Why Sleep Apnea Treatments Work for Some People and Not Others
Two people with the same AHI can have completely different underlying physiology. Four measurable traits explain much of who responds to which treatment, and none of them appears on a standard report.
Here is a question the standard severity label cannot answer: why does an oral appliance transform one person with an AHI of 25 and do nothing for another?
For most of the history of sleep medicine the answer was "anatomy," and the implicit model was a single disease of a narrow airway with one right treatment. That model is wrong, and the research replacing it is the most useful framework a patient can understand.
The four traits
Obstructive sleep apnea is now understood as arising from a combination of four measurable physiological traits (Finnsson et al., Frontiers in Sleep, 2023):
1. Collapsibility. How easily your upper airway closes when the muscles relax. This is the closest to the old anatomical model — a narrow, floppy airway collapses at higher pressure than a robust one. It is the trait CPAP directly overrides, by holding the airway open with air pressure regardless of how collapsible it is.
2. Upper airway muscle compensation. How well your dilator muscles — chiefly the genioglossus, which protrudes the tongue — respond to a narrowing airway during sleep. Two people with identical collapsibility can differ enormously here. Good compensation partly rescues a collapsible airway; poor compensation means even a modest narrowing goes uncorrected.
3. Arousal threshold. How easily a respiratory disturbance wakes you. This one is counterintuitive: waking easily is often a disadvantage. A low arousal threshold means you wake before your airway muscles have had time to recruit and stabilise breathing, so sleep fragments repeatedly without the underlying obstruction ever being resolved. It also destabilises breathing further on returning to sleep.
4. Loop gain. How aggressively your ventilatory control system responds to a change in carbon dioxide. High loop gain means overshoot — you over-breathe after an event, blow off too much CO2, and the drive to breathe drops, which lets the airway collapse again. It turns a single event into a cycle. This is the trait behind much treatment-emergent central apnea.
Why this matters for treatment
Because different treatments act on different traits, and a treatment that addresses a trait you do not have will not help you.
CPAP overrides collapsibility mechanically, which is why it works across the range and why it is the most reliably effective option. It does not fix loop gain, and in people with high loop gain it can unmask central events.
Oral appliances change airway anatomy by advancing the mandible, so they act on collapsibility — which is consistent with their variable response rate. If your dominant trait is a low arousal threshold or high loop gain, moving your jaw forward addresses the wrong mechanism. CPAP versus an oral appliance.
Positional therapy works when collapsibility is strongly posture-dependent, which is exactly what the supine-versus-lateral split on your sleep study describes. Positional therapy.
Hypoglossal nerve stimulation recruits the dilator muscles directly, so it targets muscle compensation.
The newer drug approaches are explicitly trait-targeted — the aroxybutynin/atomoxetine combination is designed around neuromuscular dysfunction rather than anatomy.
The evidence that traits predict response
The clearest demonstration comes from hypoglossal nerve stimulation.
Roughly one-third of patients treated with hypoglossal nerve stimulation are incomplete responders, despite careful selection based on baseline characteristics and drug-induced sleep endoscopy (Op de Beeck et al., American Journal of Respiratory and Critical Care Medicine, 2021).
That is the problem in one sentence. Patients are selected using the best available anatomical screening — an endoscopy under sedation, looking directly at where the airway collapses — and a third still do not respond adequately.
The researchers computed the four traits from baseline polysomnography data in 91 of 126 patients in the STAR trial and used logistic regression to test AHI-adjusted associations between response — defined as a greater than 50% AHI reduction to under 10 per hour at one year — and the underlying traits [Op de Beeck 2021].
The significance is the approach as much as the result: response was predictable from physiology that a standard report does not show, over and above the anatomy an endoscopy does show.
Why your report does not include them
The honest answer, and it is a practical one rather than a conspiracy.
Most methods for extracting these traits require specialised training and equipment not available in a standard sleep clinic, which has itself limited the ability to assess how much the traits affect outcomes (Finnsson 2023).
That is changing. The work above describes methods for estimating endotypes from ordinary polysomnography signals, including cloud-based implementations — the point being scalability, so the traits could be derived from a study you have already had rather than from a research-grade physiology session.
Tier 2 overall. The framework is well established physiologically and the trait-response associations are real; what does not yet exist is routine clinical use or a trial showing that assigning treatment by endotype beats current practice. How we grade evidence.
What to take from it as a patient
Stop reading treatment failure as personal failure. If an oral appliance did nothing for you, the most likely explanation is that your dominant trait was not the one it addresses. That is a mismatch, not non-compliance and not bad luck.
The severity label is not a treatment plan. Two people with an AHI of 25 may need different things, and nothing in the number distinguishes them.
The proxies you can see are worth reading. Your report will not name these traits, but it hints at them. A high arousal index with a modest AHI suggests a low arousal threshold. A strong supine-versus-lateral difference points at posture-dependent collapsibility. Central events appearing on treatment suggest loop gain. How to read your sleep study.
Ask the question directly. "Do you have a sense of which mechanism dominates in my case?" is a reasonable thing to put to a sleep physician, and the answer shapes what to try next. What to tell the sleep doctor.
The line
None of this makes any consumer product work. Tape, strips and pillows do not address collapsibility, muscle compensation, arousal threshold or loop gain, and none treats obstructive sleep apnea. The framework here is about matching real treatments to real physiology. Snoring versus sleep apnea.