Sleep Apnea: Why Breathing Stops at Night

The part of your airway that fails at night has no bone and no cartilage holding it open — it is a soft tube kept from collapsing by muscle tone alone. Every night that tone drops away, and in some people the tube shuts. Press play, then switch to Obstructive and watch the whole sequence: the pharynx narrows, the uvula starts to flutter, the airway seals, the chest keeps heaving against nothing, the oxygen drains out of the blood, and the brain finally throws an emergency switch to reopen it. Then do it again with CPAP on.

Try this: start on Obstructive and let one full event run — note the SpO₂ nadir. Now press 🍷 Alcohol / sedative and watch the same event get longer and the nadir get deeper. Then press 🚸 CPAP on and watch four traces that were a disaster go flat and clean in a single breath. Finally switch to Central with CPAP still on — the airway stays wide open and breathing still stops, because the problem was never the airway.

Speed Diagram is illustrative — not to scale.

Live sleep-study readout

97 % SpO₂

Nadir this event: 97% · normal 95–100%

0.0 s event

No event — airway open

0.0 AHI /hNormal

Real bands: <5 normal · 5–15 mild · 15–30 moderate · >30 severe

58 bpm

Steady sleeping rate

2.4 arousals /h

Each arousal is a few seconds of wakefulness you will not remember.

21 % REM reached

A healthy adult night is 20–25% REM. Fragmentation eats it first.

What's happening

Quiet nasal breathing. Pharyngeal dilator muscles are holding the soft tube of the throat open. Air moves freely; oxygen stays at 97%.

Airway patent. No scored events.

inhaled air exhaled air obstruction central pause

What is real: the AHI severity bands (<5 normal, 5–15 mild, 15–30 moderate, >30 severe), the scoring definitions (an apnea is a ≥90% drop in airflow lasting ≥10 s; a hypopnea is a ≥30% drop for ≥10 s with a ≥3% or ≥4% desaturation or an arousal, depending on the rule used), the typical 10–40 s event length, normal SpO₂ of 95–100%, the 20–25% of a normal night spent in REM, and the usual CPAP working range of about 5–15 cmH₂O. What is modelled: every number that moves on this page. The SpO₂ curve, the heart-rate swing, the arousal index and the REM figure are illustrative model output tuned to behave like a real recording — they are not measurements from a patient. The animation also packs events far closer together than a real night so you do not have to wait; the AHI readout is scaled back to a realistic per-hour rate rather than counted off the compressed clock.

The Science in Plain Language

The one stretch of your airway that has nothing holding it open

Run a finger down the front of your neck. You can feel the hard ridges of the larynx and, below it, the ribbed tube of the trachea. Those rings are cartilage, and they exist for exactly one reason: to stop the airway from collapsing when you suck air through it. Now go upwards, above the larynx, into the pharynx — the stretch of throat behind your tongue and behind your soft palate. There is no cartilage there. There is no bone. It is a soft, wet, collapsible tube of muscle and mucosa, and the only thing holding it open is muscle tone.

Evolution made that trade deliberately. A rigid pharynx cannot produce speech, cannot swallow a lump of food, cannot switch between nose-breathing and mouth-breathing. Flexibility is the feature. The cost of the feature is that the tube is a Starling resistor: every time your diaphragm pulls down, it generates a negative pressure inside a floppy tube, and the tube wants to implode. Awake, you never notice, because a set of pharyngeal dilator muscles — above all the genioglossus, the big fan-shaped muscle that runs from the inside of your chin into the body of your tongue — fires a fraction of a second before each inspiration and stiffens the walls. The suction arrives to find the door already braced.

The genioglossus is driven by the hypoglossal nerve, cranial nerve XII. This is worth remembering, because it is the single fact behind one of the newer treatments described below. When you fall asleep, output along that nerve drops. In most people it drops a little and nothing happens. In someone whose pharynx is already narrow — a thick tongue base, large tonsils, a small or set-back lower jaw, a crowded palate, fat deposited in the lateral pharyngeal walls, or simply the anatomy they were born with — the same drop in tone is enough for the walls to touch. That is the whole disease.

Why REM is the worst part of the night

Sleep is not one state. In non-REM sleep, muscle tone falls but does not vanish. In REM sleep — the stage where most vivid dreaming happens — the brainstem actively paralyses almost all skeletal muscle. This is REM atonia, and it is a safety mechanism: it stops you from physically acting out your dreams. Two things are spared. The extraocular muscles keep moving, which is where the name rapid eye movement comes from. And the diaphragm keeps working, because it has to.

Notice what is not on the spared list. The genioglossus is skeletal muscle. The other pharyngeal dilators are skeletal muscle. So during REM you get the worst possible combination: a diaphragm still pulling at full strength, generating negative pressure, and an airway whose bracing muscles have been switched off. This is why sleep-study reports so often show events clustering into dense runs, longer and with deeper desaturations, in the REM periods — and why the REM periods are also the ones that get destroyed, because each event ends in an arousal that kicks you out of REM.

REM is also concentrated in the second half of the night, with the longest REM period usually coming in the last couple of hours before waking. That is a large part of why people with untreated apnea say the worst hours are the early morning, and why they so often wake at 4 or 5 a.m. feeling as though they have been fighting something.

Obstructive and central apnea look identical on the airflow channel. They are opposites.

In both, breathing stops. In both, oxygen falls. From the outside — from a bed partner's point of view, from the airflow trace alone — they are indistinguishable. The difference is visible on exactly one channel, and this animation exists partly to make you see it.

Obstructive sleep apnea (OSA) is a plumbing failure. The brain is sending the signal. The diaphragm is contracting. The chest and abdomen are heaving. But the pipe is shut, so no air moves. On a sleep study the airflow channel goes flat while the chest and abdominal effort belts keep swinging — and typically swing harder, in a crescendo, as carbon dioxide accumulates and the respiratory centre shouts louder into a closed door. When the chest and abdomen start moving in opposite directions — chest in, belly out — that paradoxical pattern is a fingerprint of obstruction.

Central sleep apnea is a signalling failure. The airway is wide open. Nothing is blocking anything. But the brainstem respiratory centre has simply stopped issuing the command, usually because a control loop overshot: ventilation blew off too much carbon dioxide, CO₂ fell below the level that triggers the next breath, and the system went quiet until CO₂ climbed back. On the study, both the airflow channel and the effort belts flatline together. Nobody is trying. Central apnea shows up with heart failure (the crescendo–decrescendo pattern called Cheyne–Stokes respiration), with opioids, at high altitude, after a stroke, and sometimes emerging on CPAP itself.

This distinction is not academic, and the animation makes the reason concrete: switch to Central and then turn CPAP on. The airway stays beautifully open — CPAP is doing its job perfectly — and breathing still stops, because CPAP splints a tube; it does not tell a brainstem to breathe. A machine that fixes one disease completely can do essentially nothing for the other.

The AHI: what a sleep study actually counts

The apnea–hypopnea index is the headline number on every sleep-study report. It is a rate, not a total: the number of scored respiratory events per hour of sleep.

An apnea is a drop in airflow of at least 90% lasting at least 10 seconds.
A hypopnea is a partial event — airflow down by at least 30% for at least 10 seconds — that also produces either an oxygen desaturation or a cortical arousal. Which desaturation threshold counts (3% or 4%) depends on which scoring rule the lab uses, and this matters more than it sounds: the same night of sleep can score a meaningfully different AHI under the two rules, which is one reason two reports on the same person can disagree.

The severity bands are fixed and worth memorising, because they are what determines whether treatment is offered: under 5 is normal, 5 to 15 is mild, 15 to 30 is moderate, and above 30 is severe. Note the units. An AHI of 30 means breathing is disrupted, on average, once every two minutes, all night, every night.

A full in-lab polysomnogram records far more than airflow: EEG to stage sleep and detect arousals, EOG for eye movements, chin EMG for muscle tone, ECG, leg movements, nasal pressure and thermistor airflow, chest and abdominal effort belts, oximetry, body position, and often snore sound and video. A home sleep apnea test records a useful subset — usually airflow, effort, oximetry and pulse. Home testing is cheaper, more comfortable and validated for uncomplicated suspected moderate-to-severe OSA. Its weakness is that without EEG it cannot tell how long you actually slept, so it divides events by time in bed rather than time asleep, which systematically underestimates the AHI. A negative home test in someone with convincing symptoms is not the end of the conversation.

Two other numbers on the report deserve attention. The oxygen desaturation index and the lowest recorded SpO₂ describe how much hypoxic stress the night actually delivered, and they sometimes tell a more alarming story than the AHI. The arousal index describes how shredded the sleep architecture is. Someone with an AHI of 12 — technically mild — and an arousal index of 40 is not having a mild night.

Why it damages the heart: the surges, not the snoring

The snoring is the socially disruptive part. It is not the dangerous part. The damage comes from what happens two hundred to six hundred times a night at the end of each event.

During an obstruction, oxygen falls and carbon dioxide climbs while the chest strains against a closed airway. That combination is one of the most powerful sympathetic-nervous-system stimuli the body knows. Catecholamines surge, blood pressure spikes — often far above daytime values — and heart rate does something characteristic: it slows during the apnea (a vagally mediated bradycardia related to the diving reflex, visible on the heart-rate channel in this animation), then jumps on arousal. Cycle that pattern hundreds of times a night, for years, and the vascular system stops resetting. Blood pressure that should dip by 10–20% overnight stops dipping — the "non-dipper" pattern — and eventually daytime pressure rises too. Untreated OSA is the most common identifiable contributor to resistant hypertension: blood pressure that stays high on three or more medications. If someone's pressure will not come down on a good regimen, the sleep study is not an afterthought; it is one of the first things to order.

The same forces act on the atria directly. Each obstructed breath swings intrathoracic pressure violently negative, mechanically stretching the atrial walls; add intermittent hypoxia, autonomic whiplash and systemic inflammation, and you get an unusually good recipe for atrial fibrillation. This is why untreated sleep apnea is associated with substantially higher rates of AF recurrence after cardioversion and after ablation — and why many electrophysiologists now want apnea treated before they will schedule a repeat procedure. Untreated OSA also tracks with coronary disease, stroke, heart failure, pulmonary hypertension and poorly controlled type 2 diabetes.

Morning headache has a simpler explanation: with ventilation repeatedly impaired, carbon dioxide accumulates overnight, and CO₂ is a potent cerebral vasodilator. The classic description is a dull, band-like, bilateral headache present on waking that fades over an hour or two. Daytime sleepiness comes from the arousals rather than the hypoxia. Each event ends in a brief cortical arousal, usually too short to remember. You can spend eight hours in bed, sleep for seven and a half of them by the clock, and never assemble more than a few consecutive minutes of consolidated deep or REM sleep. Sleep quantity looks fine; sleep continuity is gone. That is why "but I slept nine hours" is not reassurance, and it is what the REM readout in this animation is dramatising.

CPAP, and the treatments for people who cannot use it

CPAP — continuous positive airway pressure — is not oxygen and it is not a ventilator. It is a pneumatic splint. A blower pushes room air at a mildly raised pressure, usually somewhere around 5 to 15 cmH₂O, through a mask; that pressure holds the floppy pharynx open from the inside so it cannot implode when the diaphragm pulls. That is the whole mechanism, and it is why it works so completely and so immediately: press CPAP on in this animation and the four traces clean up within a single breath, because the anatomy has been fixed in real time. Used properly it abolishes the events, the desaturations and the arousals, and typically improves blood pressure, sleepiness and quality of life. APAP machines vary the pressure automatically through the night; BiPAP delivers a higher pressure on inhalation than exhalation and is used when pressures get high, when exhaling against CPAP is intolerable, or when there is coexisting hypoventilation.

The honest problem with CPAP is adherence. A machine in a cupboard treats nobody. Most CPAP failure is fixable and is about the interface rather than the therapy: the wrong mask size, a nasal mask on someone who mouth-breathes, dryness that a heated humidifier solves, pressure that feels like too much at the start of the night that a ramp setting solves, or claustrophobia that gets better with a nasal pillow mask instead of a full face mask. Before accepting "I could not tolerate CPAP", it is worth working through the mask.

The real alternatives, for people who genuinely cannot use positive pressure or who have milder disease:

Mandibular advancement devices. A custom dental appliance that holds the lower jaw forward during sleep, which pulls the tongue base forward with it and enlarges the space behind it. Less effective than CPAP on average, but often better tolerated — and a treatment used every night beats a better treatment used twice a week. Best evidence is in mild-to-moderate disease. Get one made by a dentist experienced in sleep medicine; the drugstore boil-and-bite versions are not equivalent and can move teeth.
Positional therapy. In a substantial minority of people, events are dramatically worse lying on the back, because gravity loads the tongue and soft palate straight into the airway. Press Supine position in the animation to see the effect. If the sleep study shows a large supine-versus-lateral difference, a device or a garment that keeps you off your back can do real work.
Weight loss, where there is weight to lose. Fat in the tongue and in the lateral pharyngeal walls physically narrows the tube. The relationship established in large cohort studies is roughly proportional in both directions: a change of about 10% in body weight moves the AHI by roughly a quarter to a third, up or down. That is meaningful but usually not curative on its own, and it is slow — it does not remove the need to treat the apnea now.
Hypoglossal nerve stimulation. An implanted device that senses each breath and delivers a small stimulus to the hypoglossal nerve in time with inspiration, contracting the genioglossus and pulling the tongue forward — restoring, artificially, the reflex that sleep switched off. Approved in the United States since 2014, it is for selected people with moderate-to-severe OSA who cannot use CPAP, and eligibility depends on BMI limits and on a scope examination showing the right collapse pattern.
Surgery. In children, adenotonsillectomy is first-line and often curative. In adults, tonsillectomy helps when tonsils are genuinely large; palatal surgery has modest and unpredictable results; maxillomandibular advancement, which moves both jaws forward, is highly effective but a major operation. Nasal surgery relieves nasal obstruction and can make CPAP tolerable, but rarely cures apnea by itself.
Things to stop. Alcohol in the evening and sedative-hypnotics both suppress pharyngeal dilator tone and blunt the arousal response — longer events, deeper desaturations. Press Alcohol / sedative in the animation and watch the nadir drop. Opioids are worse still, and specifically provoke central apnea. Smoking inflames the upper airway. None of these are moral failings to confess; they are dials that measurably move the disease.

Myth: snoring means you have sleep apnea

It does not. Snoring is the sound of turbulent air vibrating the soft palate and uvula in a partly narrowed airway. Primary snoring — snoring with no significant airflow reduction, no desaturation and no arousals — is common, and on a sleep study it produces an AHI under 5. That is the Snoring scenario in this animation: the uvula is fluttering, the noise is real, but the oxygen trace never moves. Plenty of people are told they must have apnea because they snore, and are then surprised by a normal study. A normal study is a normal study.

The inverse error is more dangerous. Snoring is not sufficient, but it is a genuine signal, and the combination that should always trigger testing is loud habitual snoring + pauses witnessed by someone else + daytime sleepiness. If a partner has watched you stop breathing, that observation carries more weight than any amount of reassurance about how well you sleep. And the absence of snoring does not clear you: people who have had palatal surgery, and many people with central apnea, stop breathing quietly.

Myth: sleep apnea is a disease of large, older men

Male sex, higher BMI, older age and a large neck circumference are all real risk factors, and that is exactly why this myth is so damaging — it is built out of true statements. But risk factors describe a population, not a person, and using them as a filter causes three groups to be missed for years.

Thin people get sleep apnea. A narrow airway can be inherited rather than acquired. A small or set-back lower jaw (retrognathia), a high narrow palate, a long soft palate, large tonsils, or a crowded oropharynx will collapse at a perfectly ordinary body weight. In these people the airway was always marginal; falling asleep is all that was needed to close it.

Women are substantially under-diagnosed, and part of the reason is that the textbook picture was written from male patients. Women more often present with fatigue, insomnia, morning headache, low mood or anxiety rather than the classic loud snoring and observed apneas — and are correspondingly more likely to be worked up for depression, thyroid disease or anaemia first. Women's events are also more likely to be hypopneas than frank apneas, and more likely to cluster in REM, which can produce a modest-looking overall AHI on a night that was genuinely bad. Risk rises sharply after menopause. If you are a woman who has been exhausted for years with a normal thyroid panel and a normal ferritin, sleep-disordered breathing belongs on the list.

Children get it too, usually from large tonsils and adenoids rather than from weight, and they very often do not look sleepy. Paediatric apnea presents as hyperactivity, irritability, poor concentration, bedwetting, mouth breathing and poor school performance — a picture that is regularly mistaken for ADHD. Snoring every night is not normal in a child. It is worth asking about.

When to get tested

Ask for a sleep study if you recognise several of these: loud habitual snoring; pauses in breathing that someone else has witnessed; waking with a gasp or a choking sensation; unrefreshing sleep no matter how long you spend in bed; falling asleep in meetings, in front of the television, or — the red flag that should move you to the front of the queue — while driving; morning headaches; nocturia (getting up repeatedly to urinate, which apnea genuinely causes); dry mouth on waking; and difficult-to-control blood pressure.

Screening questionnaires such as STOP-BANG exist precisely so this conversation is quick. It scores eight items — snoring, tiredness, observed apneas, high blood pressure, BMI, age, neck circumference and sex — and a high score is a straightforward reason to test. It is a screening tool, not a diagnosis; the diagnosis comes from a recording.

Push harder for testing, sooner, if you have resistant hypertension, atrial fibrillation, heart failure, type 2 diabetes, a previous stroke, or if you are being scheduled for surgery under general anaesthesia — undiagnosed OSA meaningfully raises perioperative airway risk, and anaesthetists want to know beforehand. And if you drive professionally, treat sleepiness at the wheel as an emergency rather than an inconvenience.

The reason to push is that this is one of the more satisfying conditions in medicine to treat. The mechanism is mechanical, the diagnosis is objective, and the main treatment works the first night you use it correctly. People who have been exhausted for a decade routinely describe the first well-titrated night as the best sleep of their adult life. It is worth finding out.

↑ Back to the animation