König, A., Linz, N., Tröger, J., Bai, C.-H., Ritchie, C., Teipel, S., Dubois, B., Bombois, S., Teichmann, M., Robert, P., Palmqvist, S., & Hansson, O. (2021). PROSPECT-AD: Population-based screening over speech for clinical trials in AD.
Poster presented at CTAD 2021.
Background: Language and speech impairments are an early feature of neurodegenerative dementias. Consequently, digital biomarkers of language and speech performance may be promising tools for early diagnosis. The current “new era of Alzheimer’s disease (AD) clinical trials” suggests a shift to very early identification of people at risk. Hence, digital markers of language and speech may serve the screening of at-risk populations that are at a prodromal stage of AD, eventually in combination with advanced machine learning longitudinal modelling. Here, we conceived a pre-screening battery consisting of speech-based neurocognitive tests, enabling automated first-line pre-screening to be performed remotely using a telephone.
Objectives: PROSPECT-AD aims to build and validate speech-based machine learning models for the detection of the relevant phenotype through access to gold-standard phenotyped cohorts. Further, the predictive potential (sensitivity/ specificity) for differential/ prognostic diagnosis based on information extracted from the participant’s speech in cognitive vocal and narrative speech tasks and its usefulness for remote pre-screening and monitoring will be examined.
Methods: PROSPECT-AD collaborates with already ongoing cohorts such as EPAD (UK), DESCRIBE (Germany) or INSIGHT-preAD II (France) by adding the collection of speech data to existing protocols or as follow-up assessments over the telephone. Participants at preclinical stages are mainly recruited from existing parent cohorts across Europe to form a ‘probability-spectrum’ population covering the entire continuum of anticipated probability for Alzheimer’s dementia development. This characterization of cognitive, biomarker and risk factor (genetic and environmental) status of each research participants over time combined with audio recordings of speech samples will provide the necessary well-phenotyped population for developing predictive longitudinal models for Alzheimer’s disease covering the entire disease course and concurrently create a pool of highly characterized individuals for the validation analysis. 300 participants aged 50 or older will be included per cohort, with a clinical dementia rating scale (CDR) score of 0 or 0.5. The study protocol is planned to run over 18 months. The speech protocol includes the following tests which will be administered remotely: Word List, Story Retelling [Learning & Memory]; Digit Span, Phonemic Verbal Fluency, Semantic Verbal Fluency [Executive Functions], Spontaneous free speech [Psychological and/ or behavioral symptoms]. The spoken features extracted from the recordings will be compared to data from the neuropsychological evaluations, genetic profiles, biomarkers, neuroimaging, family history. Based on the analysis of vocal performances, models will be trained to predict participant’s risk to convert to AD dementia; employing advanced machine learning and different computational techniques to identify the most significant speech markers that could represent an early indicator for a pre-screening scenario.
Results: The overall study protocol is being developed and will be presented at the conference in addition to previous research findings and its resulting new hypotheses.
Conclusion: The outcome of PROSPECT-AD may have a major impact on the improvement of drug development research methodology by providing a validated telemedical solution for neurocognitive pre-screening and monitoring of participants of early AD clinical trials.