Actigraphy validation in behavioral variant frontotemporal dementia

Background: Actigraphy is increasingly being used to assess sleep in patients with neurodegenerative diseases. However, information on its accuracy relative to polysomnography (PSG) in this clinical population remains scarce. This study investigates the performance of actigraphy compared to PSG in patients with behavioral variant frontotemporal dementia (bvFTD), which is the leading form of early-onset dementia. Methods: Eighteen patients with bvFTD (10 males, mean age 70.50 ± 8.48 years) underwent overnight, in-home PSG while concurrently wearing an actigraph on their non-dominant wrist. Actigraphy performance was assessed through discrepancy analysis, Bland-Altman plots, and epoch-by-epoch analysis (EBE). Analyses were conducted separately for the Cole-Kripke and UCSD scoring algorithms. Results: Discrepancy analysis highlighted that the Cole-Kripke and UCSD algorithms overestimate total sleep time (by 43 and 60 min, respectively) and sleep efficiency (by 7.13 % and 10.33 %, respectively). The Cole-Kripke algorithm also overestimates sleep onset latency (by 7.75 min). Wake after sleep onset (WASO) showed a negative proportional bias for both algorithms, indicating that actigraphy underestimates WASO for subjects with longer PSG-measured WASO. In the EBE analysis, the Cole-Kripke algorithm shows an accuracy of 84 % (sensitivity 93 %, specificity 62 %) and the UCSD algorithm an accuracy of 85 % (sensitivity 96 %, specificity 57 %). Conclusions: In patients with bvFTD, actigraphy significantly overestimates total sleep time, sleep latency, and sleep efficiency, while underestimating WASO. Clinicians and researchers using actigraphy to study sleep in bvFTD must carefully consider these measurement biases and correct for them based on the results of previous comparison studies.