Sleep Apnea

Sleep Apnea

Sleep apnea is several mechanistically distinct disorders that share a counting statistic. The apnea–hypopnea index is a useful case-finding and trial-entry tool; it is not a complete measure of hypoxic burden, sleep fragmentation, or cardiovascular risk. Positive airway pressure reliably reduces scored events and sleepiness when it is used. In adequately powered randomized trials of secondary cardiovascular prevention, it has not reduced the composite events that observational cohorts attributed to untreated apnea. That gap is the article’s controlling fact.

Compiled by South Beach Longevity · 20 August 2026 Copyright 2026 Series SBL-41 / SP-SLEEP-APNEA · Register A scientific article Sources peer-reviewed trials, cohorts, clinical-practice guidelines, and labelled secondary reporting · verified NCBI records Constraint This document describes published research. It is not medical advice. No human use, dose, route or schedule is recommended anywhere in this document.

How to read this document Every finding is labelled, in the sentence that reports it, by the kind of study that produced it. An apnea–hypopnea index is a count, not a disease. A cohort association is not a prevention trial. A per-protocol adherent subgroup is not an intention-to-treat result. A wearable signal is not a scored respiratory event. Amounts, pressures, and durations appear only as reported study or guideline parameters, always with the population attached. Nothing here is a recommendation for any person.

01 Three syndromes, one index

Obstructive sleep apnea is recurrent upper-airway collapse during sleep. Central sleep apnea is recurrent cessation of ventilatory effort. Complex, or treatment-emergent, central sleep apnea is the appearance of central events after an obstructive disorder is mechanically opened (Morgenthaler et al., 2006; Zhang and Wang, 2020). Those three objects share an index. They do not share a mechanism, a first-line device, or a cardiovascular trial literature.

The apnea–hypopnea index (AHI) is events per hour of sleep, or of recording time on a home test. An apnea is a near-cessation of airflow. A hypopnea is a lesser reduction plus a desaturation, an arousal, or both. The rule that defines the lesser reduction is the entire claim (Redline et al., 2007; Ruehland et al., 2009; Berry et al., 2012). Change the rule and the same night becomes a different disease.

Gottlieb and Punjabi’s 2020 review is the clean contemporary map: the syndrome is common, the index is convenient, and the treatment literature is stronger for sleepiness than for hard vascular events (Gottlieb and Punjabi, 2020). This article keeps that map and refuses to collapse it.

02 The event and the scoring rule

The 2007 American Academy of Sleep Medicine (AASM) manual offered more than one hypopnea definition. Ruehland and colleagues applied the competing rules to the same recordings and showed that the AHI moves with the rule: a “recommended” 4% desaturation criterion and an “alternative” 3% or arousal criterion do not produce interchangeable caseness (Ruehland et al., 2009). Berry and colleagues later restated the scoring update (Berry et al., 2012). Grigg-Damberger recorded how the field was still living with that plurality four years after the manual (Grigg-Damberger, 2012). Redline and colleagues had already shown that respiratory-event scoring is reliable only inside a stated rule (Redline et al., 2007).

The practical consequence is not pedantry. Prevalence estimates, trial entry, device reimbursement, and “severe disease” all inherit the rule. HypnoLaus used a 3% or arousal definition in an older general population and found rates that would empty a clinic if they were treated as the 1990s syndromic figure (Heinzer et al., 2015). That is a methods fact before it is an epidemic.

03 Prevalence depends on the rule

Young’s 2002 population-health review remains the required origin story: the Wisconsin Sleep Cohort made sleep-disordered breathing a measured community fact rather than a clinic curiosity, and the classic “syndrome” figure required both a laboratory count and sleepiness (Young et al., 2002). Peppard and colleagues later re-estimated the same cohort’s laboratory prevalence for 2007–2010 and reported moderate-to-severe sleep-disordered breathing (AHI ≥15) in 13% of men and 6% of women aged 30–70 years — higher than the cohort’s earlier decades (Peppard et al., 2013). Those percentages are AHI caseness. They are not the 1990s sleepy-syndrome percentages.

Heinzer and colleagues, in HypnoLaus, scored 2,121 adults from a Swiss population sample with a 3% desaturation or arousal hypopnea rule. Moderate-to-severe sleep-disordered breathing (AHI ≥15) was present in 49.7% of men and 23.4% of women (Heinzer et al., 2015). The paper is not a claim that half of Swiss men need a machine. It is a claim that a sensitive rule in an older sample finds events everywhere.

Senaratna and colleagues’ systematic review of general-population studies found prevalence ranges that swing with age, sex, and scoring (Senaratna et al., 2017). Benjafield and colleagues then scaled published estimates to a global 30–69-year population and arrived at about 936 million adults with AHI ≥5 and about 425 million with AHI ≥15 (Benjafield et al., 2019). A literature-based global total inherits every local rule. It is a burden sketch, not a census.

04 Snoring is not the disease

Snoring is vibration of pharyngeal tissue. It can exist without scored events. Scored events can exist without a bed-partner complaint. Young’s epidemiology review treated snoring as a related symptom and a risk marker, not as caseness (Young et al., 2002). The AASM oral-appliance guideline addresses snoring and obstructive sleep apnea as overlapping but not identical indications (Ramar et al., 2015). A wearable that detects snoring has not diagnosed apnea. A night without snoring has not excluded it.

05 Sex and age

Men have higher laboratory prevalence than women at the same age in Wisconsin and in HypnoLaus (Peppard et al., 2013; Heinzer et al., 2015). The gap narrows after menopause in the epidemiologic reviews; it does not vanish (Young et al., 2002; Senaratna et al., 2017). Women are under-represented in older clinic series relative to community counts — a detection problem as well as a biology problem (Young et al., 2002; Gottlieb and Punjabi, 2020).

Age raises AHI. HypnoLaus is the blunt demonstration: a mean age in the sixties and a sensitive rule produce rates that would be unbelievable if the 1990s sleepy-syndrome number were the target (Heinzer et al., 2015). Whether those late-life events carry the same symptom and vascular meaning as events in a sleepy 45-year-old man is the overdiagnosis question, not a solved fact. Young’s Wisconsin mortality follow-up was strongest for the more severe laboratory categories in a working-age cohort (Young et al., 2008). Punjabi’s Sleep Heart Health mortality analysis likewise treated severity as graded, not binary (Punjabi et al., 2009).

06 Upper-airway anatomy

The adult obstructive event is a collapsible pharyngeal tube. Soft palate, tongue base, lateral walls, and epiglottis can each be the site of narrowing; they are not interchangeable surgical targets. Eckert and colleagues’ phenotypic series treated anatomy as one of four traits, measured as a critical closing pressure, not as a photograph (Eckert et al., 2013). Sands and colleagues later extracted related pharyngeal traits from ordinary polysomnography (Sands et al., 2018). Wellman and colleagues published a simplified physiologic method for the same purpose (Wellman et al., 2013).

Anatomy is necessary. It is not sufficient. Two people with similar collapsibility can have different AHI if loop gain, arousal threshold, and dilator-muscle responsiveness differ (Eckert et al., 2013; Eckert, 2018). That is why a counting index and a collapse site are different clinical objects.

07 Obesity

Obesity is the strongest reversible anatomic load in adult obstructive sleep apnea. Fat in the tongue and surrounding the pharynx narrows the lumen; abdominal mass reduces caudal traction and lung volume. Peppard’s updated Wisconsin estimates rose with the cohort’s weight gain across decades (Peppard et al., 2013). Benjafield’s global sketch is, in large part, an obesity sketch (Benjafield et al., 2019).

Weight is not the whole disease. Lean craniofacial obstruction exists. HypnoLaus found substantial AHI in a European sample that is not a bariatric clinic (Heinzer et al., 2015). The official American Thoracic Society weight-management statement therefore treats weight reduction as disease-modifying for many adults and as incomplete for others (Hudgel et al., 2018). Tuomilehto and colleagues’ randomized lifestyle trial in mild disease is the required early demonstration that a weight-loss programme can reduce AHI as a first intervention (Tuomilehto et al., 2009). Malhotra and colleagues’ SURMOUNT-OSA trials later showed that a dual incretin agonist that produces substantial weight loss also reduces AHI, with or without concurrent positive airway pressure (Malhotra et al., 2024). Weight loss is an apnea treatment in those trials. It is not a proof that every remaining event was “just fat.”

08 Craniofacial structure

Maxilla, mandible, hyoid position, and soft-tissue volume set the bony box in which the fat and the tongue sit. Ethnic and familial craniofacial differences appear in the epidemiologic reviews as a residual after body-mass index (Young et al., 2002; Senaratna et al., 2017). Oral-appliance therapy and maxillomandibular surgery are anatomic replies to that residual (Ramar et al., 2015; Strollo et al., 2014). They are not replies to high loop gain. Eckert’s review is explicit: targeted therapy requires knowing which trait is abnormal (Eckert, 2018).

09 Ventilatory control and loop gain

Loop gain is the ventilatory controller’s tendency to overshoot. High loop gain destabilizes breathing even when the airway is only modestly collapsible. It is a central mechanism that can look obstructive on an airflow channel if collapse is secondary (Eckert et al., 2013; Wellman et al., 2013; Eckert, 2018). Central sleep apnea in heart failure is the clinical extreme of that controller problem: Cheyne–Stokes respiration, high plant and controller gain, and a narrow carbon-dioxide reserve (Bradley et al., 2005; Cowie et al., 2015).

Opening the airway with pressure can unmask the controller. That is one account of treatment-emergent central events (Morgenthaler et al., 2006; Zhang and Wang, 2020). Acetazolamide has been studied as a loop-gain drug in physiologic protocols; this article does not treat those protocols as a licensed indication and records them only as trait evidence (Eckert, 2018). The trait is the point. The index is not.

10 Arousal threshold and muscle responsiveness

A low arousal threshold ends the respiratory event early. That can protect oxygenation and wreck sleep continuity. A high threshold lets the event deepen. Dilator-muscle responsiveness determines whether a narrowing is rescued without waking (Eckert et al., 2013; Sands et al., 2018). Eckert’s 2018 review assembled these traits as a map for non-anatomic therapy — hypnotic effects on threshold, noradrenergic/antimuscarinic combinations on muscle, carbonic-anhydrase inhibitors on loop gain — and as a warning that one drug will not treat four traits (Eckert, 2018). Taranto-Montemurro and colleagues’ atomoxetine-plus-oxybutynin nights are a pharmacologic proof of the muscle-and-threshold idea, not a completed therapeutic class (Taranto-Montemurro et al., 2019; Aishah et al., 2023).

11 Central and treatment-emergent apnea

Bradley and colleagues’ CANPAP trial randomized adults with heart failure and central sleep apnea to continuous positive airway pressure or usual care. CPAP reduced the central index, improved ejection fraction and norepinephrine, and did not improve transplant-free survival on the primary analysis (Bradley et al., 2005). Arzt and colleagues’ post-hoc analysis found better survival in the subgroup whose AHI fell below 15 — a suppression analysis, not a randomized contrast (Arzt et al., 2007). That paper is the required exhibit of adherence-and-response bias in a central-apnea trial.

Cowie and colleagues’ SERVE-HF trial randomized adults with systolic heart failure and predominant central sleep apnea to adaptive servo-ventilation plus usual care or usual care alone. The primary composite was not reduced. All-cause and cardiovascular mortality were higher with servo-ventilation (Cowie et al., 2015). A device that regularizes the nocturnal tracing is not, in that population, a survival device.

Morgenthaler and colleagues described complex sleep apnea as central events appearing once obstruction is treated (Morgenthaler et al., 2006). Zhang and Wang reviewed treatment-emergent central sleep apnea as a distinct, often transient, pattern (Zhang and Wang, 2020). Berger and colleagues showed that the pattern is not confined to positive airway pressure: mandibular-advancement devices and other non-pressure therapies can also unmask central events (Berger et al., 2021; Hellemans et al., 2023). The syndrome is a controller response to a newly patent airway, not a brand of mask.

12 Hypoxemia and fragmentation

The scored event is a duration-plus-amplitude object. The AHI discards both once the event has crossed a threshold. Azarbarzin and colleagues computed a hypoxic burden — the area under the desaturation curve associated with respiratory events — in the Osteoporotic Fractures in Men (MrOS) sleep study and found that this burden predicted cardiovascular mortality after AHI was taken into account (Azarbarzin et al., 2019). That is the required empirical crack in AHI-as-severity.

Yaffe and colleagues, in older women, found that sleep-disordered breathing and nocturnal hypoxia were associated with later mild cognitive impairment or dementia (Yaffe et al., 2011). Quan and colleagues, in APPLES, found that the cross-sectional cognitive signal of AHI was modest and not a simple linear function of the index (Quan et al., 2011). Fragmentation and hypoxia are candidate mediators. They are not interchangeable with event counts.

Punjabi and colleagues’ insulin-resistance study in middle-aged overweight men associated sleep-disordered breathing with insulin resistance after body-mass adjustment (Punjabi et al., 2002). Hirotsu and colleagues later found incident metabolic syndrome associated with obstructive sleep apnea in the joined Episono and HypnoLaus samples (Hirotsu et al., 2018). Those papers are metabolic associations. They are not CPAP-prevention trials.

13 Polysomnography

Attended polysomnography is the reference standard in the AASM diagnostic guideline: airflow, effort, oxygen saturation, electroencephalography, and the rest of the montage that can score sleep, arousals, and respiratory events under a stated rule (Kapur et al., 2017; Berry et al., 2012). It is also expensive, scarce, and a single night. Night-to-night variability is a real limit of any one-night test (Gottlieb and Punjabi, 2020). NICE NG202 likewise treats sleep-study type as a function of pre-test probability and comorbidity, not as a universal laboratory night (National Institute for Health and Care Excellence, 2021).

Polysomnography can support trait estimation (Wellman et al., 2013; Sands et al., 2018). In ordinary practice it reports AHI, oxygen-desaturation index, sleep architecture, and a narrative. The narrative is not in the index.

14 Home sleep-apnea testing

Unattended portable monitors record a subset of channels, usually without electroencephalography. Collop and colleagues’ 2007 guideline and 2011 technology evaluation set the adult rules: home testing is for high pre-test probability of moderate-to-severe uncomplicated obstructive disease; it is not a general substitute for polysomnography in heart failure, neuromuscular disease, hypoventilation, or suspected central disease (Collop et al., 2007; Collop et al., 2011). Kapur and colleagues’ 2017 AASM guideline restated that boundary (Kapur et al., 2017).

El Shayeb and colleagues’ systematic review found that level-3 portable tests can approximate level-1 AHI in the populations they were designed for, with misclassification at the edges and in less selected samples (El Shayeb et al., 2014). A home test that reports “AHI” is often reporting events per hour of recording, not of sleep. Wake time dilutes the index. That is a built-in underestimation risk, not a software bug.

15 AHI, RDI, and the oxygen-desaturation index

AHI is apneas plus hypopneas per hour. The respiratory disturbance index (RDI) adds respiratory-effort-related arousals when the montage can score them. The oxygen-desaturation index (ODI) counts desaturations of a stated depth, usually 3% or 4%, and does not require an airflow event. The three numbers answer three questions. They are not aliases.

Ruehland’s scoring paper is the required demonstration that AHI is a rule, not a vital sign (Ruehland et al., 2009). Redline’s reliability paper is the required demonstration that the rule must be stated (Redline et al., 2007). An ODI can be closer to hypoxic burden than an AHI built on arousal-only hypopneas; it can also miss non-desaturating events that wreck sleep. Neither substitution is free.

16 Hypoxic burden

Azarbarzin and colleagues’ MrOS analysis is the landmark: event-related desaturation area predicted cardiovascular mortality more informatively than AHI (Azarbarzin et al., 2019). The metric is still a research object in most laboratories. It is already a conceptual rebuke. Depth and duration are part of the exposure. Frequency is not the exposure.

17 Why AHI is incomplete

AHI ignores event duration, desaturation depth, hypoxic area, sleep-stage distribution, position, arousal intensity, and the trait mix that produced the event (Eckert, 2018; Azarbarzin et al., 2019; Gottlieb and Punjabi, 2020). It is unstable across scoring rules (Ruehland et al., 2009). It is diluted on home tests that cannot score sleep (Collop et al., 2011; El Shayeb et al., 2014). It is a superb trial-entry tool because it is standardized enough to randomize. It is a poor disease definition because two people with AHI 30 can have different hypoxia, different sleepiness, and different collapse physiology.

The red-team charge that AHI is treated as a complete disease metric is therefore a charge against a real practice, not a straw man.

18 Wearables and consumer detection

The AASM position statement on consumer sleep technology is blunt: these devices are not diagnostic instruments, are not substitutes for sleep studies, and should not be used to self-diagnose or to titrate therapy (Khosla et al., 2018). de Zambotti and colleagues reviewed wearable sleep technology as a research and consumer class whose sleep–wake performance is better than its stage performance and whose respiratory claims require separate validation (de Zambotti et al., 2019). An earlier consumer-tracker evaluation showed systematic mis-estimation of sleep relative to laboratory scoring (de Zambotti et al., 2015).

Photoplethysmography and accelerometer “apnea scores” can correlate with ODI or AHI in selected samples. Correlation is not equivalence. A false negative in a sleepy commercial driver is a different error from a false positive in an anxious self-tracker. The AASM statement exists because both errors were already arriving in clinic (Khosla et al., 2018).

19 Diagnostic matrix

The table is the first required artifact. A person can be home-test “negative,” wearable “positive,” and trait-abnormal on the same week. Those are not three measurements of one thing.

20 Sleepiness

Sleepiness is the outcome positive airway pressure treats most honestly. Weaver and colleagues’ CATNAP trial randomized sleepy patients with milder obstructive sleep apnea to CPAP or sham and found a sleepiness benefit (Weaver et al., 2012). Patil and colleagues’ AASM systematic review and guideline treat sleepiness, quality of life, and blood pressure as the endpoints with the cleanest PAP evidence (Patil et al., 2019a; Patil et al., 2019b). McEvoy and colleagues’ SAVE trial, which did not reduce its cardiovascular composite, still improved Epworth scores, mood, and work days among users (McEvoy et al., 2016).

Sleepiness is not AHI. Quan’s APPLES baseline paper found only modest cognitive-test associations with AHI (Quan et al., 2011). A nonsleepy person with a high AHI is a different trial population from a sleepy person with a milder AHI (Barbé et al., 2012; Weaver et al., 2012; Peker et al., 2016). Collapsing them into “sleep apnea treatment” is how cardiovascular nulls get misread as “CPAP does nothing.”

21 Cognition

Yaffe and colleagues followed older women and found that sleep-disordered breathing and hypoxia were associated with later mild cognitive impairment or dementia (Yaffe et al., 2011). That is a prospective association in one sex and age band. Quan and colleagues’ APPLES analysis is the required counterweight: cross-sectional neurocognitive performance was not a simple AHI gradient (Quan et al., 2011). Gottlieb and Punjabi’s review treats the cognitive literature as mixed and the randomized cognitive effect of CPAP as small outside of sleepiness (Gottlieb and Punjabi, 2020). Hypoxia is a more biologically plausible cognitive exposure than event frequency. It is still not a dementia diagnosis.

22 Blood pressure

Martínez-García and colleagues’ HIPARCO trial randomized adults with resistant hypertension and obstructive sleep apnea to CPAP or usual care. Twenty-four-hour mean blood pressure fell more with CPAP; the nocturnal increment was larger than the daytime increment (Martínez-García et al., 2013). Patil’s guideline treats blood-pressure reduction as a real, modest, PAP-supported effect (Patil et al., 2019a). Barbé and colleagues’ randomized trial in nonsleepy patients did not show a significant reduction in incident hypertension or cardiovascular events on the primary analysis (Barbé et al., 2012).

The blood-pressure literature is therefore not a slogan. In resistant hypertension with apnea, CPAP moved ambulatory pressure. In nonsleepy outpatient apnea, it did not clearly prevent new hypertension. A 2–3 mm Hg trial mean is not a clinic anecdote and is not a myocardial-infarction surrogate.

23 Arrhythmia

Gami and colleagues showed a nocturnal clustering of sudden death in obstructive sleep apnea (Gami et al., 2005) and, in a cohort, an association of apnea with incident atrial fibrillation after obesity was considered (Gami et al., 2007). Mehra and colleagues’ 2022 American Heart Association scientific statement on sleep-disordered breathing and arrhythmia reviews mechanisms — nocturnal autonomic swings, atrial stretch, hypoxia — and the observational AF literature, and it does not convert those associations into a claim that CPAP prevents stroke from AF (Mehra et al., 2022). Somers and colleagues’ 2008 AHA/ACC statement had already set the same boundary in broader cardiovascular language (Somers et al., 2008).

24 Observational cardiovascular disease

Gottlieb and colleagues’ Sleep Heart Health analysis found that obstructive sleep apnea was associated with incident heart failure in men and with coronary disease in some strata (Gottlieb et al., 2010). Yaggi and colleagues’ clinic cohort associated obstructive sleep apnea with stroke and death after adjustment (Yaggi et al., 2005). Young and colleagues’ eighteen-year Wisconsin follow-up associated severe laboratory sleep-disordered breathing with all-cause mortality (Young et al., 2008). Punjabi and colleagues’ Sleep Heart Health mortality study found a graded association, clearer in men under 70 (Punjabi et al., 2009). Somers and colleagues assembled the physiologic and observational case that sleep apnea is a cardiovascular risk marker (Somers et al., 2008).

Those papers are consistent on one point: people with higher AHI die and infarct more often than people with lower AHI, after imperfect adjustment. They are not consistent with a later claim that treating the index must therefore prevent the event. Residual confounding by obesity, socioeconomic status, adherence behaviour, and undiagnosed vascular disease is the standing alternative (Yu et al., 2017; Gottlieb and Punjabi, 2020).

25 Randomized cardiovascular prevention

McEvoy and colleagues randomized 2,717 adults with moderate-to-severe obstructive sleep apnea and established coronary or cerebrovascular disease to CPAP plus usual care or usual care alone. Mean CPAP use was 3.3 hours per night. The primary composite of cardiovascular death, myocardial infarction, stroke, or hospitalization for heart failure, acute coronary syndrome, or transient ischemic attack was not reduced (hazard ratio 1.10, 95% confidence interval 0.91–1.32) over a mean 3.7 years (McEvoy et al., 2016). Sleepiness and quality of life improved.

Peker and colleagues’ RICCADSA trial randomized revascularized coronary patients with nonsleepy obstructive sleep apnea to CPAP or no CPAP. The intention-to-treat primary cardiovascular composite was not significantly reduced. Analyses restricted to users who met an hours-per-night threshold suggested benefit (Peker et al., 2016). That contrast is the adherence-bias exhibit, not a second primary endpoint.

Sánchez-de-la-Torre and colleagues’ ISAACC trial randomized patients with acute coronary syndrome and obstructive sleep apnea to CPAP or usual care. CPAP did not reduce the recurrent cardiovascular composite (Sánchez-de-la-Torre et al., 2020). Barbé and colleagues’ nonsleepy trial, already cited, was null on incident hypertension and cardiovascular events (Barbé et al., 2012).

Yu and colleagues meta-analysed randomized trials of positive airway pressure and found no association with reduced cardiovascular events or death (Yu et al., 2017). That meta-analysis is the required synthesis. It does not say CPAP is useless. It says the observational vascular story did not survive randomization on the composites that were tested, in the populations that were tested, at the adherence that was achieved.

Two design facts travel with every null in that set. First, SAVE’s 3.3 hours of mean nightly use is not a secret failure of the trialists; it is the adherence the field actually has (McEvoy et al., 2016; Rotenberg, Murariu, and Pang, 2016). A hypothesis that “more hours would have won” is a hypothesis about a different intervention. Second, these were secondary-prevention or nonsleepy samples already receiving contemporary coronary care. They are the populations in which an incremental device effect is hardest to see, and they are also the populations to which the observational slogan was most often applied.

The honest remainder is therefore narrow. Sleepiness still moved. Ambulatory pressure can move in resistant hypertension (Martínez-García et al., 2013). Crash risk has an observational treatment signal (Tregear et al., 2010). Hard secondary-prevention composites did not. Figure 2 is that remainder, not a verdict on every sleepy adult who has never had an infarct.

26 Stroke

Yaggi and colleagues associated clinic-diagnosed obstructive sleep apnea with stroke and death (Yaggi et al., 2005). SAVE included cerebrovascular disease in its entry criteria and in its composite; CPAP did not reduce that composite (McEvoy et al., 2016). Yu’s meta-analysis likewise found no randomized stroke benefit for positive airway pressure (Yu et al., 2017). The observational stroke signal is real as an association. The prevention trial is not a confirmation.

27 Metabolic disease

Punjabi and colleagues associated sleep-disordered breathing with insulin resistance in middle-aged overweight men after body-mass adjustment (Punjabi et al., 2002). Hirotsu and colleagues associated obstructive sleep apnea with incident metabolic syndrome in two population samples (Hirotsu et al., 2018). Hudgel and colleagues’ ATS statement treats weight as both a cause and a treatment target (Hudgel et al., 2018). Malhotra and colleagues’ tirzepatide trials moved AHI by moving weight (Malhotra et al., 2024). CPAP’s randomized metabolic effects are smaller and inconsistent in the reviews that examine them (Gottlieb and Punjabi, 2020; Patil et al., 2019b). Apnea and metabolism share fat, sleep loss, and sympathetic activation. Sharing a pathway is not a CPAP-for-diabetes claim.

28 Mortality

Young and colleagues’ Wisconsin follow-up and Punjabi and colleagues’ Sleep Heart Health analysis are the required observational mortality papers (Young et al., 2008; Punjabi et al., 2009). Yu and colleagues’ randomized meta-analysis is the required negative on death as a PAP endpoint (Yu et al., 2017). SERVE-HF is a mortality harm signal for adaptive servo-ventilation in systolic heart failure with central apnea (Cowie et al., 2015). Those three sentences do not cancel. They describe three designs.

29 Accidents

Tregear and colleagues’ systematic review and meta-analysis found an increased motor-vehicle-crash risk in obstructive sleep apnea (Tregear et al., 2009). A subsequent review found that CPAP use was associated with a reduction in that crash risk (Tregear et al., 2010). The second paper is observational on the treatment side. It is still the strongest accident literature this article holds. Sleepiness, not AHI alone, is the biologically obvious mediator. Commercial-driving policy sits on that literature; it is not a cardiovascular composite.

30 Outcome and cardiovascular matrices

31 CPAP and APAP

Continuous and auto-adjusting positive airway pressure splint the pharynx. They reduce AHI when the mask is on (Patil et al., 2019b; McEvoy et al., 2016). Auto-adjusting devices vary pressure across the night; they are still pressure, not a different disease model. The AASM 2019 guideline recommends PAP for adults with obstructive sleep apnea and excessive sleepiness, and it treats blood pressure and quality of life as supported secondary gains (Patil et al., 2019a). NICE NG202 likewise places PAP in the symptomatic and more severe adult pathway (National Institute for Health and Care Excellence, 2021).

SAVE is the required limit on the cardiovascular sentence, not a limit on the sleepiness sentence (McEvoy et al., 2016). Mean use of 3.3 hours in that trial is also a limit: a therapy that works for a third of the night is not a 24-hour risk-factor drug.

32 Adherence

Weaver and Grunstein reviewed CPAP adherence as the central implementation problem: many adults use the device less than the four-hour threshold trials treat as “good,” and early use predicts later use (Weaver and Grunstein, 2008). Rotenberg, Murariu, and Pang pooled twenty years of published adherence and found a flattened curve — roughly one-third to one-half meeting common hour-and-night criteria, without a clear secular improvement (Rotenberg, Murariu, and Pang, 2016). RICCADSA’s ITT-versus-user split is the causal problem that adherence creates (Peker et al., 2016). People who wear a mask four hours a night are not a random sample of the people who were prescribed one.

33 Oral appliances

Ramar and colleagues’ 2015 AASM/AADSM guideline supports custom mandibular-advancement devices for adults with obstructive sleep apnea who prefer them or who fail PAP, and for snoring (Ramar et al., 2015). Phillips and colleagues randomized CPAP and a mandibular-advancement device and found similar short-term health-outcome scores despite a larger AHI reduction on CPAP — a reminder that the index and the symptom are not the same endpoint (Phillips et al., 2013). Hellemans and colleagues reported treatment-emergent central events with mandibular-advancement devices (Hellemans et al., 2023). Berger and colleagues’ review places those events among non-pressure therapies generally (Berger et al., 2021). An oral appliance is an anatomic therapy. It inherits craniofacial anatomy, dental contraindications, and a smaller mean AHI effect than PAP in head-to-heads.

34 Positional therapy

Supine-predominant obstruction is a real physiologic subclass (Eckert, 2018; Gottlieb and Punjabi, 2020). Devices that discourage supine sleep reduce AHI in that subclass in the reviews that discuss them; they do not treat non-positional or central disease. This bibliography’s strongest positional statement is indirect: trait and review papers name position as a modifier, and guidelines treat positional therapy as adjunctive rather than as a SAVE-level literature (Gottlieb and Punjabi, 2020; National Institute for Health and Care Excellence, 2021). The gap is recorded, not filled with unreplicated device marketing.

35 Weight reduction

Tuomilehto and colleagues randomized a very-low-calorie diet plus lifestyle counselling versus lifestyle advice in overweight adults with mild obstructive sleep apnea. The intervention arm lost more weight and reduced AHI more at one year (Tuomilehto et al., 2009). Hudgel and colleagues’ ATS statement treats comprehensive weight management as indicated for overweight and obese adults with OSA, alongside primary OSA therapy when needed (Hudgel et al., 2018). Malhotra and colleagues’ two SURMOUNT-OSA trials randomized tirzepatide versus placebo in adults with obesity and moderate-to-severe OSA, with and without PAP. Tirzepatide reduced AHI and body weight substantially relative to placebo (Malhotra et al., 2024). That is a pharmacologic weight-loss trial with an apnea primary endpoint. It is not a completed cardiovascular-outcome trial in apnea.

36 Exercise

Exercise appears in the ATS weight statement and in Gottlieb’s review as a small-to-moderate AHI adjunct that does not require the weight loss to be large (Hudgel et al., 2018; Gottlieb and Punjabi, 2020). The mechanism may include fat distribution, rostral fluid, and sleep depth. It is not a substitute for PAP in sleepy severe disease in any guideline held here.

37 Upper-airway surgery

Surgery is site-specific. Palate operations, tongue-base procedures, and skeletal advancement treat different collapse patterns. Gottlieb and Punjabi’s review and the NICE guideline treat surgery as reserved for selected anatomy after evaluation, not as a first-line population therapy (Gottlieb and Punjabi, 2020; National Institute for Health and Care Excellence, 2021). This bibliography does not include a maxillomandibular-advancement meta-analysis with a verified PMID in the locked set; the anatomic logic is taken from the trait papers and the clinical reviews, not from an unreproduced surgical table. Drug-induced sleep endoscopy is a targeting tool, not an outcome trial.

38 Hypoglossal nerve stimulation

Strollo and colleagues’ STAR trial implanted a unilateral hypoglossal-nerve stimulator in adults with moderate-to-severe obstructive sleep apnea who had not accepted or had failed PAP. At twelve months, AHI and oxygen-desaturation index fell, and a withdrawal phase showed return of events when stimulation was turned off (Strollo et al., 2014). The trial is a device-efficacy design in a selected anatomy, not a cardiovascular-outcome trial and not a therapy for complete concentric collapse or morbid obesity as later labelled in practice. It is electrical restoration of a dilator, which is one of Eckert’s four traits (Eckert, 2018).

39 Emerging pharmacologic approaches

Taranto-Montemurro and colleagues randomized a single-night combination of atomoxetine and oxybutynin versus placebo and recorded a large AHI reduction in a physiologic sample (Taranto-Montemurro et al., 2019). Aishah and colleagues tested one-month dosing and found a smaller, still present, effect (Aishah et al., 2023). Those papers are trait-targeted pharmacology. They are not a licensed chronic regimen in this document.

Schweitzer and colleagues’ TONES 3 trial randomized solriamfetol versus placebo for residual excessive daytime sleepiness in treated obstructive sleep apnea and found sleepiness-scale benefit (Schweitzer et al., 2019). Pépin and colleagues randomized pitolisant versus placebo for residual sleepiness in CPAP-adherent adults and found a sleepiness benefit (Pépin et al., 2021). Those drugs treat sleepiness. They do not treat the airway. Using them as “apnea treatment” is a category error.

Malhotra and colleagues’ tirzepatide trials, already cited, are the first large pharmacologic AHI wins that travel with weight loss rather than with residual sleepiness (Malhotra et al., 2024). Loop-gain drugs remain physiologic (Eckert, 2018). This article does not convert any of these agents into a dose, a schedule, or a consumer stack.

40 Oxygen and central-apnea approaches

Supplemental oxygen can reduce loop gain and hypoxic burden in selected physiologic work (Eckert, 2018). It is not first-line obstructive therapy in the AASM PAP or diagnostic guidelines (Patil et al., 2019a; Kapur et al., 2017). For central sleep apnea in systolic heart failure, CANPAP showed that CPAP is not a survival therapy, and SERVE-HF showed that adaptive servo-ventilation can be a harmful one (Bradley et al., 2005; Cowie et al., 2015). Those two trials are the required central-apnea limit. Optimizing heart-failure treatment remains the disease-modifying move in the reviews; it is cardiology, not a mask setting (Somers et al., 2008; Cowie et al., 2015).

41 CPAP-RCT and intervention-comparison matrices

The oral-appliance row is the required dedicated comparison: Phillips and colleagues showed that a smaller AHI win can still match CPAP on short-term health-outcome scores (Phillips et al., 2013). Ramar and colleagues set the practice-guideline boundary (Ramar et al., 2015). Hellemans and Berger record the TECSA catch (Hellemans et al., 2023; Berger et al., 2021). That is the oral-appliance literature this document will defend. It will not invent a cardiovascular appliance trial.

42 AHI as a complete disease metric

The charge is that clinics, payers, and trials treat events per hour as the disease. The evidence agrees that they do, and that they should not. Scoring rules move caseness (Ruehland et al., 2009; Heinzer et al., 2015). Hypoxic burden predicts cardiovascular death after AHI is accounted for (Azarbarzin et al., 2019). Four physiologic traits can produce the same index (Eckert et al., 2013; Eckert, 2018). Home tests dilute the index with wake (Collop et al., 2011). Phillips and colleagues showed that a smaller AHI reduction can match a larger one on the outcomes patients feel (Phillips et al., 2013). AHI remains useful for case-finding and for trial entry. Usefulness is not completeness.

43 CPAP cardiovascular-outcome assumptions

The charge is that “untreated apnea causes heart attacks, therefore CPAP prevents them” was treated as transitive. SAVE, ISAACC, RICCADSA’s intention-to-treat result, Barbé 2012, and Yu’s meta-analysis are the required break in that chain (McEvoy et al., 2016; Sánchez-de-la-Torre et al., 2020; Peker et al., 2016; Barbé et al., 2012; Yu et al., 2017). The trials do not say apnea is harmless. They say that, in the secondary-prevention and nonsleepy populations tested, at the adherence achieved, PAP did not reduce the composites that the observational literature had invited readers to expect.

Two non-excuses must be refused. First, “the trials used too little CPAP” is a hypothesis about a different intervention — more hours — not a re-analysis of the randomized contrast. Second, “the trials enrolled the wrong patients” is partly true (nonsleepy, already on statins and stents) and does not restore the old slogan. It narrows the slogan.

44 Adherence bias

The charge is that observational “CPAP users live longer” studies and per-protocol RCT slices credit the mask for the behaviour of people who wear masks. Weaver and Grunstein and Rotenberg and colleagues document that use is incomplete and stubbornly so (Weaver and Grunstein, 2008; Rotenberg, Murariu, and Pang, 2016). RICCADSA’s user-threshold benefit and Arzt’s CANPAP suppression analysis are the in-trial versions of the same bias (Peker et al., 2016; Arzt et al., 2007). A person who wears a device 6 hours a night also takes other medicines, keeps appointments, and differs in ways no covariate list finishes. Intention-to-treat exists to refuse that compliment.

45 Consumer wearable claims

The charge is that consumer devices diagnose apnea. The AASM position statement forbids that use (Khosla et al., 2018). de Zambotti and colleagues show why: sleep–wake is an easier problem than respiratory events, and published consumer validations do not support diagnostic substitution (de Zambotti et al., 2015; de Zambotti et al., 2019). A trend arrow can still be a prompt to seek a real test. A prompt is not a study.

46 Causal associations with cardiovascular disease

The charge is that the field converted a risk marker into a proven cause and then into a proven treatment target. The observational papers in this bibliography are strong enough to keep apnea on the cardiovascular list (Young et al., 2008; Punjabi et al., 2009; Gottlieb et al., 2010; Yaggi et al., 2005; Gami et al., 2005; Gami et al., 2007; Somers et al., 2008; Azarbarzin et al., 2019). The randomized papers are strong enough to block the transitive treatment claim for secondary-prevention composites (McEvoy et al., 2016; Yu et al., 2017). Both can be true. Causal language that erases the second set is the error.

Reverse causation is live in heart failure: Cheyne–Stokes respiration is as much a consequence of a long circulation time as a cause of the next admission (Bradley et al., 2005; Cowie et al., 2015). Obesity is a common prior of both AHI and infarction (Hudgel et al., 2018). Residual confounding is not a ritual caveat. It is the best current explanation of the gap between Figure 2’s columns.

47 Overdiagnosis and underdiagnosis

Both charges are true in different rooms. Underdiagnosis: Tregear’s crash meta-analysis and the Wisconsin mortality follow-up describe a disease that, when severe and sleepy, is not a wellness fashion (Tregear et al., 2009; Young et al., 2008). Women were under-counted in older clinic series relative to community prevalence (Young et al., 2002; Peppard et al., 2013). Home-test denials in high-probability sleepy adults are a missed-disease problem (Kapur et al., 2017).

Overdiagnosis: HypnoLaus’s 3% rule in an older sample labels a large fraction of the population (Heinzer et al., 2015). Benjafield’s 936 million is a scoring-rule census (Benjafield et al., 2019). Nonsleepy secondary-prevention RCTs did not reward that label with fewer infarctions (Peker et al., 2016; Sánchez-de-la-Torre et al., 2020; McEvoy et al., 2016). Wearables will widen the labelled pool without widening the validated one (Khosla et al., 2018). The honest position is not a midpoint slogan. It is two errors with different victims: the untreated sleepy driver, and the older adult converted into a patient by a rule.

48 How this document was assembled

Filing is SBL-41/SP-SLEEP-APNEA, Register A, parent SBL-41 Science Monograph Series. Identity is South Beach Longevity. Presentation follows the canonical house style with the official circular emblem and full lockup. The constraint line is verbatim. Dual editions are produced through pipeline stages 06, 07, and 07c.

HOUSE_STYLE.md was read as the production standard. Tesamorelin is the reference implementation of the presentation system. This title is not a Radix peptide article and is not filed to the peptide-article library.

Citations were bound only after NCBI confirmation of title, authors, journal, year, and identifier. Landmark searches that returned letters or author replies were discarded. General reference databases were treated as finding aids, not as citation authority. Institutional guidance was consulted directly — the AASM practice-guideline index, NICE NG202, an FDA consumer sleep-apnea page, and the AHA statements; where the 2021 AHA statement on OSA and cardiovascular disease was not retrievable, the 2008 AHA/ACC statement, the 2022 arrhythmia statement, and Gottlieb and Punjabi 2020 carry the guideline-level cardiovascular language instead.

no live adversarial panel was run. The six charges in Part Six were applied as principal objections and answered on the papers.

49 Unresolved science

A consensus clinical replacement for AHI — hypoxic burden, event duration, or a trait panel — has not been locked by a completed surrogate-endpoint trial. Whether higher nightly PAP hours would have moved SAVE’s composite is unknown because that intervention was not randomized. Whether tirzepatide’s AHI reduction will later reduce infarctions is unknown because SURMOUNT-OSA was not a cardiovascular-outcome trial. Whether hypoglossal stimulation or oral appliances prevent hard events is untested at SAVE’s scale. The correct hypopnea rule for older, minimally symptomatic adults remains a methods dispute with prevalence consequences. Adequately powered primary-prevention trials in sleepy, high-hypoxic-burden adults would change Part Four. They have not been done on the design that would settle the slogan.

50 References

51 Evidence handling

Claims in this document are tagged to a design class in the sentence that states them. Observational associations are not rewritten as prevention. Per-protocol slices are not rewritten as intention-to-treat. Scoring-rule dependence is treated as part of the result. Pressures, hours, and milligrams appear only as reported trial or physiologic parameters. No sentence recommends a device, drug, or programme for any person. Third-party figures were not reproduced; the three graphics are original schematics of cited contrasts.