Sleep in Long COVID

Sleep is not just one of the problems being managed in Long COVID — it is the medium through which most other interventions work. Pain sensitization, autonomic regulation, immune function, and cognitive capacity are all substantially modulated by sleep quality. A patient with chronically fragmented or non-restorative sleep will show limited responses to even well-targeted treatment for their other sub-syndromes. This is why sleep warrants early, active attention rather than being deferred until other problems are resolved.

The default clinical approach — sleep hygiene counseling plus a sedating medication — fails most of these patients not because it's wrong but because it's incomplete. It addresses one piece of a multi-component problem, and often not the most important piece.

The Core Problem: Non-Restorative Sleep

The thread that runs through most Long COVID sleep presentations is not insomnia or hypersomnia per se — it is sleep that does not restore. Patients describe waking after 8, 10, or 12 hours feeling no more rested than after 4. Duration is not the variable that needs optimizing; architecture is. The question is not how much sleep the patient is getting but whether the sleep they get is doing the physiological work it's supposed to do.

Most patients present in one of two patterns — often with overlap between them:

Parasomnias — vivid dreaming, nightmares, sleep paralysis, and REM sleep behavior disorder (acting out dreams) — occur at elevated rates post-COVID and may reflect brainstem involvement or sleep architecture disruption. When present, high-dose melatonin (3–12 mg) is the preferred first-line option; safety measures to prevent injury during episodes are a practical priority.

What Restorative Sleep Actually Requires

The glymphatic system — a network of perivascular channels through which cerebrospinal fluid flows during sleep to clear metabolic waste from brain tissue — is most active during slow-wave sleep and is substantially suppressed during waking. There is reasonable evidence that glymphatic flow is impaired in conditions involving poor sleep quality, and emerging interest in whether Long COVID-associated brain fog may partly reflect inadequate glymphatic clearance.

What this framework clarifies is why strong sedatives are often counterproductive as sleep aids in this population. Benzodiazepines, Z-drugs, and high-dose antihistamines do produce sedation — but sedation is not the same as restorative sleep architecture. These agents suppress slow-wave sleep and alter REM distribution, which may produce sleep duration without the glymphatic and consolidation processes that make sleep restorative. For a population where cognitive restoration during sleep is already compromised, trading architectural quality for sedation depth is a poor exchange. This is not an argument against all pharmacotherapy — it is an argument for choosing agents that support rather than suppress sleep architecture.

Barriers That Must Be Addressed First

The most important clinical insight in Long COVID sleep management is that sleep cannot be made restorative until the conditions that prevent restoration are addressed. Applying CBT-I or medication to a patient who is physiologically wired, in significant pain, or working a schedule incompatible with normal sleep architecture is unlikely to produce meaningful improvement — not because the interventions are wrong but because the ground isn't ready for them.

The barriers that most commonly need to be resolved before sleep work can succeed:

• Autonomic hyperactivation and adrenaline surges at night — Patients often describe a pattern of falling asleep adequately, then waking abruptly 1–3 hours later with palpitations, diaphoresis, and an adrenaline-like state they cannot reverse. This is a dysautonomia problem, not primarily a sleep problem, and it generally requires autonomic treatment before sleep intervention is effective.
• Rumination and cognitive hyperarousal — The hyperactivated state that prevents sleep onset in the mixed insomnia pattern is often resistant to behavioral techniques alone when the underlying autonomic and central sensitization burden is high. Addressing the adrenaline component first — with appropriate pharmacotherapy or autonomic stabilization — often makes behavioral techniques more tractable.
• Active pain and central sensitization — Sleeping in pain is nearly impossible to address with sleep-specific interventions. Migraine, widespread pain sensitivity, and allodynia disrupt sleep architecture independently of any insomnia mechanism. Pain management must be a parallel priority.
• Circadian schedule incompatibility — A patient on a night work schedule, an irregular shift schedule, or living in a household with unavoidable disruptions at their optimal sleep time cannot be expected to achieve restorative sleep regardless of what pharmacotherapy is prescribed. Acknowledging this constraint honestly — and working around it where possible — is more useful than optimizing medications in an incompatible environment.

Treatment: The Right Thing in the Right Order

Sleep treatment in this population fails most often not because the wrong tools are chosen but because they are applied in the wrong order. The sequence that tends to work has three steps, and the medications appear in two different roles within it.

First, treat the barriers. Pain, rumination, and dysautonomia are the conditions that hold restorative sleep out of reach, and each has its own targeted treatment — behavioral and pharmacologic. This is where the low-dose tricyclics earn their place: amitriptyline and nortriptyline (not doxepin, for the cognitive reasons noted above) are useful here not as hypnotics but as agents that lower pain and quiet rumination, and they carry the added benefit of dual utility in central sensitization. Autonomic stabilization for the nighttime adrenaline surges, and active pain and migraine management, belong in this same first step. Medications are genuinely helpful at this stage — but for the barriers, not yet for sleep itself.

Then, the sleep-specific behaviors. Once the ground is prepared, the behavioral interventions with the best long-term data — CBT-I, a fixed wake time held regardless of the prior night, and timed light therapy for circadian components — have something to work on. In sleep medicine these consistently outperform pharmacotherapy over the long run and do so without the architectural cost of sedating agents. Attempting them before the barriers are addressed is the most common reason they are judged, wrongly, to have failed.

Last, medications aimed at sleep architecture. Only once the barriers are controlled and the behavioral foundation is in place is it worth reaching for agents whose purpose is to improve the quality of sleep itself, and the selection principle is to support architecture rather than suppress it. Melatonin receptor agonists (ramelteon, tasimelteon) act on the circadian and sleep-onset machinery rather than sedating. Melatonin itself is best thought of as three different medications separated by dose. At very low evening doses (roughly 0.3–0.5 mg) it nudges circadian timing and may support sleep architecture without acting as a sedative at all. At medium doses (on the order of 1–3 mg) it takes on a mild sedative effect — but pushing past that tends to buy a groggy, hungover morning rather than deeper sleep. At higher doses (3–12 mg) it has a role in other problems such as parasomnias and restless legs, though tolerability is often limited by that same morning hangover carrying into the day. Orexin receptor antagonists (suvorexant, lemborexant, daridorexant) are the more mechanistically interesting option: they lower wake-promoting orexin signaling rather than globally depressing the brain, an approach more aligned with preserving natural architecture than the benzodiazepines, Z-drugs, and antihistamines discussed above. Sodium oxybate deserves brief mention as the one agent that reliably deepens slow-wave sleep, though its twice-nightly dosing, controlled-substance status, and safety requirements keep it a niche consideration rather than a routine one. It is worth being clear about the evidence: much of this is architecture-level reasoning rather than proven Long COVID outcome data, so the honest aim at this step is to improve sleep quality where possible while, at minimum, not making the architecture worse.

The Migraine Connection

Central sensitization and sleep have a bidirectional relationship that is clinically significant. Poor sleep lowers the threshold for central sensitization, increases pain sensitivity, and reliably triggers migraine in susceptible individuals. Active central sensitization, in turn, fragments sleep through night pain, arousal responses to sensory stimuli, and disrupted sleep architecture. Treating one without addressing the other tends to produce partial responses at best. Agents with dual utility — low-dose tricyclics, some anticonvulsants — are worth considering when both are prominent, because they address both targets simultaneously rather than requiring separate medications for each.

What Improving Sleep Changes

The clearest signal that sleep is genuinely improving is that other sub-syndrome symptoms begin to shift. Orthostatic heart rate increments on the NASA Lean Test decrease, pain sensitivity improves, and cognitive clarity increases. Sleep is not just one outcome — it is a mediating variable for most of the others. When it moves, it tends to move the whole picture. This is why addressing it early and specifically, rather than waiting until other problems are controlled, tends to produce disproportionate downstream returns.