Scoring apnea and hypopnea events to classify OSA severity from the Apnea-Hypopnea Index
The human upper airway is a uniquely unsupported tube: unlike the trachea, it has no rigid cartilaginous ring holding it open. From the nasal choanae to the larynx, patency depends entirely on the active, moment-to-moment contraction of roughly two dozen paired skeletal muscles. While awake, this neuromuscular scaffolding comfortably wins the tug-of-war against negative inspiratory pressure and gravity — but that margin is far smaller in people who go on to develop obstructive sleep apnea (OSA).
The pharynx is conventionally divided into three collapsible segments relevant to sleep-disordered breathing: the nasopharynx, the retropalatal oropharynx (behind the soft palate and uvula — the most common site of collapse), and the retroglossal oropharynx (behind the tongue base). Each segment is a compliant tube whose caliber is set by a dynamic balance between two opposing forces: the negative intraluminal pressure generated by the diaphragm during inspiration, which tends to suck the walls inward, and the outward mechanical support provided by surrounding soft tissue tone, fat pad volume, and skeletal (mandibular, hyoid, cervical spine) attachments.
In an anatomically favorable airway, the resting cross-sectional area is generous and the surrounding soft-tissue mass is modest, so inspiratory negative pressure never approaches the closing threshold. In a crowded airway — a setback mandible, a low-lying hyoid, tonsillar hypertrophy, or fat infiltration of the parapharyngeal and tongue tissue — the resting lumen is already narrowed, and comparatively little additional collapsing pressure or loss of tone is needed to occlude it.
Genioglossus, the largest tongue muscle, is the most studied upper-airway dilator. Its surface EMG shows two components while awake: a tonic (baseline, continuous) discharge that keeps the tongue base forward, and a phasic component that bursts with each inspiration, timed to counteract the negative pressure pulse generated by diaphragmatic contraction just before it arrives at the pharynx.
This anticipatory, pressure-triggered reflex is mediated by pharyngeal mechanoreceptors that sense wall deformation and negative pressure, signaling through a brainstem circuit back to the hypoglossal motor nucleus. Other dilators — tensor and levator veli palatini (stiffen and elevate the soft palate), geniohyoid and sternohyoid (anchor and lower the hyoid bone, indirectly stabilizing the tongue base) — act in a coordinated ensemble so that the entire pharyngeal tube stiffens synchronously with each breath.
This awake neuromuscular compensation is so effective that people with markedly narrow anatomic airways can be asymptomatic while awake — their genioglossus and palatal muscles work harder, but the airway holds. Sleep removes this compensation, which is precisely why OSA is, almost by definition, a disease that only manifests during sleep.
Several structural features chronically narrow the baseline airway and lower the pressure reserve before sleep even begins: retrognathia and micrognathia (posteriorly set mandible reduces the retroglossal space), macroglossia and lingual tonsil hypertrophy, adenotonsillar hypertrophy (the dominant cause in children), a low hyoid position, and — most prevalent in adults — parapharyngeal and tongue-base fat deposition associated with obesity and a large neck circumference (>43 cm in men, >37 cm in women is an independent OSA risk marker independent of BMI).
Craniofacial factors interact multiplicatively with soft-tissue bulk: a modestly narrow jaw combined with modest weight gain can produce the same functional airway crowding as severe obesity alone. This is why the two sliders driving this simulation — airway collapsibility (Pcrit, largely anatomic/neuromuscular) and BMI category (largely soft-tissue load) — combine to determine event frequency and severity rather than acting independently.
The transition from wake to NREM sleep is the pivotal physiological event in OSA pathogenesis. Within seconds of sleep onset, the brainstem withdraws wakefulness-dependent excitatory drive to upper-airway motor neurons. Muscles that were compensating for an anatomically compromised airway suddenly stop compensating — and the airway behaves according to its passive mechanical properties alone.
Wakefulness itself provides a tonic excitatory input to upper-airway motor neurons, independent of the reflex pathways described in Stage 1 — this is sometimes called the "wakefulness stimulus." As cortical arousal state transitions into stage N1/N2 NREM sleep, ascending reticular activating system output falls, and this wakefulness stimulus is withdrawn. Hypoglossal motor neuron firing rate to genioglossus drops within one to two breaths, well before any change in blood gases could explain it.
The mechanoreceptor-mediated negative-pressure reflex described previously also blunts during sleep — its gain (the increase in EMG activity per unit of negative pressure) falls by roughly half in NREM sleep compared with wakefulness, and falls further still in REM sleep, when skeletal muscle atonia is most profound. This is why OSA events are typically most frequent and most severe during REM sleep.
The pharynx behaves mechanically like a Starling resistor: a collapsible segment situated between two rigid tubes (the nasal airway upstream, the trachea downstream) whose flow is governed not by the pressure gradient between the two ends but by the relationship between the surrounding tissue pressure and the pressure inside the collapsible segment. The airway collapses when intraluminal pressure falls below the critical closing pressure, Pcrit — a single number that summarizes the net mechanical "collapsibility" of an individual's airway, combining anatomic loading and passive tissue compliance.
Pcrit is measured by transiently dropping applied nasal pressure (via a modified CPAP circuit) and recording the pressure at which flow ceases. Values below about −5 cmH₂O predict a stable, non-collapsing airway; primary snorers without apnea cluster around −5 to 0 cmH₂O; patients with moderate-to-severe OSA typically show Pcrit between 0 and +5 cmH₂O — meaning the airway collapses even at pressures at or above atmospheric, and CPAP must supply a positive pneumatic splint simply to keep it at zero net collapsing pressure.
Pcrit is the single best physiological predictor of OSA severity because it integrates every anatomic and neuromuscular factor into one number. In this simulator, the "Airway Collapsibility" slider is a simplified stand-in for a patient's measured Pcrit — Low corresponds to a stable, negative Pcrit airway; High corresponds to a markedly positive, easily collapsible one.
Before the airway occludes completely, it typically passes through an intermediate state called inspiratory flow limitation: as inspiratory effort increases, airflow through the narrowed segment plateaus rather than continuing to rise, producing a characteristic flattened or "sawtooth" top on the nasal pressure or pneumotachograph tracing instead of the normal rounded sinusoidal inspiratory curve. This occurs because, once flow becomes high enough at a Starling-resistor site, further increases in upstream (alveolar) driving pressure are no longer transmitted downstream — the collapsible segment "chokes" the flow.
Flow limitation alone, without a scoreable desaturation or arousal, is not counted toward the AHI, but sustained flow-limited breathing associated with arousals is captured by the broader Respiratory Disturbance Index (RDI), which adds Respiratory Effort-Related Arousals (RERAs) to the apnea and hypopnea count.
When the collapsing pressure at the retropalatal or retroglossal segment exceeds the dilator muscles' waning ability to resist it, the airway occludes completely. Under the American Academy of Sleep Medicine (AASM) scoring manual, this moment begins the countdown toward a scoreable respiratory event, and — a few heartbeats later, once already-oxygenated blood in transit finishes circulating — toward a measurable fall in peripheral oxygen saturation.
An apnea is scored when airflow (measured at the mouth/nose by a thermal sensor, or more precisely by nasal pressure transducer) drops by ≥90% from baseline for at least 10 seconds. A hypopnea, per the AASM 2012 recommended (now most widely used) rule, requires a ≥30% drop in the nasal pressure signal for ≥10 seconds, accompanied by either a ≥3% oxygen desaturation or an EEG-defined arousal.
Apneas are further sub-classified using respiratory effort belts (inductance plethysmography around chest and abdomen): obstructive apneas show absent airflow with continued, often paradoxical (out-of-phase chest/abdomen) respiratory effort, because the diaphragm keeps trying to breathe against a sealed airway; central apneas show absent airflow with absent effort — the brainstem simply stops sending the signal to breathe; mixed apneas begin as central (no effort) and transition to an obstructive pattern (effort resumes but the airway stays sealed) before ending.
From the moment of occlusion, the diaphragm and intercostal muscles typically continue rhythmic contractions against the closed airway, generating progressively larger negative intrathoracic pressure swings (visible as increasing effort-belt excursions and, if esophageal manometry is used, deepening negative deflections) without moving any air. Arterial oxygen content already in the pulmonary capillaries and left heart continues to circulate normally, so the pulse-oximeter SpO2 reading does not begin falling immediately — there is a characteristic 6–12 second lag reflecting lung-to-fingertip circulation time.
Once desaturation begins, its rate and depth depend on event duration, baseline lung oxygen reserve (functional residual capacity), and — critically — where on the oxyhemoglobin dissociation curve the patient starts. Because the curve is sigmoidal, identical drops in arterial PaO2 produce far larger SpO2 drops once saturation falls below about 90%, which is why events in patients who start with borderline hypoxemia, or with a very collapsible airway prolonging the event, can plunge to profoundly low nadirs.
The oxyhemoglobin dissociation curve is nearly flat between 100% and 90% SpO2 but extremely steep between 90% and 70% — a difference of only a few mmHg of PaO2 in that range corresponds to a huge swing in SpO2. This is why "just a few percent lower" nadir SpO2 readings in severe OSA (down into the 60s–70s) represent a dramatically larger fall in dissolved oxygen delivery than the percentages alone suggest.
Not all scored events carry equal physiological weight. Four variables jointly determine the harm done by any single apnea or hypopnea: duration (longer events allow deeper desaturation), desaturation nadir (the lowest SpO2 reached), the frequency of events per hour (which determines how much of the night is spent in a desaturated state), and the intervening reoxygenation dynamics between events. Repetitive cycles of desaturation followed by rapid reoxygenation constitute a pattern of intermittent hypoxia analogous to ischemia-reperfusion injury, generating reactive oxygen species with each cycle and driving the systemic oxidative stress and endothelial dysfunction implicated in OSA-associated cardiovascular disease.
Because standard AHI only counts events without weighting for depth or duration, researchers increasingly track supplementary indices — the Oxygen Desaturation Index (ODI, desaturations of a defined magnitude per hour) and the newer "hypoxic burden" (the total area under the desaturation curve per hour) — which correlate more tightly with cardiovascular outcomes than AHI alone (see Stage 5).
Left completely uncorrected, progressive hypoxemia and hypercapnia would be lethal — but the brain does not let that happen. Rising respiratory drive, chemoreceptor stimulation, and mechanical loading eventually trigger a cortical arousal: a brief, often subconscious awakening that restores upper-airway muscle tone just long enough to reopen the airway, followed by a burst of compensatory hyperventilation and a transient surge of sympathetic nervous system activity.
An AASM-scoreable arousal requires an abrupt shift in EEG frequency — to alpha, theta, or frequencies greater than 16 Hz (excluding spindles) — lasting at least 3 seconds, occurring after at least 10 seconds of stable prior sleep, with (in REM sleep) a concurrent increase in submental EMG tone. Arousals can occur with or without the sleeper reaching conscious awareness; most apnea-terminating arousals are never remembered the next morning, yet they still fragment sleep architecture and prevent progression into restorative slow-wave and REM sleep.
The "arousal threshold" — how much respiratory stimulus (hypoxia, hypercapnia, mechanical loading) is required to trigger arousal — varies substantially between individuals and is itself a target of emerging pharmacotherapy: patients with a low arousal threshold wake too easily, terminating events before enough time has passed to build stable ventilatory drive, which can paradoxically perpetuate a cycle of unstable breathing.
At arousal, hypoglossal and other upper-airway motor neuron output abruptly returns to near-wakeful levels, restoring genioglossus and palatal muscle tone within one to two breaths. The airway snaps back open, often audible as a loud snort or gasp as high-velocity air rushes through the still-narrowed passage. Because the preceding obstructed effort has built up substantial negative intrathoracic pressure and a chemoreceptor-driven ventilatory drive elevated by accumulated CO2 and falling O2, the first several breaths after reopening are markedly larger and faster than baseline — a compensatory hyperpnea that rapidly restores minute ventilation, clears retained CO2, and drives SpO2 back toward baseline (sometimes with a brief overshoot above the pre-event baseline).
Each arousal is accompanied by a burst of sympathetic nervous system activity — measurable directly via muscle sympathetic nerve activity (MSNA) recordings — that produces transient tachycardia and a beat-to-beat blood pressure surge, often 20–30 mmHg systolic or more, superimposed on the bradycardia that frequently accompanies the preceding apnea itself. This alternating brady-tachycardia pattern, cyclical variation in heart rate (CVHR), is distinctive enough to be used as an automated OSA-screening signal from ECG or pulse-oximetry waveforms alone.
Repeated hundreds of times nightly, this chronic intermittent sympathetic activation is a principal mechanistic link between OSA and its downstream cardiovascular sequelae: sustained (including nocturnal non-dipping and resistant) hypertension, atrial fibrillation, elevated stroke risk, and heart failure exacerbation. Sleep fragmentation from the arousals themselves — independent of the hypoxia — also drives excessive daytime sleepiness, impaired glucose tolerance, and next-day cognitive and mood impairment.
Because most arousals are not consciously remembered, patients with severe OSA — sometimes experiencing an arousal every one to two minutes all night — frequently have no idea their sleep is fragmented at all. They simply feel unrefreshed and excessively sleepy the next day, which is why the diagnosis so often requires objective polysomnography rather than patient-reported sleep quality.
Every scored apnea and hypopnea across a full night of monitoring is summed and normalized by sleep duration to produce the single number that anchors the entire clinical framework of sleep apnea: the Apnea-Hypopnea Index (AHI). Despite its well-recognized limitations, AHI remains the primary metric used to diagnose OSA, grade its severity, guide treatment decisions, and determine insurance coverage for therapy worldwide.
AHI = (number of obstructive + central + mixed apneas, plus hypopneas) ÷ total sleep time in hours, as scored from either full in-laboratory polysomnography (PSG) or a home sleep apnea test (HSAT). PSG uses EEG-confirmed total sleep time as the denominator, the gold-standard approach; HSAT devices lack EEG and instead estimate total recording time or use surrogate sleep-detection algorithms, which tends to make the HSAT-derived index (sometimes labeled REI, Respiratory Event Index) systematically lower than a PSG-derived AHI for the same patient, because time awake in bed is harder to exclude from the denominator.
A related, broader metric — the Respiratory Disturbance Index (RDI) — additionally counts Respiratory Effort-Related Arousals (RERAs): flow-limited breathing that triggers an arousal without meeting the amplitude criteria for a hypopnea. RDI is always ≥ AHI and better captures patients with primarily arousal-driven (rather than desaturation-driven) disease, sometimes labeled Upper Airway Resistance Syndrome when AHI is normal but RDI is elevated.
The AASM severity scale, near-universally adopted, is: Normal, AHI <5 events/hour; Mild, 5–15; Moderate, 15–30; Severe, >30. These thresholds were set somewhat arbitrarily from early epidemiologic and symptom-correlation studies and, while enormously useful for standardizing care, correlate only imperfectly with symptom burden and cardiovascular risk at the individual level — some patients with an AHI of 12 are profoundly symptomatic and hypertensive, while others with an AHI of 28 are comparatively asymptomatic, depending on event duration, desaturation depth, arousal intensity, and individual physiological resilience.
Severity classification interacts with the two sliders driving this simulation: a highly collapsible airway (high Pcrit) shortens the time to occlusion and deepens each event, while a higher BMI category adds parapharyngeal soft-tissue load that both increases event frequency and worsens desaturation depth per event — together they drive the projected AHI from the normal range up through severe.
Severity grading directly shapes the treatment algorithm. Mild OSA with minimal symptoms may be managed with weight loss, positional therapy (avoiding supine sleep, since gravity worsens retroglossal collapse), and mandibular advancement oral appliances. Moderate-to-severe OSA — especially with cardiovascular comorbidity, excessive daytime sleepiness, or occupational safety concerns (commercial drivers, pilots) — is generally treated first-line with continuous positive airway pressure (CPAP), which pneumatically splints the airway open at a pressure exceeding Pcrit. Alternatives and adjuncts for CPAP-intolerant patients include hypoglossal nerve stimulation (an implanted device that senses inspiration and electrically stimulates the genioglossus to protrude the tongue), maxillomandibular advancement surgery, and, in children, adenotonsillectomy as first-line therapy.
Because AHI simply counts events without weighting for duration, desaturation depth, or arousal intensity, two patients with identical AHI can have very different disease burdens. This has driven development of newer composite indices, most notably the Sleep Apnea-Specific Hypoxic Burden (SASHB — the total area under the desaturation curve associated with respiratory events, per hour), which several large cohort re-analyses have shown predicts cardiovascular mortality more accurately than AHI alone. Nonetheless, AHI remains entrenched as the regulatory and clinical standard because of its simplicity, decades of accumulated outcome data, and ease of automated scoring — newer metrics are gradually being incorporated as adjuncts rather than replacements.
A single AHI number can never fully capture a night of sleep-disordered breathing — but as a reproducible, standardized rate, it remains the backbone of OSA diagnosis and severity communication between sleep physicians, primary care providers, and payers worldwide.
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| Normal | |||
| Mild | |||
| Moderate | |||
| Severe |