Which Patients Actually Benefit From CPAP?
Two independent analyses found the same split: cardiovascular benefit concentrated in patients with a high hypoxic burden, and a signal pointing the wrong way in those with a low one.
For a decade, trials of CPAP with cardiovascular endpoints kept coming back null, and the field's explanations were unsatisfying — poor adherence, too-short follow-up, the wrong endpoints.
A better explanation has emerged, and two independent datasets now point at it: CPAP's cardiovascular benefit is not distributed evenly across people with sleep apnea. It appears concentrated in one group, and in the other group the signal does not merely vanish.
The first dataset
A secondary analysis of a randomised trial in patients with acute coronary syndrome and sleep apnea who were not sleepy — 362 assigned to CPAP, 365 to usual care, mean age 59.7 years, 84.5% male — split them by baseline hypoxic burden (Eur Respir J, 2023).
A significant interaction was found between treatment arm and hypoxic burden category [ERJ 2023].
In the high hypoxic burden group, CPAP significantly reduced cardiovascular events: HR 0.57 (95% CI, 0.34–0.96) [ERJ 2023].
In the low hypoxic burden group, CPAP-treated patients showed a trend toward higher cardiovascular risk than usual care: HR 1.33 (95% CI, 0.79–2.25) [ERJ 2023].
And the differential effect followed a dose-response relationship with baseline hypoxic burden [ERJ 2023].
Read the second number carefully. Its confidence interval crosses 1, so it is a trend rather than demonstrated harm — but the point estimate sits above 1, and the dose-response across categories is what makes the pattern hard to dismiss as noise.
The second dataset
A separate 2026 analysis, in a different population, found the same shape.
Among 3,549 participants, major adverse cardiovascular and cerebrovascular events occurred in 16.6% on CPAP against 16.3% on usual care — null overall. But the treatment effect was significantly greater in high-risk sleep apnea (interaction HR 0.69; 95% CI, 0.50–0.95; p = 0.024), and the authors reported CPAP benefits in high-risk apnea "observed alongside harm in low-risk OSA" (Eur Heart J, 2026).
Two analyses, different cohorts, different risk definitions, same structure: benefit above a threshold, nothing or worse below it. The 2026 analysis in full.
Why this explains the null trials
If benefit is confined to a subgroup and the opposite may hold in the rest, then a trial enrolling both produces an average of roughly nothing — which is exactly what happened repeatedly.
And the enrolment criterion was the problem. Most trials recruited by apnoea-hypopnoea index, which is a poor predictor of cardiovascular outcomes: in two cohorts totalling nearly 8,000 people, hypoxic burden strongly predicted cardiovascular mortality while the AHI did not. What hypoxic burden measures.
Recruit on a number that does not track the risk you are trying to reduce, and you fill a trial with a mixture of people who stand to benefit and people who do not. The real effects cancel. That is a coherent account of a decade of disappointing results, and it locates the problem in patient selection rather than in CPAP.
What would explain harm in the low-burden group
Honestly: nobody has established a mechanism, and this is the weakest part of the story.
Candidate explanations exist — sleep disruption from the device in people who were not symptomatic to begin with, and the possibility that mild intermittent hypoxia has some conditioning effect that is removed by treatment. Both are speculative.
It is also possible the signal is not real. Two independent analyses seeing it is suggestive; neither was designed to detect harm, and subgroup findings mislead often enough to warrant caution.
Tier 2. Consistent, dose-responsive, reproduced across two datasets — and derived from subgroup and secondary analyses rather than from a trial that allocated treatment by burden in the first place. How we grade evidence.
What this does not change
If you have symptoms, treat. Every bit of this concerns cardiovascular endpoints in people who were largely not sleepy — the first analysis says so explicitly. CPAP's best-supported uses are relieving daytime sleepiness, improving function and quality of life, and treating the apnoea itself. Those are unaffected and are why most people are prescribed it. Getting used to CPAP.
If you have severe apnoea with deep desaturations, you are the high-burden group. The evidence here argues for treating you, not against.
Do not self-assess. "High risk" and "high hypoxic burden" are study classifications, not something to judge from an article. Hypoxic burden is not on standard reports at all.
And do not stop treatment on the strength of this. Adherence decisions belong with your clinician, and for commercial drivers with a medical examiner. Sleep apnea and your DOT physical.
What it does change
The reason for treating matters more than it used to. Treating severe apnoea with deep desaturations in someone with cardiovascular disease rests on different evidence from treating mild, asymptomatic apnoea to prevent a future event. The first is supported. The second is where these analyses counsel caution.
Ask what your oxygen numbers look like, not just your AHI. Oxygen nadir, time below 90% and desaturation index are crude proxies for hypoxic burden and they are on most reports. A mild AHI with repeated drops into the low 80s is a different situation from a mild AHI with shallow dips. How to read your sleep study.
And expect the field to move. If treatment allocation shifts from AHI toward measures of hypoxic load, that changes who gets offered what — which is the practical significance of all this.
The line
No consumer product treats obstructive sleep apnea or affects cardiovascular risk from it. This page is about which patients a real treatment helps, not a reason to substitute anything for it. Snoring versus sleep apnea.