Alzheimer's

Poster

Association of a Digital Speech Biomarker for Cognition (SB-C) with Functional Outcomes in Early Preclinical Alzheimer's Disease: Functional Equivalence to In-Clinic PACC Within the B-AARC Cohort

Alexandra König, Clàudia Porta-Mas, Andreea Rǎdoi, Elisa Malick, Johannes Tröger, Oriol Crau-Rivera, Gemma Salvadó, Gonzalo Sánchez-Benavides

*Poster presented at CTAD 2026

Introduction: Automated speech analysis has emerged as a scalable, low-burden approach for detecting early cognitive change in preclinical AD, but its relationship with patient-meaningful function remains unclear. This is an important gap, as decline in instrumental activities of daily living (IADL) is what patients and care partners consider most consequential, and is recognized as a clinically relevant outcome. In this study we hypothesized that a validated Speech Biomarker for Cognition (SB-C), administered via brief remote phone calls, would demonstrate associations with IADL-relevant outcomes in a population at increased-risk for AD.

Methods:

96 participants from the PROSPECT-AD study were included.

  • Participants completed a remote, automated, phone-
    administered speech tasks. Speech was processed using an AI-
    driven pipeline (ki:elements), generating the SB-C scores for overall cognition and sub-scores (memory, executive function,
    processing speed).

  • Functional outcomes were operationalized using a composite
    of ADL-anchored items from the Cognitive Function Instrument
    (CFI): driving, financial management, work, social involvement,
    following television and films, hobbies, and use of household
    appliances and electronics.

  • The functional outcome was defined as the presence of any
    functional difficulty across the seven items (any non-zero
    response) versus none.

  • The CFI was completed in-clinic by both the participant and by a
    study partner. The modified Preclinical Alzheimer Cognitive
    Composite (mPACC) was also administered in-clinic.

  • Associations between SB-C (global and sub-scores), mPACC and
    CFI-derived functional outcome were examined using logistic
    regression models, conducted separately for CFI from the
    study-partner and self-report.

  • Secondary analyses were repeated at the individual item level.
    CSF Aβ42/40 status was examined as a moderator through
    interaction models.

  • All continuous predictors were standardized, and models were
    adjusted for age, sex, and education. FDR correction was
    applied for the primary analyses.

Results: SB-C is associated with functional difficulty, on in partner-reports.

  • Based on study-partner report, 39 participants were classified as having any functional difficulty and 57 as having no difficulty. Based on self-report, 32 were classified as having any difficulty and 64 as having no difficulty.

  • Higher SB-C global cognition score was associated with lower odds of study-partner reported functional difficulty (OR = 0.56, pFDR = 0.028), with the strongest sub-score association observed for the memory score (OR = 0.55, pFDR= 0.027) (Figure 1A).

  • SB-C executive function and processing speed scores showed associations in the
    same direction that did not reach significance (Figure 1A).

  • PACC was also significantly associated with study-partner reported difficulty (OR = 0.48; pFDR = 0.027), with SB-C similar in magnitude to what was obtained with mPACC (Figure 1A).

  • In contrast, no SB-C and mPACC measures were associated with self-reported functional difficulty (Figure 1B).

  • Interaction models showed no significant interaction with Aβ status.

Secondary item-level analyses showed nominally significant associations for specific items from the partner-report: higher SB-C executive function score with lower difficulty following news/television (OR = 0.52; p = 0.037), higher SB-C memory score with lower disorientation (OR = 0.55; p = 0.04), and higher SB-C processing speed with lower work-related difficulty (OR = 0.21, p = 0.05).

Conclusion:

  • A brief, remote, automatically administered speech biomarker (SB-C) was associated with study-partner–reported functional difficulty, at a magnitude comparable to the in-clinic mPACC.

  • These findings extend speech-biomarker validation to functional outcomes recognized as clinically and regulatory-relevant in AD.

  • SB-C offers pharmaceutical sponsors a scalable, low-burden tool for remote monitoring, cohort enrichment, and functional endpoint capture in decentralized AD trials.

Related Articles