Evidence Trace, complete trial record

In plain terms

In plain terms: CPAP for obstructive sleep apnea has some human trial evidence supporting it, but the evidence is not yet strong enough for EBL to call it established. It is worth knowing about and worth discussing with your clinician.

CPAP for obstructive sleep apnea

B Supported: a good trial, backed by more

Category Devices

Evidence Randomized trials and meta-analyses of CPAP against usual care, sham CPAP or inactive control

Reviewed October 2026

PMIDs 27571048 · 31839558 · 26914592 · 22618923 · 28697252 · 37787793 · 39638419 · 40254559 · 26624827 · 39401854 · 24327037 · 24947425 · 23204602 · 16467072 · 16437429 · 26497082

The endpoint this grade is based on: Surrogate biomarker endpoint

The evidence as the record states it

Evidence record

PMID 27571048 PMID 31839558 PMID 26914592 PMID 22618923 PMID 28697252 PMID 37787793 PMID 39638419 PMID 40254559 PMID 26624827 PMID 39401854 PMID 24327037 PMID 24947425 PMID 23204602 PMID 16467072 PMID 16437429 PMID 26497082

Benning et al. (J Sleep Res, 2025; PMID 40254559), Bratton et al. (JAMA, 2015; PMID 26624827), Pengo et al. (Eur Respir J, 2025; PMID 39401854), Martinez-Garcia et al. (JAMA, 2013; PMID 24327037) and Bratton et al. (Thorax, 2014; PMID 24947425) on blood pressure; Kushida et al. (Sleep, 2012; PMID 23204602), Marshall et al. (Thorax, 2006; PMID 16467072), Giles et al. (Cochrane Database Syst Rev, 2006; PMID 16437429) and Bratton et al. (Lancet Respir Med, 2015; PMID 26497082) on sleepiness; McEvoy et al. (N Engl J Med, 2016; PMID 27571048), Sanchez-de-la-Torre et al. (Lancet Respir Med, 2020; PMID 31839558), Peker et al. (Am J Respir Crit Care Med, 2016; PMID 26914592), Barbe et al. (JAMA, 2012; PMID 22618923), Yu et al. (JAMA, 2017; PMID 28697252) and Sanchez-de-la-Torre et al. (JAMA, 2023; PMID 37787793) on cardiovascular events. Bradley et al. (Eur Respir J, 2024; PMID 39638419) set the observational and randomized cardiovascular evidence side by side: nine observational studies favored treatment and 12 randomized trials, which mostly excluded sleepy patients, did not.

Evidence by specific use

Blood pressure in adults with obstructive sleep apnea

Grade B

Benning et al. (J Sleep Res, 2025; PMID 40254559) pooled 75 randomized trials: office systolic pressure 2.5 mmHg lower and nighttime systolic 3.5 mmHg lower than control. An individual-patient analysis of 9,434 people found the fall confined to those whose pressure was uncontrolled at the start (Pengo et al., Eur Respir J, 2025; PMID 39401854), and minimally sleepy patients showed no significant change (Bratton et al., Thorax, 2014; PMID 24947425). Blood pressure is a surrogate, so the letter stops at B.

PMID 40254559 PMID 26624827 PMID 39401854 PMID 24327037 PMID 24947425

Daytime sleepiness measured by the maintenance of wakefulness test

Grade B

The maintenance of wakefulness test measures how long a person stays awake in a dim room, scored by instrument. CPAP improved it against sham in APPLES, most clearly at an apnea-hypopnea index above 30 (Kushida et al., Sleep, 2012; PMID 23204602), and by 2.1 minutes in pooled trials of mild to moderate sleep apnea (Marshall et al., Thorax, 2006; PMID 16467072). Held at B: the second objective test, the multiple sleep latency test, did not change in the same pooled trials, the effect in milder disease was small, and the trials behind the pooled figure are not itemized here.

PMID 23204602 PMID 16467072 PMID 16437429

Self-rated daytime sleepiness (Epworth scale) in adults with obstructive sleep apnea

Grade B

Bratton et al. (Lancet Respir Med, 2015; PMID 26497082) pooled 67 randomized trials: Epworth scores 2.5 points lower on CPAP than on inactive control, and the Cochrane review agrees (Giles et al., 2006; PMID 16437429). The effect was larger in sleepier patients. Self-reported, so the letter stops at B.

PMID 26497082 PMID 16437429 PMID 27571048

Heart attack, stroke and cardiovascular death in adults with moderate to severe obstructive sleep apnea, mostly without daytime sleepiness

Grade D

Four randomized trials with hard endpoints found no reduction in cardiovascular events: SAVE in 2,717 adults (McEvoy et al., N Engl J Med, 2016; PMID 27571048), ISAACC in 1,264 after acute coronary syndrome (Sanchez-de-la-Torre et al., Lancet Respir Med, 2020; PMID 31839558), RICCADSA in 244 (Peker et al., Am J Respir Crit Care Med, 2016; PMID 26914592) and a Spanish trial in 723 (Barbe et al., JAMA, 2012; PMID 22618923). Pooling SAVE, ISAACC and RICCADSA gave a hazard ratio of 1.01 (Sanchez-de-la-Torre et al., JAMA, 2023; PMID 37787793). Lower event rates in people who used CPAP 4 hours or more appear only in on-treatment analyses, which lose the protection of randomization. The letter does not reach sleepy patients, who were largely excluded, or consistent all-night use, which the trials did not achieve.

PMID 27571048 PMID 31839558 PMID 26914592 PMID 22618923 PMID 28697252 PMID 37787793

Why Grade B and not Grade A

Grade B reflects moderate human evidence -- typically a single strong trial, multiple smaller trials, or surrogate endpoint data without hard clinical outcome replication. To reach Grade A, CPAP for obstructive sleep apnea would require at least one additional independent randomized trial demonstrating a hard clinical endpoint or objectively measured functional outcome in a comparable population.

What this grade covers and what it does not

Grade B on blood pressure (Benning et al., 2025; PMID 40254559) and on daytime sleepiness, both on the maintenance of wakefulness test (Kushida et al., 2012; PMID 23204602; Marshall et al., 2006; PMID 16467072) and on the Epworth scale (Bratton et al., 2015; PMID 26497082). Grade D on Route 2 for heart attack, stroke and cardiovascular death: four randomized trials and their pooled analysis found no reduction (PMIDs 27571048, 31839558, 26914592, 22618923, 37787793). WHY NOT A: blood pressure is a surrogate and the Epworth scale is self-reported; the wakefulness test is instrument-scored, but the multiple sleep latency test did not change in the same pooled trials and the trials behind the pooled figure are not itemized here. WHAT THIS GRADE DOES NOT COVER: cardiovascular outcomes in sleepy patients who use CPAP all night, a group the large trials excluded or did not achieve.

What would move this grade

An independent randomized trial demonstrating a hard clinical endpoint -- mortality, major adverse event, or objectively confirmed functional improvement -- in an adequately powered sample, with results independently replicated, would support Grade A.

How to track whether this is working Apnea-hypopnea index on the device download and on retesting, nightly hours of use, the Epworth Sleepiness Scale, and home blood pressure.