During deep sleep, your brain's cells shrink by 60%, opening channels that flush out the proteins that cause Alzheimer's disease. Sleep is not rest. It is maintenance. We are in the middle of a global maintenance crisis.
In 2013, Maiken Nedergaard and her colleagues at the University of Rochester published a paper in Science that changed how neuroscience thinks about sleep. Using a new imaging technique that allowed them to watch cerebrospinal fluid moving through the living mouse brain in real time, they discovered something that had not been predicted by any existing model: during sleep, the brain's supporting cells — astrocytes — shrink by approximately 60%, opening channels between them through which cerebrospinal fluid flows in a rapid pulsing wave.
The fluid doesn't merely circulate. It flushes. It sweeps through the brain's interstitial spaces, carrying metabolic waste products — including amyloid-beta and tau, the proteins that aggregate into the plaques and tangles of Alzheimer's disease — out of brain tissue and into the body's lymphatic drainage system. Nedergaard named the system the glymphatic system: a portmanteau of glial cells and the lymphatic system it resembles in function. It is, in essence, the brain's sewage network, and it operates almost exclusively during deep non-REM sleep.
The implications were immediate and significant. Amyloid-beta accumulation had long been known to precede Alzheimer's diagnosis by decades — the protein begins depositing in brain tissue twenty years or more before cognitive symptoms appear. What was not understood was why. The glymphatic discovery suggested a mechanism: insufficient sleep impairs the system that clears amyloid-beta, allowing it to accumulate. In 2017, a study by Lucey and colleagues found that even a single night of sleep deprivation caused a measurable increase in amyloid-beta concentration in the human brain. The finding was replicated. The connection between sleep disruption and Alzheimer's pathology was no longer theoretical.
The glymphatic system is 10 times more active during sleep than during waking hours. Cerebrospinal fluid enters brain tissue along channels surrounding arteries, flows through the interstitium, and exits along channels surrounding veins — a directional flow driven in part by the slow, large-amplitude oscillations of deep NREM sleep. The arterial pulsations that drive CSF movement are larger and more synchronised during deep sleep, which is why the quality of deep sleep, not merely its duration, determines how effectively the brain is cleaned.
The clinical significance is compounded by a second discovery: that sleep deprivation is not an individual aberration but a structural feature of modern life. The CDC's surveillance data shows that more than one-third of American adults regularly sleep fewer than seven hours per night. Similar figures apply across most high-income countries. The causes are well-understood — artificial light at night suppressing melatonin, work schedules misaligned with circadian biology, the stimulant effects of caffeine and digital screens, and a cultural consensus that treating sleep as expendable is a mark of productivity rather than a biological error. Against the backdrop of glymphatic science, this consensus looks increasingly like a slow civilisational mistake.
Sleep is not a single state. It is a structured cycle of distinct neurological phases, each performing different functions, and the consequences of deprivation depend substantially on which phases are disrupted.
Non-REM sleep is divided into three stages of increasing depth. Stage 3 — slow-wave or deep sleep — is characterised by large, synchronised oscillations across the cortex, reduced heart rate and body temperature, and the highest activity of the glymphatic system. It is during this stage that the brain consolidates declarative memories, transferring information from the hippocampus to the cortex for long-term storage. Growth hormone is secreted almost exclusively during slow-wave sleep. The immune system performs much of its maintenance, including the production of cytokines that regulate inflammation. Deep sleep declines naturally with age — a 70-year-old may get less than half the slow-wave sleep of a 20-year-old — which aligns temporally with the window of elevated Alzheimer's risk.
REM sleep — the stage associated with vivid dreaming — serves different functions. Emotional memories are processed and consolidated during REM; the amygdala, which encodes fear and emotional responses, is highly active. Matthew Walker's research at UC Berkeley has shown that REM sleep strips the emotional charge from difficult memories — a mechanism that may underlie why sleep deprivation is so strongly associated with mood disorders, anxiety, and post-traumatic stress. Disrupted REM architecture is a consistent feature of PTSD. Alcohol, which suppresses REM sleep even at moderate doses, may worsen emotional dysregulation precisely through this pathway.
Sleep spindles — brief bursts of oscillatory neural activity during Stage 2 NREM sleep — are now understood to play a critical role in memory consolidation. They mark the moments when the hippocampus replays memories to the cortex for long-term encoding. People who generate more sleep spindles retain more of what they learned the previous day. The density of sleep spindles declines with age and is reduced in people with early Alzheimer's pathology, suggesting that the memory impairment of dementia may partly reflect a sleep architecture problem rather than a purely structural one.
The cardiovascular consequences of sleep deprivation are among the most robustly established in medicine. Short sleep duration — defined as fewer than six hours — is associated with a 20% elevated risk of heart attack, a significantly higher risk of stroke, impaired glucose metabolism that accelerates toward type 2 diabetes, and dysregulation of the hormones ghrelin and leptin that control appetite, which drives caloric overconsumption. Chronic partial sleep restriction does not produce the subjective experience of extreme sleepiness that complete sleep deprivation does — people adapt to feeling moderately impaired and rate their alertness as normal — while objective performance measures continue to deteriorate. This is among the more dangerous features of the sleep crisis: the people most impaired by it have the least awareness of their impairment.
We built modernity on the premise that sleep is time stolen from productivity. The glymphatic system suggests the opposite: that everything you do while awake is preparation for the maintenance that only happens while you sleep.
The relationship between sleep and Alzheimer's disease is now understood to be bidirectional — which means it is also potentially self-reinforcing in a way that is deeply concerning at a population level.
Sleep deprivation, as described above, impairs glymphatic clearance and allows amyloid-beta and tau to accumulate. But Alzheimer's disease itself disrupts sleep architecture — damaging the neurons in the basal forebrain that generate the slow-wave oscillations of deep sleep, and affecting the locus coeruleus, which regulates REM sleep cycling. The result is that Alzheimer's pathology impairs the sleep that would otherwise clear the proteins driving the pathology forward. A person in the early stages of Alzheimer's — which begin, neurologically, two decades before any cognitive symptom appears — is already sleeping less well, clearing amyloid-beta less effectively, and thereby accelerating the process that will, eventually, produce the disease they have no idea is developing.
A 2021 study published in Nature Communications followed nearly 8,000 adults in the UK over 25 years and found that consistently sleeping six hours or fewer at age 50 was associated with a 30% increased risk of developing dementia later in life, independent of other health and behavioural factors. A 2023 meta-analysis in Sleep Medicine Reviews pooled data from 21 studies and confirmed the association between both short and long sleep duration and elevated dementia risk, with the strongest signal for short sleep in midlife. The researchers noted that the associative evidence is now sufficiently consistent to treat sleep as a modifiable risk factor for Alzheimer's disease — alongside blood pressure, physical inactivity, and smoking.
This framing matters because Alzheimer's disease has no cure and no disease-modifying therapy with meaningful efficacy in clinical use. The anti-amyloid antibodies lecanemab and donanemab, approved by the FDA in 2023 and 2024 respectively, show modest slowing of cognitive decline in early-stage patients at significant cost and with serious side effects in a subset of users. They do not reverse existing damage. They do not work in moderate or late-stage disease. In that context, the possibility that improving population sleep quality could reduce Alzheimer's incidence is not a minor finding. It may be the most tractable intervention available — and it requires no drug development, no clinical infrastructure, and no regulatory approval. It requires only that the culture take sleep seriously.
The public health infrastructure around sleep is remarkably thin given the scale of the evidence. There is no globally coordinated effort to reduce sleep deprivation analogous to anti-smoking campaigns or cardiovascular disease prevention programmes. Workplace cultures in most high-income countries continue to implicitly reward those who work long hours at the expense of sleep. School start times in the United States remain misaligned with adolescent circadian biology despite decades of research showing that later start times improve academic performance, mental health, and physical safety. Shift work — which forces workers onto schedules incompatible with circadian rhythms — affects approximately 20% of the workforce in industrialised countries and is associated with elevated rates of cancer, cardiovascular disease, and metabolic disorders, yet is regulated primarily as a labour issue rather than a health one.
The practical implications of glymphatic science are not complicated. Deep sleep is most effectively protected by: consistent sleep and wake times aligned with natural light-dark cycles; reducing artificial light exposure, particularly blue-spectrum light, in the hours before bed; avoiding alcohol and cannabis, both of which suppress slow-wave sleep; maintaining physical fitness, which is one of the strongest predictors of deep sleep architecture; and treating sleep disorders — particularly obstructive sleep apnoea, which fragments deep sleep — as medical priorities. None of these interventions are inaccessible. Most of them cost nothing. What they require is a shift in the cultural frame from treating sleep as a variable to be compressed when other demands press against it, to treating it as non-negotiable biological maintenance.
The glymphatic system was only described in 2013. The science of what it does is still being established — there are open questions about its precise mechanics in humans (most of the direct evidence comes from mouse models), about the relative importance of glymphatic clearance versus other waste-removal mechanisms, and about the degree to which interventions that improve sleep architecture actually reduce Alzheimer's risk at the population level. What is not uncertain is the direction of the evidence. Every study that has looked at the relationship between deep sleep, glymphatic function, and Alzheimer's pathology has found the same thing: the brain that does not sleep deeply enough accumulates the proteins that cause Alzheimer's disease. This is not a story about what might happen. It is a story about what is already happening, every night, in one-third of the population.
© 2026 Lisa Pedrosa · lisapedrosa.com
All articles cited to primary institutional or peer-reviewed sources
Buy me a coffee