Sensitivity of objective and subjective sleep features to tau and Aβ burden in healthy older adults

Abstract Background Impaired sleep is a risk factor for Alzheimer’s disease (AD) as well as a symptom of AD pathological progression. Recent work has established associations between disrupted sleep electrophysiology and AD pathology before the onset of AD symptoms. However, it is not understood how real‐life sleep features, both objective and subjective, are associated with tau and Aβ burden in the context of early AD. Method We compared objective and self‐reported measures of sleep quantity and quality to 18 F‐FTP tau PET and 11 C‐PIB Aβ PET in healthy older adults (N=80, age 78.9 ± 6.9, 65% female). Objective sleep was assessed using 7 nights of at‐home wristwatch actigraphy, with analyses focusing on sleep duration, efficiency, fragmentation and timing (midpoint). Subjective sleep was assessed using the Pittsburgh Sleep Quality Index. 18 F‐FTP SUVRs (80‐100 min, inferior cerebellar gray reference) were calculated for each subject in a composite ROI of temporal cortical regions. A global cortical 11 C‐PIB DVR (0‐90 min, cerebellar gray reference) was also calculated. Result Shorter average sleep duration was associated with significantly greater cortical Aβ (Figure 1A; r=‐.23, p=0.04). In contrast, worse sleep efficiency, rather than duration, was associated with greater tau burden (Figure 1B ; r=‐.24, p=.03). Of clinical interest, neither self‐reported sleep duration or efficiency were associated with tau and Aβ burden. Indeed, greater tau and Aβ burden were associated with worse accuracy in self‐reported estimation of sleep, relative to objective amount (Figure 2; r=0.23, p=0.04). Conclusion These findings suggest that specific actigraphy‐based measures of sleep, assessed in the ecological at‐home setting, are differentially predictive of tau and Aβ burden. This dissociation may reflect distinct effects of these anatomically segregated pathologies in early stages of AD and their respective impact on different aspects of sleep‐regulating brain networks. Of translational relevance, participants’ own subjective sleep estimates did not predict tau or Aβ burden, suggesting that objective sleep assessments may be more clinically appropriate, reflecting real‐life behavior patterns in the context of aging. Finally, the severity of both tau and Aβ burden predicted the degree of inaccuracy in participants’ assessment of their own sleep duration, the underlying neural mechanisms of which remain unclear.