Sleep Apnea Market Sizing Is Measuring the Wrong Population

Alternative sleep apnea therapies are not taking share from CPAP. They are being sold, for the most part, into the far larger population that CPAP never held — and most published sleep apnea market sizing cannot tell those two motions apart. That distinction decides everything downstream. Substitution has a hard ceiling set by the installed base of positive-airway-pressure users. Expansion has a ceiling set by diagnostic throughput and reimbursement policy, both of which moved more in the last twenty-four months than in the prior decade.

The rest of this piece is an attempt to earn that opening.

The denominator nobody has actually measured

The number underneath nearly every forecast in this category comes from Benjafield et al. (2019), The Lancet Respiratory Medicine: an estimated 936 million adults aged 30 to 69 worldwide with obstructive sleep apnea at an apnea-hypopnea index (AHI) of 5 or greater, and roughly 425 million at moderate-to-severe thresholds (AHI ≥ 15). It is the most careful global estimate available. It is also a model — extrapolated from a limited set of country-level prevalence studies using heterogeneous scoring criteria, which the authors state plainly. The confidence intervals in the original paper are wide enough to swallow entire regional markets.

On top of that sits a second number, usually stated with less hedging: that most cases are undiagnosed. The common citation is the 2016 economic burden analysis published by the American Academy of Sleep Medicine with Frost & Sullivan, which put roughly 23.5 million of an estimated 29.4 million affected U.S. adults outside the diagnostic system. Treat it as directionally sound and precisely unknowable.

Against a prevalence denominator that loose, the device market is comparatively legible. ResMed's annual revenue now runs near $5 billion; Philips Respironics has been operating under the shadow of its 2021 recall of roughly 5.5 million devices ever since. Published estimates of the global sleep apnea device market cluster somewhere between $6 billion and $9 billion, and the spread between them is mostly definitional — whether masks and consumables count, whether diagnostics count, whether home testing counts. The variance is a taxonomy problem wearing the costume of an empirical one. A forecast quoted to one decimal place is telling you about the vendor's formatting conventions, not their measurement precision.

Adherence is a market-structure problem, not a compliance problem

The industry has spent twenty years describing CPAP nonadherence as a patient behavior issue. It is more usefully read as a demand signal.

The canonical review, Weaver and Grunstein (2008) in Proceedings of the American Thoracic Society, put nonadherence between 46% and 83% depending on the definition, where adherence itself was defined generously: four or more hours per night on 70% of nights. That threshold is arbitrary and low. In the SAVE trial — McEvoy et al. (2016), New England Journal of Medicine, 2,717 patients with moderate-to-severe OSA and established cardiovascular disease — mean CPAP use was 3.3 hours per night, and the trial found no reduction in cardiovascular events. The most common reading is that undertreatment, not therapeutic failure, explains the null result. That reading is plausible and not proven.

The commercially relevant metric is one the sleep literature already has and market models almost never use: mean disease alleviation, the product of therapeutic efficacy and objective hours of use. CPAP has near-total efficacy and mediocre adherence. Oral appliances have partial efficacy and better adherence. On the composite, they converge far more than a per-night AHI comparison suggests. Any forecast that segments by device category rather than by delivered therapeutic hours is measuring shipments and calling it treatment.

Where the airway actually fails, in the order it happens

Sleep onset reduces tonic drive to the pharyngeal dilator muscles — genioglossus chief among them. The airway, now less stiffened, narrows or closes at whatever pressure its anatomy dictates; this collapsibility is measurable as critical closing pressure. Airflow falls. Oxygen desaturates and CO₂ climbs. Chemoreceptors detect the drift and increase ventilatory drive. If that control loop overshoots — high loop gain — the correction itself destabilizes breathing. Meanwhile the accumulating respiratory effort approaches the threshold at which the brain arouses. A low arousal threshold ends the event early, before dilator muscles have had time to recruit, and the cycle restarts.

Eckert et al. (2013), American Journal of Respiratory and Critical Care Medicine, phenotyped 75 subjects across these four traits and found that impaired anatomy alone accounted for the disease in only a minority; most patients carried two or more contributing endotypes.

This is the structural reason the alternatives segment does not behave like a single market. Positional therapy addresses one anatomic subgroup. Nerve stimulation addresses dilator responsiveness. A sedative-hypnotic that raises arousal threshold addresses a different patient entirely. These are not competing products. They are partially disjoint markets sharing a diagnostic code.

What are the alternatives to CPAP?

In clinical practice, five: mandibular advancement oral appliances, positional therapy, hypoglossal nerve stimulation, upper airway surgery (including maxillomandibular advancement), and — new as of 2024 — pharmacotherapy. Everything else is a variation on one of those five.

Commercially they behave nothing alike. Oral appliances are dentist-delivered, low-capital, and reimbursed inconsistently. Hypoglossal nerve stimulation is a surgical implant with dedicated Category I CPT coding since 2020 and a hospital-procedure economic profile. The pivotal STAR trial — Strollo et al. (2014), NEJM, 126 patients — reported median AHI falling from 29.3 to 9.0 at twelve months. Excellent numbers, and the enrollment criteria are where the market sizing lives: BMI at or below 32, AHI between 20 and 50, and no complete concentric palatal collapse on drug-induced sleep endoscopy. The eligible fraction of the diagnosed population is a small single-digit-to-low-double-digit percentage, and no forecast should be read without an explicit assumption stated for it.

The trials that move the denominator

SURMOUNT-OSA — Malhotra et al. (2024), NEJM, 469 participants with obesity and moderate-to-severe OSA across two trials — reported AHI reductions on tirzepatide of roughly 25 events per hour against roughly 5 on placebo. The FDA approved the drug for OSA in December 2024. Separately, Apnimed's AD109 program, building on the atomoxetine-plus-antimuscarinic mechanism tested in MARIPOSA (Schweitzer et al., 2023, AJRCCM), targets dilator muscle activation pharmacologically rather than mechanically.

A device forecast built before December 2024 has a structural error in it, and the direction of that error is genuinely ambiguous. Weight-loss pharmacotherapy could shrink the severe-disease pool that drives PAP revenue. It could also pull millions of undiagnosed patients into sleep medicine for the first time, most of whom will still have residual apnea after treatment. The honest position is that nobody has three years of persistence data yet, and the models asserting either outcome are asserting.

Diagnosis stops being the bottleneck

Home sleep apnea testing already moved diagnosis out of the lab. In 2024 the FDA authorized sleep apnea detection features on both Samsung Galaxy and Apple Watch devices. These are not diagnostic instruments and their manufacturers do not claim they are. They are funnel machinery, and they sit on hundreds of millions of wrists.

If even a modest fraction of the undiagnosed pool is converted to referrals, the binding constraint on the alternatives market stops being clinical evidence and becomes sleep physician capacity and payer policy. That is a very different bottleneck to model, and it is not one that device-shipment data can see.

An honest rule of thumb

Before accepting any forecast in this category, find the eligible-fraction assumption. If it isn't stated, the number is a prevalence figure wearing a market's clothes.

What this doesn't answer

It doesn't answer the question that actually determines the segment's value: how long patients stay on any of these therapies. Adherence at ninety days is a regulatory artifact; persistence at three years is the economics, and almost nobody publishes it. It also doesn't answer how payers will price a device once a pill with a comparable AHI effect exists. Look for real-world persistence registries, CMS coverage determinations following the 2024 drug approval, and — most informative of all — the eligible-fraction assumptions buried in the methodology appendices that vendors would rather sell you than show you.

The market that gets sized is the one that can be counted, and in sleep apnea those have never been the same market.