← Back to all articles

Physiology · Phenotyping · Precision sleep medicine

Upper airway instability: why the same AHI can have four different causes

Written by Dr Ari Manuel, Consultant in Sleep & Respiratory Medicine · Last reviewed March 2026

Two patients, each with an AHI of 20. One responds to CPAP immediately — symptoms resolve, adherence is high, objective data confirms effective treatment. The other struggles: poor adherence, residual symptoms, treatment-emergent central apnoeas, intractable insomnia induced by therapy. The AHI was identical. The underlying physiology was not.

For most of sleep medicine's history, OSA was understood primarily as an anatomical problem — a narrow pharynx, a large tongue, excessive soft tissue — and treatment was correspondingly anatomical: open the airway with pressure, a device, or surgery. This model explains some patients well and others poorly. The last two decades of physiological research have produced a more nuanced framework, now widely referred to by the acronym ALZHAR (or the related Eckert/Wellman phenotyping model), that identifies four distinct physiological traits which, in various combinations, produce sleep-disordered breathing — and which have fundamentally different treatment implications.

The four traits

A — Anatomy (Passive Critical Closing Pressure, Pcrit)

Pcrit is the luminal pressure at which the passive, unprotected upper airway collapses. It is a measure of pure anatomical vulnerability — the structural tendency of the airway to obstruct when muscle tone is removed, as it is during sleep. Patients with a highly negative Pcrit (e.g. −10 cmH₂O) have structurally robust airways that are unlikely to collapse; patients with a positive or mildly negative Pcrit (e.g. −2 to +2 cmH₂O) have anatomically compromised airways that collapse readily. Obesity increases Pcrit by loading the airway with peripharyngeal fat; retrognathia, macroglossia, and adenotonsillar hypertrophy do the same through structural crowding.

Pcrit is the dominant trait in most patients with moderate-to-severe OSA — which is why CPAP, by mechanically splinting the airway open with positive pressure that counteracts the collapsing tendency, works so well in this group. If anatomy is the whole story, CPAP is close to a perfect treatment.

L — Loop Gain

Loop gain describes the stability of the ventilatory control system — specifically, the ratio of the ventilatory response to a disturbance compared with the disturbance itself. A system with high loop gain overreacts: a small reduction in ventilation triggers a disproportionately large ventilatory response, which overshoots, drives CO₂ below the apnoeic threshold, and triggers another central or obstructive event. The result is a cyclical, self-perpetuating breathing instability.

High loop gain is the physiological mechanism underlying periodic breathing, Cheyne-Stokes respiration, and many cases of central sleep apnoea. But it also contributes significantly to obstructive OSA in patients where the anatomical compromise is mild — it is the instability of the control system, not just the narrowness of the airway, that drives respiratory events. Patients with high loop gain tend to have positional or REM-predominant OSA, respond poorly to CPAP alone (which can worsen loop gain instability at high pressures), and may do better with acetazolamide (which reduces loop gain by causing a mild metabolic acidosis that raises CO₂ reserve), supplemental oxygen, or adaptive servo-ventilation.

Z — Arousal Threshold

The arousal threshold describes how easily the brain wakes in response to a respiratory event. A low arousal threshold means the patient wakes — and therefore terminates the apnoea — at relatively modest levels of respiratory loading. This sounds protective, but it is paradoxically harmful: the arousal prevents the oxygen desaturation from becoming severe, but the repeated arousals fragment sleep architecture profoundly, and the wake state triggered by each arousal causes hyperventilation that blows off CO₂, re-establishes pharyngeal muscle tone, causes upper airway reopening, and then — as sleep resumes — the cycle begins again.

Patients with low arousal threshold are characterised by frequent but short apnoeas, relatively modest oxygen desaturations, high arousal indices on polysomnography, and often prominent insomnia as a comorbidity. They may have only mild-to-moderate AHI but severely disrupted sleep architecture. Clinically, they often complain more of insomnia and non-restorative sleep than of sleepiness — which means the diagnosis is frequently delayed. The pharmacological approach of raising the arousal threshold with agents such as trazodone or eszopiclone has evidence in selected patients and is a consideration in those who fail CPAP or have contraindications.

H — Upper Airway Muscle Responsiveness

The upper airway is not a passive tube. The genioglossus, tensor palatini, and other dilator muscles actively stiffen and open the pharynx in response to the negative pressure generated during inspiration. This neuromuscular compensation is the primary defence against airway collapse, and its adequacy determines whether a patient with a given Pcrit actually obstructs. In patients with robust neuromuscular responsiveness, the muscles compensate for a moderately compromised anatomy and events do not occur. In patients where this responsiveness is diminished — whether through reduced reflex sensitivity, muscle fatigue, or REM-related atonia — the same anatomical vulnerability becomes clinically manifest.

Upper airway muscle responsiveness is the therapeutic target of hypoglossal nerve stimulation (HNS), which directly activates the genioglossus via the hypoglossal nerve, bypassing the reflex pathway entirely. HNS is now NICE-approved for selected patients with CPAP-intolerant moderate-to-severe OSA. It is also the target of myofunctional therapy — structured oropharyngeal exercise programmes that, in RCTs, reduce AHI by approximately 50% in adults with predominantly muscle-responsive airway compromise.

Why this matters clinically

The ALZHAR framework has direct treatment implications that go well beyond academic interest. It explains several phenomena that the anatomical model cannot:

Why some patients fail CPAP despite apparently adequate pressure. If high loop gain is the dominant mechanism, elevated CPAP pressure can worsen ventilatory instability — increasing the propensity for central events, hypocapnia, and treatment-emergent complex sleep apnoea. These patients may need lower pressures, bilevel therapy, or ASV rather than escalating CPAP.

Why some patients with low AHI have severe symptoms. A patient with a low arousal threshold may have an AHI of 10 but 40 arousals per hour and severely fragmented sleep. Their AHI understates the clinical significance of their disease. Conventional threshold-based treatment decisions will miss them entirely.

Why mandibular advancement devices work better for some patients than others. MADs work primarily by advancing the mandible to increase pharyngeal dimensions — an anatomical mechanism. They are most effective in patients where anatomy (Pcrit) is the dominant trait, where loop gain is low, and where upper airway muscle responsiveness is preserved. Phenotyping can predict MAD response with reasonable accuracy.

Why weight loss can produce disproportionate AHI reductions in some patients. In patients with high loop gain, weight loss — by reducing the metabolic load and improving respiratory mechanics — may reduce loop gain substantially, producing AHI reductions beyond what anatomical improvement alone would predict. GLP-1 receptor agonists may have additional effects on upper airway muscle tone beyond their weight-reducing effects, a hypothesis under active investigation.

Why positional therapy works in a specific subgroup. Position-dependent OSA — disease that is predominantly or exclusively supine — reflects a phenotype where Pcrit is only marginally above the collapse threshold in the supine position and sufficiently below it when lateral. These patients may achieve near-complete resolution of OSA with positional interventions alone, without CPAP or any device.

Phenotyping in clinical practice

Formal measurement of Pcrit, loop gain, and arousal threshold requires research-grade equipment not available in routine practice. However, several clinical and polysomnographic surrogates allow phenotypic inference from standard data. The pattern of events on a sleep study — their duration, oxygen desaturation depth, arousal association, and positional distribution — carries phenotypic information that a skilled interpreter can use. Tools such as the PALM score (Pcrit, Arousal threshold, Loop gain, Muscle responsiveness) operationalise this approach. Ambulatory monitoring platforms that generate cardiopulmonary coupling data — including SleepImage — provide additional phenotypic signal through autonomic and ventilatory stability indices.

At this service, standard home sleep study data is interpreted in this physiological context — not just as an AHI number, but as a pattern of breathing behaviour that reflects the underlying mechanisms. Treatment recommendations follow the phenotype, not just the severity category.