Brain clearance and slow waves

Updated 13 Oct 2026

One attractive frontier is measuring a restorative process more directly than a sleep-stage label. Brain-fluid transport is a candidate. The glymphatic hypothesis describes exchange between cerebrospinal fluid, the fluid surrounding the brain and spinal cord, and interstitial fluid within brain tissue. The proposed exchange helps move soluble material through and out of the brain.

This makes a precise question worth pursuing: can we measure a process that matters for brain health, improve it, and demonstrate a useful result? The evidence provides promising parts of that chain, with important gaps between them.

First define the quantity

The endpoint determines what an improvement would establish:

A local concentration can change through arrival, departure or redistribution. Protein concentrations also depend on production and degradation. The study’s endpoint and assumptions therefore belong beside its result.

For this project, a measurement should identify which part of that chain it observes. A proposed “brain-clearance score” would need a validation argument, a stable measurement protocol and a defined relationship to an outcome we care about.

Human sleep changes brain-fluid dynamics

Fultz and colleagues combined EEG with fast functional MRI in 13 people. During NREM sleep, slow electrical activity was coupled to changes in blood-related signals and waves of cerebrospinal fluid inflow. This demonstrated coordinated human physiology. It did not measure removal of amyloid, tau or other solutes, or test future dementia risk. NIH funded the study; three authors disclosed patent-related interests. Fultz et al., 2019.

A 2026 study used EEG and ultrafast MRI in 22 healthy volunteers to estimate the velocity and direction of physiological brain pulsations. Respiratory and vasomotor components increased in particular regions during NREM sleep, mainly N1/N2 in the selected segments, while cardiac-related velocity decreased. The analysis was calibrated against a flow phantom. Its endpoint was pulsation dynamics, rather than the amount of a metabolite removed. Elabasy et al., 2026.

These studies make breathing, vascular rhythms and neural activity interesting mechanisms. They also show why a single “lower is better” target is inadequate: different components change in different directions. A respiratory-rate estimate from a ring does not establish those intracranial dynamics.

Imaging closer to the pathway

CSF-STREAM is a contrast-free MRI method that isolates CSF movement from blood and tissue signals at high spatial resolution. Its 2025 human study characterized regional mobility and included a small visual-entrainment experiment that changed that signal. An August 2026 analysis mapped the spaces around cerebral arteries in 11 healthy participants, reusing earlier data. This provides a more anatomically precise measurement opportunity. The method cannot determine net flow or protein removal by itself. The MRI center disclosed Philips research support. Hirschler et al., 2025, Fultz et al., 2026.

A human tracer experiment

Eide and colleagues compared seven sleep-deprived participants with 17 matched controls allowed to sleep, following a tracer injected into the cerebrospinal fluid with repeated MRI. Clearance was slower after deprivation, and the difference persisted after the subsequent recovery opportunity within the study window. Participants were patients undergoing work-up for suspected CSF disorders; allocation was not randomized. This was a small, clinically selected population using an invasive, off-label tracer procedure. Eide et al., 2021.

The study supports concern about acute deprivation and molecular transport. It does not determine an optimum deep-sleep duration in a healthy young adult, or establish that recovery sleep can never compensate for lost sleep. Those broader propositions exceed its design and follow-up.

Protein production is another part of the process

A human pilot study used stable-isotope labeling and serial CSF samples to investigate amyloid-β kinetics across repeated sleep, deprivation and drug-induced-sleep conditions. Eight cognitively normal adults participated. Deprivation increased amyloid concentrations by about 25–30%; the kinetic model supported increased production, without a significant fractional-turnover difference. This provided a mechanism to investigate beyond fluid motion, while leaving individual clearance routes unresolved. NIH and foundation support, donated isotope and assay-related financial interests were disclosed. Lucey et al., 2018.

A later analysis of five participants from that work found higher CSF concentrations alongside lower plasma concentrations during deprivation. The authors proposed impaired transport to blood and acknowledged that production and clearance required further separation. It reused an earlier cohort, rather than supplying an independent clinical replication. Centene/NIH support and assay-related financial interests were disclosed. Liu et al., 2023.

This illustrates why the sample compartment belongs beside the result: a lower blood concentration can coexist with a higher CSF concentration. Neither direction alone establishes better brain health.

Breathing offers a controllable input

A 2025 human MRI study compared regular and slow deep breathing in 20 long-term breathing practitioners and 25 untrained controls. Changing breathing altered CSF displacement and estimated directional net flow during wakefulness. The order of breathing conditions was alternated across participants. This makes respiratory behavior an experimentally accessible input, rather than only a nightly statistic. The long-term training comparison was not randomized, and no solute-removal or cognitive benefit was measured. NIH and foundation support were acknowledged; the breathing organization helped recruitment, and the authors declared no competing interests. Lim et al., 2025, reporting correction.

The next question is whether changing those dynamics changes a meaningful transport or functional endpoint. A slower respiratory rate alone would leave that question unanswered.

Mouse experiments reveal a mechanism and a dispute

Hauglund and colleagues measured norepinephrine, vascular signals and fluid transport in mice. They linked slow norepinephrine oscillations during NREM sleep to vasomotion and glymphatic transport; experimental manipulations helped test the mechanism. Zolpidem suppressed aspects of the rhythm and flow. Funding included public agencies and research foundations; an author disclosed consultancy for unrelated work. Hauglund et al., 2025.

The relevant insight is that sleep’s internal organization can matter beyond time assigned to NREM. The findings do not establish a human medication’s effect on dementia or a target that a consumer ring can measure.

There is also a substantive methodological disagreement. Miao and colleagues inferred reduced clearance during sleep and anesthesia from fluorescent-dye experiments in male mice. Their tracer was placed in brain tissue, and the clearance inference relied on its distribution and a mathematical model. The paper received public and charitable support and declared no competing interests. Miao et al., 2024.

A published critique questioned whether tracer redistribution and the model could establish clearance, and whether injection and post-anesthesia sleep conditions affected the comparison. In reply, the authors defended their empirical clearance term, evidence for equivalent initial tracer amounts and interpretation of retention. A separate correction amended a source-data column. This is a dispute about how an endpoint is measured and inferred; the published exchange does not settle the whole glymphatic hypothesis. Plá et al., 2025, Franks and Wisden’s reply, source-data correction.

Can we improve a slow-wave process?

Closed-loop acoustic stimulation tests a promising intervention: time a sound to the brain’s ongoing slow oscillation, then assess what changes. In Ngo and colleagues’ principal experiment with 11 volunteers, phase-timed sounds enhanced slow oscillatory activity and next-morning word-pair retention. Sounds delivered at the wrong phase were ineffective. This provides a positive example of changing a process within sleep and observing a specific functional gain. The experimental protocol required EEG timing; ordinary background audio is a different intervention. Ngo et al., 2013.

A later study included a 19-person double-blind overnight replication. Stimulation enhanced slow-oscillation and spindle activity without improving the selected memory tasks. A separate nap experiment also found no memory benefit. The result makes the next research step clearer: measure the desired function alongside the physiological response, and investigate which protocols and populations benefit. Neither a changed EEG rhythm nor extra minutes labeled deep sleep can substitute for that test. Henin et al., 2019.

These experiments concern memory, rather than direct solute clearance. Their value here is to show that processes within a stage can be manipulated and that a successful physiological intervention still needs its intended outcome measured.

Emerging human devices: closer to the process, still a proxy

In 2025, Dagum and colleagues described an investigational in-ear device combining EEG with electrical-impedance measurements intended to reflect changes in brain parenchymal resistance. Two clinical studies compared sleep with deprivation; validation included contrast-enhanced MRI after intravenous injection. The reference measure itself required a model of contrast movement. Applied Cognition funded the study, and several authors disclosed financial or stock-option interests. Device study.

A 2026 paper analyzed plasma amyloid and tau alongside physiological recordings from those same cohorts. The abstract reports 39 participants; the results describe 38 in the current analysis. The raw overnight biomarker changes did not significantly differ between sleep and deprivation. The clearance interpretation emerged from multivariate models linking physiology, release and transport. This is more specific than a direct experimental demonstration that sleep increased every measured blood biomarker. Applied Cognition funded this work too, with financial and stock-option disclosures. Dagum et al., 2026.

These publications develop a possible measurement approach; they do not provide independent replications in wholly new samples. Clinical benefit, dementia prevention and a personalized optimum remain additional questions.

A pharmacological test of the proposed mechanism

A March 2026 preprint tested dexmedetomidine alone and ACX-02, a combination with midodrine. The ACX-02 cohort enrolled 22 older adults; 17 completed both visits. Treatment always preceded placebo. The intervention was monitored morning recovery sleep after overnight deprivation, and the registry was posted after study completion. Physiological signals and plasma biomarker ratios changed; the authors modeled those ratios as enhanced clearance. Most p-tau217 concentrations required below-detection imputation. These are important conditions for interpreting the finding, which included no cognitive or disease-prevention endpoint. Applied Cognition funded the study; several authors disclosed financial or stock-option interests. Preprint, official registry.

The experiment shows a route to testing several proposed drivers together, beyond simply increasing a stage total. Independent, prospectively specified studies with stronger allocation control and meaningful outcomes would determine what benefit that route can deliver. The broader intervention investigation considers other emerging approaches and their intended endpoints.

What would earn a place among my targets

The next step is a measurement and intervention question, rather than choosing a desired clearance number. I would require:

• A defined observable: which fluid, solute, region and time window does the measurement concern?
• Repeatability and validation: how much changes with fitting, placement, ordinary physiology or measurement error?
• An intervention test: can a feasible change improve the relevant process under controlled comparison?
• A functional connection: does that improvement accompany better memory, attention or another meaningful endpoint?
• An applicable population: does the evidence extend to a healthy young person with my circumstances?

Better sleep remains worth pursuing on the human performance evidence. Brain-clearance research expands what we can investigate. Until those measurement questions are answered, neither wearable deep-sleep minutes nor a blood biomarker should be translated into a personal brain-cleaning achievement.

All sleep investigations