Plasma Neurofilament Light (NfL) associated with retest learning effects on ambulatory cognitive assessments of processing speed: Results from the Einstein Aging Study (EAS)

Background Measurement burst designs, using ambulatory cognitive assessments facilitate the dissociation of retest effects from optimal (“asymptotic”) performance. Dissociating these features via cognitive process modeling may reveal indices of subtle cognitive impairment that signal risk for neurocognitive decline. We investigated cross‐sectional associations between plasma NfL (a biomarker of neurodegeneration), retest improvements, and optimal performance on the Symbol Match task, which assesses processing speed by response time (RT), in a racially/ethnically diverse community‐based EAS Study cohort. Method Analyses included 283 EAS participants free of dementia (Mean age = 77.4 years, SD = 4.8, Range = 70‐91; 67% Female; 46% Non‐Hispanic White, 40% Non‐Hispanic Black). Among them, 82 met the Jak/Bondi criteria of mild cognitive impairment (MCI). Participants completed a smartphone cognitive battery, resulting in up to 96 assessment sessions (six times per day for 16 days). Plasma NfL concentration was measured using a Single molecule array assay. A mixed‐effect negative exponential model was used to examine the associations between plasma NfL and indices of subtle cognitive impairment adjusting for covariates including age, gender, education, and race/ethnicity. Retest improvement was measured by the exponential decrease in RT across sessions and optimal performance was measured by asymptotic RT (fastest performance). Result On average, the RT of the Symbol Match task was 3.4 seconds (SD = 0.9). Mean plasma NfL level was 23.9 pg/mL (SD = 17.5). Higher plasma NfL levels were associated with a reduced retest improvement rate per session (estimate = ‐0.011, 95% confidence interval (‐0.013, ‐0.008) corresponding to one SD increase in plasma NfL, p<0.001). No significant association between plasma NfL and optimal cognitive performance was found. Stratified analyses showed NfL’s association with retest improvement was larger in non‐Hispanic Whites than in non‐Hispanic Blacks (‐0.009 vs ‐0.002), although formal testing of the interaction between NfL and race was not significant. Results remained very similar excluding participants with MCI. Conclusion Higher plasma NfL was associated with a reduced retest improvement rate on a smartphone processing speed task, highlighting its potential value for identifying subtle cognitive impairment. Longitudinal data from larger samples, a broader array of biomarkers and of ambulatory cognitive tests, are required to fully explore these issues.