Banking Sleep and Biological Sleep Need

Dr. Vyazovskiy is supported by Wellcome Trust Strategic Award 098461/Z/12/Z. The authors have indicated no financial conflicts of interest. Notwithstanding its elusiveness, the concept of biological sleep need is highly relevant for understanding neurophysiological mechanisms underlying sleep timing and architecture, maximal capacity for sleep,1 homeostatic sleep regulation,2,3 sleep-dependent renormalization of optimal brain functions,4 and many other aspects. It is also central to the discussion concerning the amount of sleep needed for optimal health,5 especially given the widely held albeit debated view that the modern society is chronically sleep deprived.6–8 Investigating “sleep need” usually entails various methods of sleep curtailment, but experimental paradigms involving “sleep extension” have also been used in multiple studies.9–12 Notably, several studies have reported that extended sleep reduces sleep propensity and improves performance and mood,13–15 which raises an intriguing possibility that sleep does not merely enable a discharge of previously accumulated sleep need, but can provide a reserve of sleep, which could be utilized during subsequent waking. Relatively few studies, however, have addressed this hypothesis directly, and the study by Arnal and colleagues,16 in this issue of SLEEP, provides important novel insights. The first paper suggesting the possibility of “sleep banking”17 came as a surprise for most sleep researchers. In a well-controlled study where 24 subjects were randomly assigned to sleep either 7 or 9 hours for a week, it was shown that the latter group was more resilient to subsequent chronic sleep restriction when sleep timing was restricted to 3 hours per night. In addition, the data suggested a faster recovery of several neurobehavioral variables in subjects who had obtained “excess sleep” prior to sleep restriction.17 The study by Arnal16 provides further support for the notion that sleep extension improves the tolerance to total sleep deprivation. This was manifested in a reduced number of PVT performance lapses, faster psychomotor speeds and prolonged sleep latency on the MSLT during sleep deprivation after a period of 6 days, where 10 out of 24 hours were spent in bed, as compared to after a similar period with habitual sleep. Some of these effects persisted on the following “recovery” day. Furthermore, the intrusions of microsleeps during wakefulness were less frequent in subjects from the extended sleep group, as compared to habitually sleeping control subjects. Consistent with the study by Rupp et al.,17 sleep extension did not affect subjective sleepiness. While the results are interesting, the question remains whether the effects observed provide unequivocal evidence for “banking sleep,” and several important aspects need to be considered before more assertive conclusions can be drawn. These concern the experimental design, the parameters under scrutiny, and how the data are interpreted in light of the underlying regulatory mechanisms. The two-process model18 appears to be a useful framework for understanding the results reported by Arnal, even though it is traditionally used to describe the dynamics of sleep-wake dependent Process S on a relatively short time scale of ∼24 hours, rather than several days. Importantly, this model implies an existence of a stable set-point, such that any deviation from a certain level of sleep (e.g., sleep deficit or excess) would trigger a proportional increase or decrease in sleep amount and intensity, until the equilibrium, over a certain time span, is reestablished. The notion of sleep banking therefore may appear inconsistent with this fundamental postulate, as it implies that the set point is altered. To our knowledge, existing data neither demonstrate convincingly, nor rule out this possibility, and even if the shift of the homeostatic set point was to ensue as a result of a prolonged period of sleep extension, it remains to be determined how many days of extended sleep is necessary and for how long the effect would persist. Moreover, it is reasonable to assume that the return to habitual sleep time could rapidly restore the original set point. An alternative scenario is a temporary change in the parameters of Process S, which would be reflected, for example, in a slower build-up of sleep pressure during wakefulness following a period of sleep extension. The putative existence of such “metaregulation” of the homeostatic process would be advantageous, because physiological variations of the parameters of Process S within a broad dynamic range, rather than being “fixed” or “hardwired,” would enable both stability and flexibility with respect to preceding history, current demands and anticipation of future states. Yet, the simplest scenario that needs to be considered is that sleep extension merely reduces the initial levels of sleep pressure at the beginning of sleep deprivation, resulting in subjects spending longer time in a “comfort zone” of reduced sleep pressure. Previous studies are consistent with this notion,13–15 and Arnal16 found that at baseline sleep extension reduced PVT lapses, improved psychomotor speed, and reduced sleep propensity. Intriguingly, the difference between the groups in PVT response speed, which was apparent during baseline, effectively diminished in the course of sleep deprivation, suggesting that the rate of deterioration is even faster after sleep extension, although the subsequent re-emergence of a faster response speed after recovery suggests an existence of some “long-term” benefits of previously cumulated sleep. A merit of the study by Arnal is that the instruction to adhere to a habitual sleep schedule with at least 8 hours in bed was more likely to allow sufficient time for most subjects to obtain their desired amount of sleep, as compared to ∼7 hours only in the study by Rupp.17 However, the finding that the subjects had increased slow wave sleep in the sleep extension condition suggests that even 8 hours in bed was still not enough for saturating sleep need fully in the habitual sleep condition. Indeed, subjective estimation of a habitual sleep need is often not accurate when compared against objective measures, and there is good support for many individuals being too optimistic in their judgement of how much sleep they need.10 Asymptotic sleep duration has been reported to be as long as 8.2 hours19 or 8.7 hours9 in healthy young men. Thus, it is possible that some of the effects reported in sleep banking studies are to some extent confounded by insufficient sleep affecting most or at least some of the individuals in the habitual sleep condition. There is little doubt that sleep affects subsequent waking and vice versa, but the critical question is how extra sleep changes the symmetry and dynamics between them. Specifically, it remains to be determined how efficient late sleep is (in comparison to early deep sleep) in making a difference with respect to discharging sleep need, creating a “positive sleep balance” and altering the build-up rates of sleep pressure or restoring neurobehavioral functions during subsequent waking. Future “sleep banking” studies involving sleep/wake EEG and mathematical modelling of Process S will likely provide important insights. One animal study showed that sleep-wake history prior to 6-h sleep deprivation was related to the magnitude of increase of slow wave activity (SWA) during subsequent sleep,20 suggesting that the initial levels of sleep pressure are an important determinant of the dynamics observed relatively far into the future. The fact that extended sleep contains little SWA12 raises the question of the underlying neuronal/network mechanisms. One possibility is that low SWA during late sleep reflects an occurrence of predominantly local ON and OFF periods in neuronal spiking, corresponding to smaller amplitude slow waves.21 However, even when SWA is low, it is still possible that individual neurons and small cortico-subcortical networks undergoing local OFF periods benefit from sleep.22 On the other hand, even if extra sleep is not contributing actively to the dissipation of sleep need, it may still do it indirectly by simply preventing wakefulness, thus delaying manifestation of any potentially deleterious consequences on performance.23 In summary, the existing literature provides tentative support for the concept of “banking sleep,” but it remains to be determined whether the underlying mechanisms and temporal dynamics are compatible with the tenets of the two-process model. From a mechanistic perspective, future studies of sleep banking will benefit from more detailed EEG analysis, both during waking and during sleep, brain imaging studies, and mathematical modelling.18,24,25 Further studies where the effects of sleep banking are compared with other manipulations, such as napping or caffeine, are undoubtedly necessary,26 especially given that potential negative effects of sleep extension cannot be excluded.27 There is also a need to address practical aspects, such as how much extra sleep is needed to yield noticeable benefits for subsequent wakefulness.