Abstract
BACKGROUND: Analysis of selected research cohorts has highlighted an association between plasma neurofilament light (NfL) protein and cross-sectional cognitive impairment as well as longitudinal cognitive decline. However, the findings have yielded inconsistent results regarding its possible application in clinical practice. Despite its potential prognostic significance, the role of plasma NfL in daily clinical practice with unselected patients suffering from cognitive impairment remains largely unexplored. METHODS: This retrospective, cross-sectional and longitudinal monocentric study enrolled 320 patients with Alzheimer's disease ([AD], n = 158), dementia with Lewy body ([DLB], n = 30), frontotemporal dementia ([FTD], n = 32), non-neurodegenerative diseases ([NND], n = 59) or subjective cognitive decline ([SCD], n = 41). Plasma NfL levels were measured at baseline on the Simoa platform. AD, DLB, and FTD patients were also analyzed altogether as a 'degenerative conditions' subgroup, whereas SCD and NND were grouped as a 'non-degenerative conditions' subgroup. We assessed the relationship between plasma NfL levels and cross-sectional cognitive performance, including global cognition and six specific cognitive domains. A subset of 239 patients had follow-up mini-mental state examinations (MMSE) up to 60 months. Models were adjusted on age, education level, glomerular filtration rate and body mass index. RESULTS: In 320 patients, baseline plasma NfL levels were negatively associated with global cognition (β=-1.28 (-1.81 ; -0.75) P < 0.001), memory (β=-1.48 (-2.38 ; -0.59), P = 0.001), language (β=-1.72(-2.49 ; -0.95) P < 0.001), praxis (β=-2.02 (-2.91 ; -1.13) P < 0.001) and executive functions (β=-0.81, P < 0.001). Across diagnosis, plasma NfL levels were negatively associated with cross-sectional global cognition in all but the SCD subgroup, specifically with executive functions and memory in AD (respectively β=-0.71(-1.21 ; -0.211), P = 0.005 and β=-1.29 (-2.17 ; -0.42), P = 0.004), and with attention in LBD (β=-0.81(-1.16 ; -0.002), P = 0.03). Linear mixed-effects models showed that plasma NfL predicted MMSE decline in the global population (β(PlasmaNfLxTime)=-0.15 (-0.26 ; -0.04), P = 0.006), as in the neurodegenerative condition subgroup (β(PlasmaNfLxTime)=-0.21 (-0.37 ; - 0.06), P = 0.007), but not in non-neurodegenerative condition subgroup. CONCLUSION: In our clinical cohort, plasma NfL was associated with faster cognitive decline in neurodegenerative dementia, which corroborates data obtained in research cohorts. Yet, plasma NfL was not predictive of accelerated cognitive decline in individuals without neurodegeneration, suggesting its use as a neurodegeneration-specific predictive biomarker.