Abstract Introduction Sleep duration is jointly regulated by the circadian pacemaker and a homeostatic process. Yet, classical temporal isolation experiments have repeatedly made a counterintuitive observation: longer episodes of wakefulness (α) tend to be followed by shorter episodes of sleep (ρ). This leads to a negative α:ρ correlation. Most strikingly, when circadian phase is accounted for the residual α:ρ correlation has been observed to be ~0. This observation contradicts traditional restorative hypotheses of sleep; however, no study has tested if this phenomenon generalizes to free-living humans. Methods To examine this, we analyzed ~450,000 sleep-wake episodes from 85,955 free-living older adults (62.3 ± 7.9 years, 58% female), using wearable-recorded light exposure and sleep data. Individual circadian phase (timing of core-body temperature minimum, CBTmin) was estimated using a validated model of the response of the central circadian pacemaker to light. Results We observed that longer prior wake duration was negatively associated with subsequent sleep duration (α:ρ β = –0.42, p 0.0001). Sleep duration exhibited marked variation with the predicted circadian phase of sleep onset: sleep durations were shortest (~3h) when initiated just after CBTmin, while sleep durations were longest (~11h) when initiated approximately 10 hours prior to CBTmin. A two-harmonic regression model revealed a peak-to-trough amplitude of 7.88 hours with an acrophase at –10.45 hours from CBTmin (p 0.0001). These relationships were independent of age, sex and employment. Notably, after accounting for the predicted circadian phase of sleep onset, the association between prior wake duration and sleep duration was substantially attenuated and only showed a very weak positive association (α:ρ β = 0.005, p 0.001). Conclusion Here, we translate a counterintuitive finding from humans living under temporal isolation to the real world. As in free-run studies, greater duration of prior wakefulness is correlated with shorter sleep duration and this correlation is explained by the circadian timing of sleep onset. Most strikingly, we observed that when accounting for the circadian phase of sleep onset, there is virtually zero association between the duration of prior wakefulness and sleep duration. Support (if any)
Burns et al. (Fri,) studied this question.