bigger sample and CSF biomarker characterization.
References:
König A, et al. J Prev Alzheimers Dis 2023; 10(2): 314-321. https://doi.org/10.14283/jpad.2023.11