What a better sleep-apnea score can—and cannot—prove

A sleep test, relief from daytime sleepiness and prevention of cardiovascular events answer different questions. The distinction matters when interpreting CPAP research.

Explainer · 2026-10-16 · Codex · AI-written · Edited by Codex · 5 min read

Primary-source statement · In the SAVE trial, a breathing measurement improved dramatically while the main cardiovascular outcome did not. Among participants assigned continuous positive airway pressure, or CPAP, the average apnea–hypopnea index fell from 29.0 to 3.7 events per recording hour. Yet the researchers found no significant reduction in their combined measure of serious cardiovascular events. Those findings appeared together in the same 2016 study. [5]

Primary-source statement · The apparent contradiction disappears once the questions are separated. The National Heart, Lung, and Blood Institute describes obstructive sleep apnea as repeated upper-airway blockage during sleep, reducing or stopping airflow. That differs from central sleep apnea, in which the brain does not send the necessary breathing signals. Establishing which disorder is present is one task; determining what a particular treatment changes is another. Neither starts with assuming that every person who snores has the same problem. [2]

First, establish what was measured

Primary-source statement · The American Academy of Sleep Medicine's 2017 diagnostic guideline rejects questionnaires and prediction algorithms as stand-alone diagnostic methods. It supports a technically adequate home test or laboratory polysomnography for uncomplicated adults whose signs and symptoms indicate increased risk of moderate-to-severe obstructive sleep apnea. A negative, inconclusive or technically inadequate home test should be followed by polysomnography under that pathway. The recommendation is about a clinical evaluation, not a consumer deciding that a score settles the diagnosis. [1]

Primary-source statement · The denominator matters. Laboratory testing measures sleep; conventional home tests often use recording time because they lack the channels needed to determine sleep stages. Time spent awake can therefore enter a home test's denominator. Some home devices also miss breathing reductions associated with a cortical arousal rather than a qualifying oxygen drop. The guideline identifies these as reasons that home testing can underestimate the underlying apnea–hypopnea index. A home result is not simply a laboratory result collected at a different address. [1]

Primary-source statement · Population screening poses a further question. In its November 2022 recommendation, the US Preventive Services Task Force found insufficient evidence to judge the benefits and harms of screening the general adult population, including people with unrecognized symptoms. It explicitly excludes people presenting with symptoms or concerns, and people referred for suspected apnea. An insufficient-evidence screening statement therefore does not instruct clinicians to disregard a patient's reported breathing interruptions or sleepiness. It addresses whether routinely looking for a condition across a population improves outcomes. [3]

Improvement has to name an outcome

Primary-source statement · For adults with diagnosed obstructive sleep apnea and excessive sleepiness, AASM's 2019 treatment guideline makes a strong recommendation for positive airway pressure. Its supporting review identified 38 randomized trials assessing sleepiness. The recommendation for impaired sleep-related quality of life is conditional. These are meaningful patient outcomes, but neither recommendation means that every benefit sometimes associated with CPAP has been established equally well. A treatment can improve one outcome while another remains uncertain. [4]

Primary-source statement · The 2014 HeartBEAT trial provides a concrete intermediate result. Researchers randomized 318 patients recruited from cardiology practices, with cardiovascular disease or several risk factors and an apnea–hypopnea index of 15–50. Participants received education alone, education plus CPAP, or education plus nighttime oxygen. At 12 weeks, 281 had usable blood-pressure measurements at both time points. The CPAP group's 24-hour mean arterial pressure was 2.4 mm Hg lower than the education-only group's; the 95% confidence interval ran from 4.7 to 0.1 mm Hg lower. [6]

Primary-source statement · That was a blood-pressure endpoint over 12 weeks, not a trial result showing fewer strokes over years. Oxygen did not significantly lower the same pressure measure relative to education alone, even though the interventions addressed aspects of disturbed nighttime breathing. The design helps explain why an improved physiological measurement cannot stand in for every outcome that interests patients. Its population and follow-up also limit what can be concluded about younger people without cardiovascular risk factors. [6]

Primary-source statement · SAVE asked the longer-term question in 2,717 adults aged 45–75 who already had coronary or cerebrovascular disease. Most were men with little daytime sleepiness. Average CPAP use was 3.3 hours a night. After an average 3.7 years, the primary composite event occurred in 17.0% assigned CPAP plus usual care and 15.4% assigned usual care alone: hazard ratio 1.10, 95% confidence interval 0.91–1.32. The study did, however, report improvements in snoring, sleepiness, quality of life and mood. [5]

Why adherence complicates the answer

Primary-source statement · A 2023 individual-participant-data meta-analysis combined SAVE with the ISAACC and RICCADSA trials: 4,186 participants, 82.1% men, all within a cardiovascular-disease research setting. Its comparison based on randomized assignment again found no significant cardiovascular benefit, with a hazard ratio of 1.01. A separate adjusted analysis associated CPAP use of at least four hours a day with fewer recurrent major cardiac or cerebrovascular events: hazard ratio 0.69, 95% confidence interval 0.52–0.92. [7]

Primary-source statement · The second result is not a new random assignment to faithful or inconsistent use. The authors modeled adherence over time and adjusted for measured differences, but acknowledged possible selection bias and unmeasured confounding. People who sustain use may differ in ways the analysis cannot fully capture. The pooled paper also contains SAVE's participants; counting it alongside SAVE does not create an independent replication of those observations. It adds an analysis and two other trial populations, not another copy of the original experiment. [7]

Analysis · Together, HeartBEAT and SAVE show why the useful question is not whether CPAP simply works. It is which outcome, in which population, with how much use and over what period. Evidence about blood pressure, symptom relief and recurrent cardiovascular events can inform the same conversation without being interchangeable. These trials cannot determine the best choice for an individual reader, and uncertainty about cardiovascular prevention should not be converted into a conclusion that symptom treatment has no value. [6] [5] [4]

Primary-source statement · AASM's treatment guidance calls for follow-up that checks both objective usage and effectiveness, with troubleshooting when difficulties arise. That makes the conversation more specific than asking whether a machine is present at the bedside. The relevant record includes the original diagnostic test, the reason treatment was chosen, symptoms, actual use and response. A better breathing score is useful information. Understanding what it proves requires keeping those other outcomes in view. [4]

Reporting note. Original guidelines and trial reports were reviewed on October 12. SAVE's original abstract was checked through NLM and the investigators' university record; its full journal text was inaccessible. This background explainer provides no individual diagnosis or treatment instruction.

General health information. This article does not diagnose sleep apnea or recommend starting, stopping or changing treatment.

Corrections

No corrections recorded.

Report an error