AbstractBackgroundThere is an urgent need for accessible biomarkers that can predict cognitive impairment in older individuals. There is also a complete lack of biomarkers for TAR DNA‐binding protein‐43 (TDP‐43), a frequent and important neurodegenerative pathology in older individuals, currently referred to as Limbic‐predominant Age‐related TDP‐43 Encephalopathy‐Neuropathologic Change (LATE‐NC). Here we explored the utility of linguistic markers for prediction of future cognitive impairments and LATE‐NC.Method134 participants of the Leisure World cohort, an epidemiologic study in a California retirement community, produced written descriptions of the cookie‐theft picture from the Boston Aphasia Diagnostic Examination. All had normal cognition at the time. Samples were quantitatively analyzed for 31 linguistic markers (see table 1 for definitions of select relevant variables). Participants subsequently joined the 90+ study and underwent cognitive assessments every 6 months and autopsy after death. Cognitive impairment and its onset date were determined at postmortem conferences. LATE‐NC was considered present in those with TDP‐43 in hippocampus or neocortex, and Alzheimer’s disease neuropathologic change (ADNC) in those with high likelihood ADNC according to NIA‐ AA guidelines. To examine the relationships between linguistic markers and future cognitive impairment and presence of LATE‐NC and ADNC, we used logistic regression analysis. Models were adjusted for sex, education, age at the time of writing the samples and age at death.ResultTable 2 summarizes demographic and neuropathology findings. Additionally, mean age at cognitive impairment was 93.2 in 72% who developed cognitive impairment. Future cognitive impairment was significantly associated with less information content (OR:0.02) and fewer clauses (OR:0.93) (Table 3). Also, presence of LATE‐NC was significantly associated with higher proportions of content words (OR:7.2), higher proportion of function word errors (OR:10.5), and fewer complete units (OR:0.15). Lower number of words also trended towards a significant association (OR:8, p = 0.07) with LATE‐NC. No language measure was associated with ADNC.ConclusionOur results suggest that linguistic markers derived from writing samples obtained on average 8 years prior to development of cognitive impairment and 12 years before death might serve as biomarkers for future cognitive impairment and presence of LATE‐NC at autopsy.