Obstructive sleep apnea is the most under-diagnosed serious sleep disorder in the adult population. Estimates from the Wisconsin Sleep Cohort and the HypnoLaus study converge on roughly 25 to 30 percent of middle-aged adults meeting criteria for at least mild OSA, with a meaningful percentage having moderate-to-severe disease and no diagnosis at all. Continuous positive airway pressure — CPAP — is the first-line treatment, and its evidence base for daytime sleepiness, cognitive function, and quality of life is stronger than the evidence for essentially any other sleep intervention on any market anywhere.
That is the headline. The complication is the compliance rate. Somewhere between 30 and 50 percent of patients prescribed CPAP abandon it within the first year. That is not a rounding error. That is the majority of a good therapy failing at delivery, and any honest article about CPAP has to hold both facts at the same time.
What OSA actually is
Obstructive sleep apnea is repeated collapse of the upper airway during sleep, causing partial or complete cessation of airflow, oxygen desaturations, arousals from sleep, and — over years — measurable increases in cardiovascular disease, metabolic dysfunction, cognitive decline, and mortality.
The apnea-hypopnea index (AHI) counts obstructive events per hour of sleep. Standard cutoffs: 5 to 14 events per hour is mild, 15 to 29 is moderate, 30 or more is severe. AHI is measured by polysomnography — an overnight lab study — or, increasingly, by home sleep apnea testing.
The classic phenotype is a middle-aged man with a large neck circumference who snores, feels unrefreshed on waking, and reports daytime sleepiness. That phenotype is real, but incomplete. Women present differently and are systematically under-diagnosed, often with more prominent insomnia and fatigue and less prominent snoring. Post-menopausal women approach male prevalence rates. Non-obese OSA is common — craniofacial anatomy alone can cause it. "You do not look like a sleep apnea patient" is a bad reason to skip screening.
Screening: STOP-BANG
The STOP-BANG questionnaire is the workhorse screening tool. Eight yes/no questions.
- Snoring loudly
- Tired during the day
- Observed apneas by a partner
- Pressure — high blood pressure
- BMI over 35
- Age over 50
- Neck circumference over 40 cm
- Gender male
Three or more yeses is a positive screen for possible OSA. Five or more is a high-probability screen. It is not a diagnosis — a positive STOP-BANG means you should be tested. Chung and colleagues validated STOP-BANG in surgical populations and it now has broad primary-care use.
The CPAP evidence base, honestly
This is where the picture gets interesting, because CPAP absolutely works for some outcomes and is more mixed for others.
Daytime sleepiness — large, clear benefit
Meta-analyses of randomized trials of CPAP versus sham in patients with moderate-to-severe OSA show robust improvements in daytime sleepiness. Effect sizes on the Epworth Sleepiness Scale are large and consistent across trials — roughly 2 to 3 points of reduction, which is clinically meaningful. Patti and colleagues, Patel and colleagues, and multiple Cochrane reviews all converge on this. Daytime sleepiness is the single outcome where CPAP has the strongest and most reliable effect.
Cognitive function and mood — meaningful improvement
Executive function, attention, and processing speed show measurable improvement on CPAP in patients who were previously symptomatic. Mood improves. Quality of life improves. These are consistent findings across multiple randomized trials, particularly in moderate-to-severe disease with baseline sleepiness.
Blood pressure — modest, real
CPAP produces small but real reductions in 24-hour blood pressure — roughly 2 to 3 mmHg on average, larger in resistant hypertension and in patients with severe untreated disease. Not transformational. Real.
Cardiovascular outcomes — this is where it gets complicated
The SAVE trial, published by McEvoy and colleagues in the New England Journal of Medicine in 2016, is the largest randomized trial of CPAP for cardiovascular outcomes to date. 2,717 patients with moderate-to-severe OSA and established cardiovascular disease were randomized to CPAP plus usual care versus usual care alone. Over a mean follow-up of about 3.7 years, the primary composite outcome — cardiovascular death, myocardial infarction, stroke, hospitalization for heart failure — showed no significant benefit from CPAP.
The finding surprised the field. Several explanations get discussed, and none fully resolves the result.
- Adherence was modest. Mean CPAP use was 3.3 hours per night — well below the 4-hour threshold commonly used to define adequate use, and below the 6+ hours that observational data suggests is needed for full cardiovascular benefit.
- Very sleepy patients were excluded. Those most likely to comply and benefit were not in the trial for ethical reasons — you cannot randomize a severely sleepy driver to no treatment.
- 3.7 years may be too short for a meaningful cardiovascular signal in secondary prevention.
Subsequent analyses of SAVE and observational cohorts suggest that patients who use CPAP more than 4 hours per night, particularly more than 6, may see cardiovascular benefit. But the randomized-trial evidence for hard cardiovascular endpoints is not the slam-dunk many sleep medicine textbooks claimed pre-SAVE. The honest read: CPAP is an excellent treatment for the symptoms of OSA and probably reduces long-term cardiovascular risk in adherent users, but the strongest available RCT evidence for hard cardiovascular outcomes is modest.
None of that changes the case for CPAP in a symptomatic patient. The daytime sleepiness benefit alone is worth the therapy.
The compliance problem
Kribbs, Weaver, and Grunstein established across multiple studies that a substantial fraction of prescribed CPAP users do not use the therapy adequately. Current estimates put one-year abandonment at 30 to 50 percent depending on the population and the definition. The typical reasons are boring and mechanical.
- Mask fit. The single largest reversible cause of CPAP failure. Wrong size, wrong style, air leaks around the seal, pressure sores on the bridge of the nose. Trying multiple masks with a good DME provider fixes most of this.
- Dry mouth and nasal congestion. Heated humidification, integrated in most modern machines, largely solves this. Older devices without heated humidification are much worse.
- Claustrophobia and psychological adjustment. Real. Desensitization protocols and cognitive-behavioral support meaningfully improve adherence in patients who struggle with this. Nasal pillow masks feel less enclosing than full-face masks and help some patients.
- Noise. Modern machines are quieter than they were 15 years ago, but partner sleep is still a legitimate concern.
- Air pressure discomfort. Auto-titrating machines (APAP) adjust pressure dynamically and are better tolerated than fixed-pressure CPAP for many users. Bi-level (BiPAP) drops the pressure on exhale and helps a subset.
The honest read on compliance: most CPAP failure is not the patient's fault and is not intrinsic to the therapy. It is a mask-fit problem, a humidification problem, or a psychological-adjustment problem, all of which have known solutions. Aggressive follow-up in the first 90 days, with a real DME provider and a sleep clinic that answers the phone, is the single largest predictor of long-term adherence.
Alternatives when CPAP truly does not work
Some patients genuinely cannot tolerate CPAP after a thorough trial. The alternatives are worth knowing.
Mandibular advancement devices
Custom-fit oral appliances that hold the lower jaw forward during sleep, keeping the airway more open. Effective for mild-to-moderate OSA, less effective for severe. Better adherence than CPAP because they are less intrusive. Should be fit by a qualified dentist with sleep-medicine training, not ordered off the internet. Follow-up sleep study is standard to confirm the device is actually reducing AHI.
Hypoglossal nerve stimulation (Inspire)
An implanted device that stimulates the hypoglossal nerve during sleep, moving the tongue forward and keeping the airway open. STAR trial showed real reductions in AHI in selected moderate-to-severe patients who had failed CPAP. Not for everyone — anatomic and severity criteria apply, and it is a surgical implantation. Real option for patients who genuinely cannot tolerate CPAP.
Positional therapy
A subset of OSA is positional — worse on the back, much better on the side. Positional devices that prevent supine sleep (from tennis-ball-in-a-shirt to purpose-built vibrating positioners) can substantially reduce AHI in the right patient. Home sleep testing that documents positional dependence identifies these patients.
Weight loss
Meaningful weight reduction — 10 percent or more of body weight — reduces AHI in obese OSA patients. Not a substitute for CPAP in severe disease, but a real disease-modifying intervention over time. Bariatric surgery produces the largest reductions and can occasionally lead to OSA resolution.
Positional and lifestyle
Alcohol and sedatives make OSA meaningfully worse and should be reduced. Sleeping on the side rather than the back helps a subset. Smoking cessation modestly improves upper airway function.
Home sleep apnea testing has expanded access
Home sleep apnea tests (HSAT) — small devices patients wear at home for a night or two — have expanded testing access substantially in the past decade. They are less sensitive than in-lab polysomnography and can under-detect events, but they are cheaper, faster to arrange, and appropriate for adults with a moderate-to-high pre-test probability of OSA and no complicating conditions.
The workflow now is often: STOP-BANG positive, HSAT ordered, HSAT confirms OSA, CPAP prescribed. That is a legitimate path for uncomplicated cases. Complex patients — significant cardiopulmonary disease, suspected central apnea, unexplained hypoventilation, prior treatment failures — still need lab-based polysomnography.
Where a sleep tracker fits (and where it does not)
Consumer sleep trackers are not diagnostic devices. They cannot make an OSA diagnosis, they should not be used to titrate treatment, and their oxygen desaturation detection is far below medical-grade pulse oximetry. That is the mandatory disclaimer, and it is real.
What a good non-contact tracker can do is flag breathing disturbances that warrant medical follow-up. Withings' under-mattress mat detects respiratory patterns and identifies breathing interruptions, and the company publishes their approach openly. It is not a diagnosis. It is a flag — a signal that some percentage of nights show breathing patterns that look like they could be OSA, at which point the sensible next step is to see a physician and get real testing.
Read that carefully. "Not diagnostic" is doing real work in that sentence. A consumer device that says "possible sleep apnea" is telling you to go get tested. It is not telling you that you have OSA. The evidence in this article — the SAVE trial, the compliance data, the alternatives — applies to people who have been properly diagnosed by a sleep physician.
The honest bottom line
CPAP is one of the strongest interventions in sleep medicine for one specific condition. The daytime sleepiness benefit is large and reproducible. The cognitive and quality-of-life benefits are meaningful. The cardiovascular benefit in the largest RCT — SAVE — was smaller than the field expected, but the therapy still has clear symptomatic value and probably long-term value in adherent users.
The compliance problem is real and mostly solvable. Mask fit, humidification, and 90-day follow-up account for most of the improvable adherence. Alternatives — MADs, Inspire, positional therapy, weight loss — exist for the patients who cannot make CPAP work after a serious trial.
If you snore, feel unrefreshed, and score positive on STOP-BANG, get tested. A consumer sleep tracker that shows unusual breathing patterns is not a diagnosis, but it is a legitimate reason to escalate. OSA is common, damaging, and treatable. The therapy works. The delivery of the therapy is what needs the work.
