Deling He, Rebecca E Langhough, Carol Van Hulle, Kristin E Basche, Erin M. Jonaitis, Rachael Wilson, Bruce P. Hermann, Henrik Zetterberg, Sterling C. Johnson, Kimberly D. Mueller
BACKGROUND: While language deficits are among the earliest detectable signs of Alzheimer`s disease (AD), no existing composites integrate connected speech - essential for capturing real-world communication - limiting precise detection and personalized interventions. METHODS: We analyzed cohort data from 824 non-demented adults and constructed three language composites: a theoretical composite based on expert knowledge, a connected speech-specific composite derived from the theoretical composite, and an empirical language composite optimized for predicting cognitive progression. We compared their longitudinal sensitivity to plasma tau phosphorylated at threonine 217 (p-tau217), neurofilament light (NfL), and glial fibrillary acidic protein (GFAP). RESULTS: The empirical language composite showed superior and consistent sensitivity (faster decline rate) with all three biomarkers. Proper name recall declined uniquely with increased p-tau217, while animal fluency declined selectively with higher NfL and GFAP. DISCUSSION: An empirical language composite captures holistic cognitive-communicative decline, while individual measures may imply a biomarker-specific language profile. Our findings support the clinical utility of plasma biomarkers and language-specific composites as sensitive early indicators of disease progression. HIGHLIGHTS: Empirical language composite had optimal sensitivity to plasma p-tau217, NfL, GFAP. It comprises words/minute, filled pauses, proper name recall, and animal fluency. It offers potential for language-dominant ADRD precise detection and monitoring.