People who have had a traumatic brain injury sleep badly. A meta-analysis of the prevalence studies estimated that about half report some form of sleep disturbance, with insomnia, hypersomnia and sleep apnoea all more frequent than in the general population (Mathias and Alvaro, 2012). The problems appear after mild as well as severe injury, they can last for years, and they are associated with slower recovery, worse mood and poorer cognition (Ouellet et al., 2015). For a long time they were treated as a consequence of injury that deserved management in its own right and little more.

A second observation has given them greater weight. Repeated head injury is associated with a distinctive tauopathy, chronic traumatic encephalopathy, in which the earliest deposits lie around small vessels at the depths of the cortical sulci (McKee et al., 2013). The perivascular position of those deposits invites a mechanical explanation, and the glymphatic model, proposed in 2012, supplies one. If the brain clears protein waste along perivascular routes, if that clearance depends on sleep, and if injury damages both, then disturbed sleep after injury is a candidate mechanism of later disease and not only a symptom. This review asks how much of that chain the evidence supports.

The article is a narrative review of published work. It takes the argument in stages: the glymphatic model and the objections to it, the nature of sleep disturbance after injury, the experimental and human evidence that connects them, and the implications for treatment and research. Animal and human findings are kept apart throughout, because most of the mechanistic evidence comes from anaesthetised rodents and most of the clinical evidence is indirect. Reviews by Rasmussen et al. (2018) on the glymphatic pathway in neurological disease and by Sandsmark et al. (2017) on sleep after injury cover the neighbouring ground in more detail.

The model has three parts, each resting on different evidence and each open to a different objection.

Sleep disturbance after injury is not one condition. In the first weeks, increased sleep need and daytime sleepiness predominate. Later, insomnia becomes more prominent, along with circadian rhythm disorders and sleep-disordered breathing (Ouellet et al., 2015). Polysomnographic studies have reported reduced sleep efficiency and altered sleep architecture, though findings on slow-wave sleep are inconsistent across studies and between patients and animal models (Sandsmark et al., 2017). Pain, medication, mood disorder and the hospital environment all contribute, which makes it hard to isolate the effect of the injury itself.

Why injury disturbs sleep is only partly understood. Diffuse axonal injury affects the brainstem and hypothalamic nuclei that control arousal, and a post-mortem study found fewer hypocretin neurons in the hypothalamus of people who had died after severe head injury (Baumann et al., 2009). The observation is of interest here because the same peptide, also called orexin, influences interstitial amyloid-β in mice (Kang et al., 2009). Damage to one small group of cells could in principle affect wakefulness and protein dynamics together.

The inconsistency matters for the present argument. The glymphatic hypothesis is specific about which aspect of sleep counts: slow-wave activity, with its low noradrenergic tone and large slow oscillations. A patient who sleeps for more hours with less slow-wave activity, and a patient who sleeps for fewer hours with normal architecture, would be predicted to differ in clearance, yet both are recorded as having disturbed sleep in most clinical studies. Questionnaire measures, which dominate the literature, cannot distinguish them.

The evidence divides by species. In rodents, perivascular transport can be measured directly, at the price of anaesthesia and invasive tracer delivery. In people it has to be inferred.

Several authors have drawn these strands into a self-reinforcing loop (Piantino et al., 2019), shown in Figure 1. Injury disturbs sleep and, separately, disrupts astrocytic water channels. Poorer sleep and damaged perivascular pathways both reduce the removal of solutes. Retained tau and amyloid-β, with the inflammatory response they provoke, impair neuronal function, including that of the brainstem and hypothalamic systems that regulate sleep. Worse sleep then feeds back on clearance.

The diagram leaves out influences that act on several steps at once. Age, vascular disease and sleep apnoea are each associated with poorer sleep, and each could plausibly alter fluid movement through effects on arterial pulsation or on astrocytes. All three are common among older adults admitted after falls, a group that makes up a growing share of head injury. A study that finds worse sleep and worse clearance markers in injured than in uninjured people has to show that these shared influences do not account for the association. Few of the available studies were designed to do so, and the groups in whom imaging markers have been examined after injury are small and mostly studied at a single time point.

The loop is coherent, and each arrow has some support. It should still be read as a hypothesis. The arrows from injury to reduced exchange and from reduced exchange to tau pathology rest on mouse experiments. The arrow from sleep loss to protein retention is supported in people, with the caveat about production noted above. The arrow from retained protein back to disturbed sleep is the least documented in the setting of injury. No longitudinal human study has yet followed sleep, a clearance measure and protein burden together after trauma.

If the loop is real, treating sleep after injury would protect the brain as well as relieve a symptom. That prospect has encouraged interest in sleep hygiene, cognitive behavioural therapy for insomnia, treatment of sleep apnoea and the timing of sedative drugs in this population. All of these have their own justification, since better sleep improves daytime function whatever happens to clearance (Ouellet et al., 2015). What cannot yet be claimed is that any of them alters glymphatic function or lowers the risk of later neurodegeneration. No trial has tested that outcome, and the choice of hypnotic drug on glymphatic grounds would be premature, given that the effects of different sedatives and anaesthetics on fluid transport are themselves disputed.

There is also a question of when to act. The mouse data suggest that perivascular function is depressed for weeks after a single injury. In patients this is the period in which sleep is most irregular and in which hospital routines, pain and noise interrupt it most often. If clearance matters, the early weeks would be the time to protect sleep, by reducing disturbance at night on the ward and treating pain adequately. Those measures are inexpensive and are justified on ordinary grounds of care. They deserve mention because they do not depend on the hypothesis being right.

Three kinds of study would move the field. The first is measurement: agreement on which human marker of perivascular function to use, with validation of the non-invasive indices against intrathecal tracer studies in the same individuals. The second is longitudinal observation after injury that pairs objective sleep recording, including slow-wave activity, with repeated imaging and fluid biomarkers, so that the order of events can be established. The third is intervention: trials of sleep treatment after injury that include a clearance marker among their outcomes.

Work in animals has gaps of its own. Most experiments use young male mice under anaesthesia, whereas the clinical population is of both sexes and all ages, and anaesthetic agents differ in their effect on tracer movement. Whether repeated mild injury, the pattern linked to chronic traumatic encephalopathy (McKee et al., 2013), produces cumulative impairment of clearance is a question the existing models are suited to answer and have only begun to address.

The proposal that sleep disruption after traumatic brain injury impairs the clearance of waste proteins, and so contributes to later disease, draws on real findings: perivascular solute transport exists, injury damages the astrocytic machinery thought to support it in mice, and sleep loss raises amyloid-β and tau in people. The chain has not been demonstrated end to end in injured patients, and its central physiological claims are still argued over. Sleep problems after injury deserve treatment for the distress and disability they cause. Whether treating them also protects the brain is a question for studies that have yet to be done.

To be declared by the author(s) before publication.

To be declared by the author(s) before publication.