Traditionally, the presence and severity of obstructive sleep apnea (OSA) have been defined by the apnea–hypopnea index (AHI). Continuous positive airway pressure is generally first-line therapy despite low adherence, because it reliably reduces the AHI when used, and the response to other therapies is variable. However, there is growing appreciation that the underlying etiology (i.e., endotype) and clinical manifestation (i.e., phenotype) of OSA in an individual are not well described by the AHI. We define and review the important progress made in understanding and measuring physiological mechanisms (or endotypes) that help define subtypes of OSA and identify the potential use of genetics to further refine disease classification. This more detailed understanding of OSA pathogenesis should influence clinical treatment decisions as well as help inform research priorities and clinical study design. In short, treatments could be individualized on the basis of the underlying cause of OSA; patients could better understand which symptoms and outcomes will respond to OSA treatment and by how much; and researchers could select populations most likely to benefit from specific treatment approaches for OSA.
Obstructive sleep apnea (OSA) is a common sleep disorder characterized by the repetitive collapse of the upper airway during sleep, leading to transient hypoxemia and arousals from sleep. Surges in sympathetic activity, repeated oxygen desaturations, and sleep fragmentation are linked to cardiovascular (e.g., hypertension, stroke, and/or myocardial infarction), metabolic (e.g., diabetes), and neurocognitive (e.g., excessive daytime sleepiness and/or automobile/workplace accidents) consequences (1). Recent estimates suggest that roughly 10% of the U.S. population has clinically important OSA (roughly 13% of middle-aged men and 6% of U.S. women), the high prevalence reflecting aging of the population and the obesity pandemic, with both aging and obesity being known contributors to OSA risk (2). In older populations, the prevalence may be much higher (3).
For many reasons, clinicians and researchers have long relied on the apnea–hypopnea index (AHI) for diagnosis and assessment of OSA severity, rather than focusing on symptoms or other markers of disease that might be more useful. First, many of the symptoms of OSA are nonspecific, such as fatigue, and can be difficult to both objectively and reliably characterize. Second, even “asymptomatic” people with OSA may report improvements after a trial of therapy. This approach has been colloquially framed as “when in doubt, pressurize the snout.” Third, payors generally make decisions about treatment coverage on the basis of an individual’s AHI rather than their symptomology. Fourth, some providers were financially rewarded for testing and treatment even in subjects with few symptoms attributable to sleep-disordered breathing. This one-size-fits-all approach has had major repercussions for the sleep medicine field. First, many patients with OSA are unable to tolerate continuous positive airway pressure (CPAP), resulting in suboptimal treatment effectiveness. Second, even among those patients who show good adherence, some will perceive little or no benefit from therapy, further reinforcing the negative perception of CPAP, which may discourage other patients who really would benefit. More concerning for the future of the field is that several large-scale multicenter randomized controlled trials failed to show CPAP definitively reducing the risk of stroke, myocardial infarction, or death or improving neurocognitive outcomes (4–8). The reasons for the disconnect between the epidemiological studies and the interventional ones are likely multifactorial, including low adherence to therapy and long duration of disease—usually years—before treatment, with potential end-organ damage that is not fully reversible. Another possibility is that OSA is in fact multiple different diseases, as is increasingly recognized in other illnesses such as asthma (9), acute respiratory distress syndrome (10), and particularly in cancer, with tumors classified not only by their distinct histology but also by the molecular markers reflecting tumor biology and responsiveness to therapy. The U.S. Preventive Services Task Force recently recommended against testing for OSA in asymptomatic people, based in part on some of these data (11, 12). Finally, few clinicians may enter the field if they see their role as writing prescriptions solely on the basis of an AHI number and with limited opportunity to apply diagnostic skills, which may be one reason for fewer clinicians entering the field (13). Conceptual advances are needed to ensure that the practice and science of sleep medicine remain relevant and responsive to patient and public health needs while integrating modern and emerging scientific innovations.
At the American Thoracic Society’s International Conference held in San Diego (2018), a symposium called “New Paradigms for Resolving Sleep Apnea Heterogeneity: Deep Phenotyping and Genomics” was convened. The goals of the session were to challenge clinicians and researchers to consider a broader array of OSA phenotypes, both for clinical assessments and in research studies. Currently, patients with OSA are lumped into broad categories according to their OSA severity as assessed typically using the AHI. However, it is increasingly recognized that patients with similar AHIs may have vastly different underlying pathophysiological determinants of OSA, symptoms, and prognosis (especially in terms of cardiometabolic disease and mortality). In other words, OSA is a heterogeneous disorder in terms of its pathogenesis and clinical expression. This heterogeneity is not captured by the AHI alone. Furthermore, it would be logical to presume that these varying OSA subtypes might require different types and intensities of treatment.
The current management approach for most people with OSA is a diagnostic sleep study followed by a trial of CPAP therapy if OSA is confirmed. If patients do not tolerate CPAP, treatment reverts to one of several potential alternative therapies, with treatment selection based on many situational factors, including third-party reimbursement policies, and with no guarantee of success. Currently, there is very little done in the way of personalizing OSA therapy in the clinic. In many other areas of medicine, it is routine to characterize diseases according to clinical subtypes that have different treatment options depending on underlying mechanisms, severity, and symptoms. In the case of diabetes, the treatment of the clinical phenotype (hyperglycemia) is dependent on the underlying pathobiological mechanisms or endotypes (i.e., insulin deficiency vs. insulin resistance) that reflect genetic and behavioral risk factors (see Figure 1). Thus, there is a very plausible model which suggests that clinicians should be able to treat OSA in much the same way. The personalized medicine approach to treating OSA must therefore require recognition of this individual variability. Progress has been made recently toward resolving much of the heterogeneity of OSA, and the field is poised to make significant progress toward precision sleep medicine. The purposes of this article are therefore to 1) synthesize the evidence presented in the American Thoracic Society session, 2) highlight the expected benefits of this precision medicine approach, and 3) provide clarity and consensus to relevant terms (e.g., phenotype vs. endotype) used in describing a sleep precision medicine paradigm.
The term “phenotype” has been used differently by different authors. Historically, the word derives from the Greek “phaino” meaning “appearance,” and it was first used in the 1900s in the scientific literature usually to differentiate between the underlying genetic code (genotype) and the observable appearance of the organism that was thought to result from an interaction between genotype and environment. Most recently, Zinchuk and colleagues (14) defined an OSA phenotype broadly as “a category of patients with OSA distinguished from others by a single or combination of disease features, in relation to clinically meaningful attributes (symptoms, response to therapy, health outcomes, quality of life).” Notably, these distinguishing features may be very different, depending on whether we use the patient’s or the clinician’s perspective or base the features on diagnostic testing.
From the patient’s perspective, people with OSA may present with very different symptoms and signs: excessive daytime sleepiness, cardiovascular morbidity (e.g., new atrial fibrillation), disturbed sleep, or few symptoms other than snoring/apneas. Ye and colleagues (15) used cluster analysis based on response to questionnaires and found three different subjectively reported, patient-centric phenotypes: disturbed sleep, excessive daytime sleepiness, and minimally symptomatic. Notably, the clusters did not differ in regard to sex, body mass index, or AHI. These findings have also been reproduced in other cohorts (16), and importantly, the symptom cluster is associated with response to CPAP therapy (17). That is, individuals with excessive daytime sleepiness are those often experiencing the greatest relief from sleepiness. A challenge in this approach is that potential symptoms of OSA may also be due to other common disorders such as insomnia or depression. Thus, attributing disturbed sleep solely to OSA may be difficult in patients with concomitant insomnia, and it may lead to an unsuccessful trial of CPAP with low adherence and worsened insomnia. Currently, there is insufficient information to predict which symptoms identify individuals likely to experience treatment benefits across the domains of sleepiness, quality of life, cognition, cardiometabolic function, and longevity.
Diagnostic characteristics may also be used to define a phenotype. Although the AHI may have some utility in the prediction of cardiovascular events, it does not predict other clinically relevant outcomes well (see Table 1). Moreover, recent research indicates that the AHI may perform less well in women than in men (18), suggesting its limitations across a spectrum of the population. Similarly to Ye and colleagues, Zinchuk and colleagues (19) performed a cluster analysis on a large dataset of polysomnography (PSG) recordings, looking for associations with cardiovascular morbidity and mortality. Inclusion in the identified clusters, which were not defined a priori but on the basis of the data itself, was more predictive of cardiovascular risk than AHI. Although expected associations were found, such as increased risk in those with more profound hypoxemia for the same AHI, other associations, such as highest risk in those with periodic limb movements (PLMs), were more surprising. Although this may be an important advantage of an agnostic or unsupervised approach, it is still not clear if PLMs represent a manifestation of OSA (i.e., a phenotype of OSA) or if they directly result in cardiovascular disease independent of breathing disturbances or are a consequence of another process, such as subclinical renal dysfunction, which itself might convey increased risk of cardiovascular mortality and can lead to PLMs. Other recent research has identified alternative single metrics of OSA severity, such as indices describing the depth and duration of hypoxemia (20) or duration of respiratory events (18), that improve mortality prediction over the AHI.
|Metric for OSA Severity||Associated Clinical Outcome (Reference)|
|4% desaturations||Hypertension (75)|
|2% desaturations||Insulin resistance (76)|
|Integrated depth and duration of oxygen desaturation during the respiratory events||Cardiovascular mortality (20)|
|Respiratory event duration||All-cause mortality (18)|
|Percentage of time below 90%||Platelet aggregation (77)|
|Arousal frequency||Memory consolidation (78), fatigue (79)|
|Epworth Sleepiness Scale score||Subjective improvement with CPAP (80)|
|None||Motor vehicle accident risk (81)|
An alternative way to determine OSA phenotypes is to assess the response to therapy. For example, although CPAP can improve many consequences of OSA, the impact can be quite variable in terms of factors such as blood pressure and sleepiness. For example, some patients with OSA have increased blood pressure when CPAP is applied (21). It is becoming clear that treating “all comers” with CPAP will have modest benefits in terms of blood pressure, but with individual variability that is, again, not well predicted by AHI (22). Instead, stratification by some means is necessary to understand if the desired benefit is likely to result from OSA treatment (i.e., whether blood pressure is expected to improve with CPAP). However, it is not clear how such a strategy would be implemented without better knowledge of which patients are at risk for which outcomes. In addition, an understudied area is whether the emergence of central apneas with treatment that may be associated with higher residual AHIs and may lead to reduced CPAP adherence (23) identifies a subset of individuals with underlying physiology/endotype, specifically those with a more unstable ventilatory control system (higher loop gain) (24). However, determining OSA phenotypes by assessing responses to therapy is limited by the multiplicity of reasons that treatment response may be inadequate. Treatment response reflects not only the efficacy of treatment when therapy is used as prescribed but also variable adherence with prescribed therapy due to behavioral or other factors and/or persistence or emergence of symptoms due to treatment side effects or comorbid conditions. This is a common clinical scenario: Will the patient struggling with CPAP eventually experience symptom relief with good adherence to CPAP (i.e., redouble efforts at adherence) or not (i.e., reasonable to abandon therapy)?
Although much work remains to be done, there is increasing evidence that the use of biomarkers (blood, urine, or other) might augment phenotyping approaches. One of the first applications of biomarkers for understanding OSA-related disease susceptibility was a study by Sánchez-de-la-Torre and colleagues (25), who showed that the presence of certain microRNA sequences predicted blood pressure responses to CPAP in patients with resistant hypertension and OSA. Conceivably, at some point in the future, such information might inform decisions such as initiating CPAP versus start of an antihypertensive medication. Careful measurement of the impact of OSA through use of frequently sampled metabolites during sleep may also help identify individuals at risk for metabolic sequelae of OSA. For example, Chopra and colleagues (26) demonstrated that plasma free fatty acids, glucose, and cortisol in the blood dynamically increased in response to OSA throughout the night. Notably, in post hoc analyses, the AHI did not distinguish a group of individuals who experienced large rises in free fatty acid and glucose in the blood during sleep (and presumably at high risk for cardiometabolic disease) from the group without a large increase, a finding that challenges the reliance on the AHI for OSA risk stratification. Yet, other circulating molecules might be used to predict other sequelae of OSA, such as liver fibrosis (27). Overall, these emerging data suggest that future precision sleep medicine approaches may incorporate information from biomarkers rather than from the AHI (or other PSG-derived markers) to target individuals most likely to benefit from OSA treatment.
A final caveat is that most methods for phenotyping to date have been based on clinical observations or on clustering using a supervised approach with modest sample sizes, in which groups are determined a priori. However, newer unsupervised “machine learning” approaches may yield differences between groups that are not readily apparent to humans (19). These analyses will benefit from access to large, multidimensional, and well-annotated data resources that contain the physiological, imaging, multi-omic, and clinical data needed to identify and validate unique, comprehensive disease signatures. It is critical that emerging informatics platforms follow FAIR (findability, accessibility, interoperability, and reusability) principles (28). Using FAIR principles, the NIH is developing virtual platforms for sharing research data and tools, including Data Commons and Data STAGE (storage, toolspace, access, and analytics for big-data empowerment). These resources will include data on almost 160,000 individuals from NIH’s TOPMed (Trans-Omics for Precision Medicine) program, including whole-genome sequences and a variety of phenotypic data, with a subgroup containing PSG data. The NIH has also supported the development of the National Sleep Research Resource (www.sleepdata.org), a data repository following FAIR principles that contains over 30,000 PSG records and various phenotypic data from well-characterized cohorts (29), providing 1) annotated datasets, 2) user interfaces for accessing data, and 3) computational tools. Approximately 1 TB of data per week is made available to registered users, who include a growing number of internationally based investigators who propose application of machine learning to the data in this repository. To allow further analysis of physiological and clinical data that are integrated with molecular data, NIH is considering leveraging the combined resources of the National Sleep Research Resource and STAGE, potentially providing unparalleled resources for comprehensively characterizing OSA subtypes across the population. Thus, a major challenge in the field is to make such different phenotypes clinically useful. Ultimately, the future of phenotyping will require reliable biomarkers that predict important patient-centric outcomes.
An endotype is defined as a subtype of a condition that has a distinct functional or pathobiological mechanism. In the case of OSA, several physiological endotypes have been identified as playing a likely causal role. To develop OSA, all patients must have some degree of anatomical compromise (i.e., making the airway prone to collapse). What contributes to this compromise is likely a heterogeneous combination of many factors (e.g., obesity, craniofacial structure, lung volumes, fluid shifts, nasal resistance, or upper airway surface tension). In addition to this anatomical predisposition, at least three nonanatomical endotypes also play a key causal role in many patients with OSA (30). These nonanatomical factors include the following:
An oversensitive ventilatory control system (i.e., ventilatory control instability or high loop gain): The sensitivity of the ventilatory control system is described by the engineering term “loop gain.” In OSA physiology, loop gain represents the inherent stability of the negative feedback loop that controls breathing and can be quantified by assessing the ratio of any ventilatory response to a ventilatory disturbance.
A low respiratory arousal threshold: The arousal threshold is the level of ventilatory drive at which a patient arouses from sleep. The lower the arousal threshold, the more readily increases in ventilatory drive can cause arousal from sleep, which can destabilize breathing (and sleep).
Poor pharyngeal dilator muscle effectiveness/responsiveness during sleep: Pharyngeal dilator muscle effectiveness or responsiveness is the amount of additional ventilation or muscle activity, respectively, that can be achieved by compensatory activation of the pharyngeal dilator muscles of the upper airway in response to increases in ventilatory drive.
The importance of these nonanatomical traits is predicated by the underlying anatomy. That is, very robust muscle responses are needed to overcome upper airway collapsibility in very obese patients (31). With more modest collapsibility (e.g., critical closing pressure of −2 to 0 cm H2O), the nonanatomical traits become relatively more important, with unfavorable traits predisposing to OSA and favorable ones protecting from OSA. Prior work (30, 32) has shown that in approximately one-third of people with OSA, the nonanatomical traits are important for OSA pathogenesis, and conversely, that such patients might benefit/be treated without the use of CPAP therapy. Furthermore, these anatomical and nonanatomical “endotypes” can be very different for patients categorized as having comparable OSA severity when assessed by the AHI (33). This multifactorial nature of OSA pathogenesis may explain why CPAP-alternative interventions such as oral appliances, upper airway surgery, and agents/pharmacological interventions (i.e., supplemental oxygen or sedatives) have to date shown only a modest improvement in OSA severity when administered to unselected patients. Recent detailed physiological studies in relatively small numbers of patients have highlighted that knowledge of a patient’s endotype is crucial for understanding which patients are most likely to show OSA resolution with non-CPAP interventions. (A detailed summary of the available literature on this topic is presented in Table 2.) Key highlights of this collective body of work show the following:
Those with poor muscle compensation at baseline experienced greater benefit from a drug that stimulated the upper airway muscles (40).
|Endotype Targeted||Treatment Intervention||Study||N||Duration of Intervention||Responder Definition Used||Endotypes Measured||Key Findings Predicting Response|
|Anatomical trait||Weight loss||Schwartz et al. (82)||13||∼17 mo||Correlation with ΔAHI*||C||Those with milder collapsibility at baseline experienced greater reductions in their AHI with weight loss|
|Oral appliance||Ng et al. (83)||10||Single night||AHItreatment <5/h||C||Collapsibility alone was not a predictor of response|
|Chan et al. (84)||69||Single time point after 6- to 8-wk acclimatization period||50% ↓ AHI||Anat||No difference at baseline between responders and nonresponders in the volumes of the airway and soft tissue structures, skeletal class, or cephalometric measurements|
|Sutherland et al. (85)||18||Singe time point||50% ↓ AHI||Anat||No differences in upper airway structure between groups|
|Edwards et al. (34)||14||Single night†||50% ↓ AHI + AHItreatment <10/h||C, A, L, M||The combination of loop gain and collapsibility predicted response to therapy with 100% sensitivity and 87.5% specificity|
|UA surgery||Schwartz et al. (86)||13||>2 mo||50% ↓ NREM AHI||C||Collapsibility alone was not a predictor of response|
|Joosten et al. (35)||46||∼3 mo||50% ↓ AHI + AHItreatment <10/h||L (n)||A low loop gain at baseline predicted surgical response|
|Li et al. (36)||31||∼4 mo||Predicted post-treatment AHI||L (n)||A low loop gain at baseline predicted surgical response|
|Hypoglossal nerve stimulation||Schwab et al. (87)||13||12 mo||50% ↓ AHI + AHItreatment <20/h||Anat||Smaller soft palate volumes at baseline and greater tongue movement anteriorly predicted response|
|Vanderveken et al. (88)||21||6 mo||50% ↓ AHI + AHItreatment <20/h||Anat||The absence of complete concentric collapse at the level of the palate was a predictor of response|
|Loop gain||Carbon dioxide||Xie et al. (39)||26||Single night||30% ↓ AHI||C, L||No difference in collapsibility between groups; responders had a higher loop gain (driven by ↑ controller gain)|
|Hyperoxia||Wellman et al. (37)||12||Single night||N/A||L||Patients with higher loop gain showed greater reduction in AHI (46%↓ vs. 16%↓)|
|Xie et al. (39)||26||Single night||30% ↓ AHI||C, L||No difference in collapsibility or loop gain between groups|
|Wang et al. (89)||20||2 mo||Correlation with ΔAHI||L||Controller gain (awake) was lower in those who experienced greater reductions in the AHI|
|Sands et al. (38)||36||Single night||50% ↓ AHI||C, A, L, M (n)||The combination of elevated loop gain, less severe collapsibility, and greater muscle compensation at baseline predicted response|
|Muscle response||Desipramine||Taranto-Montemurro et al. (40)||12||Single night||20/h ↓ AHI||C, A, L, M||Those with poor muscle compensation at baseline experienced greater reduction in AHI while receiving therapy|
|Arousal threshold||Triazolam||Berry et al. (90)||12||Single night||N/A||A||N/A|
|Eszopiclone||Eckert et al. (91)||17||Single night||Compared ΔAHI in patients with low vs. high arousal threshold||A||Patients with a low arousal threshold showed greater reduction in AHI (43%↓)|
|Trazodone||Eckert et al. (92)||7||Single night||Correlation with ΔAHI*||A||Arousal threshold alone did not predict treatment response|
|Trazodone||Smales et al. (93)||13||Single night||Upper 50th percentile of subjects based on the %ΔAHI between groups||A||Arousal threshold alone did not predict treatment response|
|Zopiclone||Carter et al. (94)||12||Single night||Correlation with ΔAHI*||A||Arousal threshold alone did not predict treatment response|
Although these findings show major promise, two key limitations of these approaches are that they 1) require difficult and time-consuming protocols to define the endotypes that are not clinically practical or available and 2) have typically assessed only acute intervention effects after one night of therapy. To successfully translate personalized OSA treatments to mainstream clinical practice, the field requires simple, clinically deployable endotyping methods that reliably identify patients most likely to accept, use, and benefit from nonstandard treatments.
Most of the evidence demonstrating that the OSA endotypes predict response to non-CPAP interventions comes from specialized physiology laboratories that have the sophisticated equipment and highly trained personnel required to measure these factors. Several groups have also used imaging techniques (magnetic resonance imaging, computed tomography, and/or drug-induced sleep endoscopy) to assess the degree of anatomical compromise and/or site of collapse. The current “gold standard” methods to quantify all four endotypes simultaneously require complex manipulations of CPAP and/or require patients to sleep while heavily instrumented (e.g., pressure catheters in the airway/esophagus, and EMG wires into key pharyngeal muscles) (30, 41, 42). Not surprisingly, the clinical applicability of these methodologies is limited by their highly specialized and relatively invasive nature. For OSA endotyping to be clinically useful, it will require techniques that allow endotypes to be determined noninvasively. Ideally, these indices 1) could be derived using currently collected clinical data (i.e., better use of the wealth of physiological information contained within an in-laboratory/home diagnostic PSG or CPAP titration) and 2) are amenable to automation.
As such, important progress has been made to noninvasively quantify several of the pathophysiological endotypes either in isolation or together (see Table 3). Such techniques vary by whether they can use the information contained within a routine clinical PSG or will likely require an additional test, as well as whether they need to be performed during wakefulness or sleep. Perhaps more important, recent evidence suggests that these surrogate measures of the OSA endotypes help predict responses to both CPAP and non-CPAP therapies. For instance, Landry and colleagues (43) demonstrated that a patient’s therapeutic CPAP level requirement (as a measure of an individual’s degree of anatomical compromise ) predicts whether they are likely to respond to the combination of oxygen and sedative just as well as critical closing pressure, which is usually only measured in the research setting. Furthermore, using the method proposed by Edwards and colleagues (45), Zinchuk and colleagues (46) demonstrated that the presence of a low arousal threshold predicts poor long-term (>5 yr) CPAP adherence, further highlighting the importance of understanding a patient’s physiological endotype, not only for treatment but also more broadly for the management of OSA. Such an endotype may suggest populations at risk for low CPAP adherence or may identify potential targets to improve CPAP adherence via efforts to improve the arousal threshold (47, 48). Although in these instances, measuring a single trait or endotype via these simplified methods has been shown to predict adherence, in many other settings (i.e., predicting the response to oral appliances), the available evidence suggests that knowledge of multiple traits is required to inform treatment. For example, a low arousal endotype might be important for adherence, but a low loop gain and mild collapsibility endotype might predict success of non-CPAP therapies. Perhaps most excitingly in this regard, techniques now exist to simultaneously quantify the four key endotypes using data collected from a routine diagnostic clinical PSG (i.e., EEG monitoring to assess sleep/wake state and cortical arousals, as well as nasal pressure to provide a semiquantitative measure of airflow), and they correlate reasonably well with gold standard measures (49–51).
|Endotype||Clinical Variable||Study||Performed in Wake/Sleep?||Requires an Additional Test?|
|Anatomical trait/site of collapse||Therapeutic CPAP level||Landry et al. (44)||Sleep||Yes/no|
|Negative expiratory pressure||Hirata et al. (95)||Wake||Yes|
|Flow or nasal pressure signal from a clinical PSG||Sands et al. (50), Azarbarzin et al. (96, 97), and Genta et al. (98)||Sleep||No|
|UA visualization (e.g., Mallampati/Friedman/MRI-derived images)||Islam et al. (99), Li et al. (100), Smith et al. (101), Chi et al. (102), and Schwab et al. (103)||Wake||Yes/no|
|Craniofacial characteristics||Sutherland et al. (104) and Schwab et al. (105)||Wake||Yes|
|Loop gain||Breath-hold duration||Trembach et al. (106) and Messineo et al. (107)||Wake||Yes (but could be done during consult)|
|Chemoreflex test||Wang et al. (89)||Wake||Yes|
|Nasal pressure signal from a clinical PSG||Terrill et al. (49)||Sleep||No|
|Muscle response||Nasal pressure signal from a clinical PSG||Sands et al. (50)||Sleep||No|
|Arousal threshold||3 PSG characteristics (AHI < 30 h−1, nadir SpO2 > 82.5%, and % hypopneas > 58.3%)||Edwards et al. (45)||Sleep||No|
|Nasal pressure signal from a clinical PSG||Sands et al. (51)||Sleep||No|
To date, there is very little concrete evidence in OSA showing that the endotype predicts the phenotype, partly driven by the heterogeneity in the phenotype and the difficulty of measurement of the endotypes. However, one example that illustrates the link between these two concepts is plausible: the association of OSA endotypes and phenotypes across a range of ages. OSA is particularly prevalent in the elderly, and treatment is likely to be important. However, the symptoms of OSA experienced by younger and older adults are often different (52). For example, Kobayashi and colleagues showed that although OSA severity as assessed by AHI or oxygen desaturation was similar between younger and older patients with OSA, those diagnosed only later in life were less sleepy objectively or subjectively than those who were diagnosed at a younger age (53). Also, epidemiological data have clearly shown that there are fewer OSA-related cardiovascular complications in the elderly than in younger people for the same degree of OSA severity (54). Thus, the presenting symptoms and the rationale for treatment may be different in younger versus older patients with OSA. Interestingly, previous work has also shown that older (age > 60 yr) compared with younger (ages 20–40 yr) people with OSA have different underlying traits/endotypes that predispose them to OSA (55). The endotype of OSA in older populations has a more collapsible airway than in matched (sex, body mass index, with equivalent AHI) younger individuals with OSA, whereas more ventilatory instability (elevated loop gain) was seen in younger than in older individuals with OSA. The different endotypes may have physiological consequences. Kobayashi and colleagues observed that for a given severity of OSA, intrathoracic pressures were markedly less negative in older than in younger patients with OSA, implying differing underlying pathophysiology (53). Because negative intrathoracic pressure is a known component of cardiac wall stress and left ventricular afterload, we have speculated that the endotype underlying OSA in the elderly leads to minimal cardiac wall stress and may result in fewer cardiovascular consequences in the elderly (phenotype) than observed in middle-aged individuals (56). Similarly, people with OSA who wake up too easily (low arousal threshold) less frequently have hypertension than those with a normal arousal threshold (57). This relatively novel concept requires further study, including careful consideration of potential confounding by sex, age, and other factors, but at least in theory, the highly variable clinical manifestations of OSA (e.g., sleepy vs. nonsleepy, hypertensive vs. normotensive, and high vs. low cardiovascular risk) may reflect the known variability in OSA pathogenesis (30).
It is worth noting that the relationship between the endotype and phenotype may also be a bidirectional one; that is, the phenotype may actually drive the endotype. For instance, chronic hypoxemia and/or high left atrial pressure in heart failure may elevate loop gain via increases in chemosensitivity (i.e., controller gain, a determinant of overall loop gain). Furthermore, the fluid retention and pulmonary edema experienced as a result of heart failure may worsen upper airway collapse (58). Such a bidirectional pathway suggests that some of the measured physiological traits may also be a consequence of disease. Some (but not all ) studies show that controller gain decreases after 1 month of CPAP therapy (60, 61) and that the arousal threshold (60, 62), too, appears to change. Nonetheless, measurement of the traits may be useful for characterizing unique subgroups of patients even if they do not directly provide information on etiological mechanisms for OSA. Given the limited evidence on this topic, more research is clearly needed to understand the complex relationship between endotypes and phenotypes.
Finally, as described above, important work using PSG data to better risk stratify patients with OSA in terms of cardiovascular mortality is also likely to reflect underlying endotype. For example, the depth of hypoxemia and duration of events likely reflect arousal threshold and ventilatory control traits, among other factors.
Although simpler techniques to measure physiological endotypes and phenotypes that are scalable to clinical arenas and large population cohort studies are promising, considerable research is needed to prove the clinical utility of these approaches. This includes generation of high-level evidence that identifies the ability of measurements to reliably characterize the OSA mechanistic traits and unique clinical features and to predict responses to treatment interventions. Consistent with principles of rigor and reproducibility, there is a need to ensure that methods are validated and clearly documented and that they provide consistent results across settings and subjects. Finally, much of the work to date has focused on defining a “good” response to treatment interventions based on the AHI alone. However, there are many factors that likely contribute to a successful treatment outcome (e.g., whether a patient will adhere to the treatment) that will require consideration. A deeper understanding of how the OSA endotypes drive the clinical phenotype of OSA is needed to better guide the interpretation and use of diagnostic and prognostic markers. Integrative approaches are needed to also comprehensively link clinical phenotypes to physiology and underlying molecular and genetic mechanisms.
The genotype characterizes the genetic makeup of a cell and therefore of an organism or individual, which contributes to its characteristics (i.e., endotype and ultimately phenotype). Given numerous endotypes and phenotypes, OSA is best conceptualized as a chronic, complex disease that likely reflects the influence of multiple genetic factors. Research over the past decade has established that OSA has a strong genetic basis, given the strong familial aggregation of AHI (63–67), with heritability estimates (the proportion of the variance in the trait estimated to be attributable to genetic factors) ranging from about 0.20 to 0.35 (68). However, stronger genetic associations are likely for traits reflective of specific endotypes. For example, recent analyses have shown higher heritability indices (0.40–0.62) with respiratory event duration and a multicomponent score comprised of six traits (AHI, event duration, two oxygen desaturation indices, sleepiness, and snoring), suggesting that these traits may better characterize genetically determined phenotypes than the AHI alone (69). The significant heritability for event duration is of particular interest, given that this trait correlates with an individual’s ventilatory control sensitivity (55) and also varies across ancestral groups and in men versus women. Although genome-wide association studies have been limited by the very large sample sizes needed for discovery and replication of genetic signals, recent work identified significant associations between several OSA phenotypes and genetic variants. Cade and colleagues (70) have taken the first big step forward in a study of three OSA phenotypes (AHI, mean sleep oxygen saturation as measured by pulse oximetry, and the average respiratory event length) in over 12,000 individuals. Two novel loci were significantly associated with the AHI and respiratory event length. Furthermore, these loci overlapped with genes that had biologically plausible functions, which could contribute to the observed phenotype. Another large study identified that variants in RAI1 were associated with the non-REM AHI level (71). Not only was this finding state specific, suggesting that different genetic mechanisms influence REM and non-REM OSA, but the association also was specific to men, consistent with other studies of this gene showing sexual dimorphisms. Other genome-wide association studies of traits characterizing overnight oxygen saturation levels showed significant associations with a number of genetic variants, including in genes influencing vascular integrity (ANGPT2 [angiopoietin 2] ), iron metabolism (FECH [ferrochelatase] ), and the NLRP3 (NACHT, LRR, and PYD domains–containing protein 3) inflammasome (74). Although this work highlights the value of genetic studies analyzing phenotypes more specific than the overall AHI, a key next step will be to identify the genes linking phenotypes to validated endotypes, which now may be more accessible with the suite of tools (described above and in Table 3) that are being developed to measure these endotypes.
Although we provide an overview of the potential for precision medicine to transform the treatment of OSA, a chronic disease with multiple risk factors, it is possible that future research may uncover one or a few fundamental mechanisms for OSA that can be targeted using one therapy or a combination of therapies. Ultimately, we believe that the increasing interest in the OSA genotypes, endotypes, and phenotypes and the links between them, as well as their relationships to treatment (hypothesized relationships depicted in Figure 1, bottom panel), will revitalize the field of sleep medicine in both research and clinical practice. The transition to a precision-based approach for treatment of OSA will require the following:
Generation of validated and reproducible measurements of endotypes that can be scaled to clinical and large research settings;
Linking these endotypes to genome-wide data to enable discovery of molecular mechanisms that influence these traits, thus identifying new intervention targets and elucidating biological pathways that interconnect OSA to other disease processes; and
Linking both endotype and genetic data to outcome and treatment data to identify individuals most likely to be susceptible (or resilient) to OSA-related stresses and also to identify those most likely to benefit from specific interventions.
|1.||Jordan AS, McSharry DG, Malhotra A. Adult obstructive sleep apnoea. Lancet 2014;383:736–747.|
|2.||Peppard PE, Young T, Barnet JH, Palta M, Hagen EW, Hla KM. Increased prevalence of sleep-disordered breathing in adults. Am J Epidemiol 2013;177:1006–1014.|
|3.||Mehra R, Stone KL, Blackwell T, Ancoli Israel S, Dam TT, Stefanick ML, et al.; Osteoporotic Fractures in Men Study. Prevalence and correlates of sleep-disordered breathing in older men: Osteoporotic Fractures in Men Sleep Study. J Am Geriatr Soc 2007;55:1356–1364.|
|4.||Gottlieb DJ, Punjabi NM, Mehra R, Patel SR, Quan SF, Babineau DC, et al. CPAP versus oxygen in obstructive sleep apnea. N Engl J Med 2014;370:2276–2285.|
|5.||Kushida CA, Nichols DA, Holmes TH, Quan SF, Walsh JK, Gottlieb DJ, et al. Effects of continuous positive airway pressure on neurocognitive function in obstructive sleep apnea patients: the Apnea Positive Pressure Long-term Efficacy Study (APPLES). Sleep (Basel) 2012;35:1593–1602.|
|6.||Loffler KA, Heeley E, Freed R, Anderson CS, Brockway B, Corbett A, et al.; SAVE (Sleep Apnea Cardiovascular Endpoints) Investigators. Effect of obstructive sleep apnea treatment on renal function in patients with cardiovascular disease. Am J Respir Crit Care Med 2017;196:1456–1462.|
|7.||McEvoy RD, Antic NA, Heeley E, Luo Y, Ou Q, Zhang X, et al.; SAVE Investigators and Coordinators. CPAP for prevention of cardiovascular events in obstructive sleep apnea. N Engl J Med 2016;375:919–931.|
|8.||Quan SF, Chan CS, Dement WC, Gevins A, Goodwin JL, Gottlieb DJ, et al. The association between obstructive sleep apnea and neurocognitive performance: the Apnea Positive Pressure Long-term Efficacy Study (APPLES). Sleep (Basel) 2011;34:303–314B.|
|9.||Moore WC, Meyers DA, Wenzel SE, Teague WG, Li H, Li X, et al.; National Heart, Lung, and Blood Institute’s Severe Asthma Research Program. Identification of asthma phenotypes using cluster analysis in the Severe Asthma Research Program. Am J Respir Crit Care Med 2010;181:315–323.|
|10.||Calfee CS, Delucchi K, Parsons PE, Thompson BT, Ware LB, Matthay MA; NHLBI ARDS Network. Subphenotypes in acute respiratory distress syndrome: latent class analysis of data from two randomised controlled trials. Lancet Respir Med 2014;2:611–620.|
|11.||Bibbins-Domingo K, Grossman DC, Curry SJ, Davidson KW, Epling JW Jr, García FA, et al.; US Preventive Services Task Force. Screening for obstructive sleep apnea in adults: US Preventive Services Task Force recommendation statement. JAMA 2017;317:407–414.|
|12.||Redline S. Screening for obstructive sleep apnea: implications for the sleep health of the population. JAMA 2017;317:368–370.|
|13.||Quan SF. Graduate medical education in sleep medicine: did the canary just die? J Clin Sleep Med 2013;9:101.|
|14.||Zinchuk AV, Gentry MJ, Concato J, Yaggi HK. Phenotypes in obstructive sleep apnea: a definition, examples and evolution of approaches. Sleep Med Rev 2017;35:113–123.|
|15.||Ye L, Pien GW, Ratcliffe SJ, Björnsdottir E, Arnardottir ES, Pack AI, et al. The different clinical faces of obstructive sleep apnoea: a cluster analysis. Eur Respir J 2014;44:1600–1607.|
|16.||Keenan BT, Kim J, Singh B, Bittencourt L, Chen NH, Cistulli PA, et al. Recognizable clinical subtypes of obstructive sleep apnea across international sleep centers: a cluster analysis. Sleep (Basel) 2018;41:zsx214.|
|17.||Pien GW, Ye L, Keenan BT, Maislin G, Björnsdóttir E, Arnardottir ES, et al. Changing faces of obstructive sleep apnea: treatment effects by cluster designation in the Icelandic Sleep Apnea Cohort. Sleep (Basel) 2018;41:zsx201.|
|18.||Butler MP, Emch JT, Rueschman M, Sands SA, Shea SA, Wellman A, et al. Apnea-hypopnea event duration predicts mortality in men and women in the Sleep Heart Health Study. Am J Respir Crit Care Med 2019;199:903–912.|
|19.||Zinchuk AV, Jeon S, Koo BB, Yan X, Bravata DM, Qin L, et al. Polysomnographic phenotypes and their cardiovascular implications in obstructive sleep apnoea. Thorax 2018;73:472–480.|
|20.||Azarbarzin A, Sands SA, Stone KL, Taranto-Montemurro L, Messineo L, Terrill PI, et al. The hypoxic burden of sleep apnoea predicts cardiovascular disease-related mortality: the Osteoporotic Fractures in Men Study and the Sleep Heart Health Study. Eur Heart J 2019;40:1149–1157.|
|21.||Castro-Grattoni AL, Torres G, Martínez-Alonso M, Barbé F, Turino C, Sánchez-de-la-Torre A, et al. Blood pressure response to CPAP treatment in subjects with obstructive sleep apnoea: the predictive value of 24-h ambulatory blood pressure monitoring. Eur Respir J 2017;50:1700651.|
|22.||Montesi SB, Edwards BA, Malhotra A, Bakker JP. The effect of continuous positive airway pressure treatment on blood pressure: a systematic review and meta-analysis of randomized controlled trials. J Clin Sleep Med 2012;8:587–596.|
|23.||Pépin JL, Woehrle H, Liu D, Shao S, Armitstead JP, Cistulli PA, et al. Adherence to positive airway therapy after switching from CPAP to ASV: a big data analysis. J Clin Sleep Med 2018;14:57–63.|
|24.||Stanchina M, Robinson K, Corrao W, Donat W, Sands S, Malhotra A. Clinical use of loop gain measures to determine continuous positive airway pressure efficacy in patients with complex sleep apnea: a Pilot Study. Ann Am Thorac Soc 2015;12:1351–1357.|
|25.||Sánchez-de-la-Torre M, Khalyfa A, Sánchez-de-la-Torre A, Martinez-Alonso M, Martinez-García MA, Barceló A, et al.; Spanish Sleep Network. Precision medicine in patients with resistant hypertension and obstructive sleep apnea: blood pressure response to continuous positive airway pressure treatment. J Am Coll Cardiol 2015;66:1023–1032.|
|26.||Chopra S, Rathore A, Younas H, Pham LV, Gu C, Beselman A, et al. Obstructive sleep apnea dynamically increases nocturnal plasma free fatty acids, glucose, and cortisol during sleep. J Clin Endocrinol Metab 2017;102:3172–3181.|
|27.||Mesarwi OA, Shin MK, Drager LF, Bevans-Fonti S, Jun JC, Putcha N, et al. Lysyl oxidase as a serum biomarker of liver fibrosis in patients with severe obesity and obstructive sleep apnea. Sleep (Basel) 2015;38:1583–1591.|
|28.||Wilkinson MD, Dumontier M, Aalbersberg IJ, Appleton G, Axton M, Baak A, et al. The FAIR Guiding Principles for scientific data management and stewardship. Sci Data 2016;3:160018.|
|29.||Zhang GQ, Cui L, Mueller R, Tao S, Kim M, Rueschman M, et al. The National Sleep Research Resource: towards a sleep data commons. J Am Med Inform Assoc 2018;25:1351–1358.|
|30.||Eckert DJ, White DP, Jordan AS, Malhotra A, Wellman A. Defining phenotypic causes of obstructive sleep apnea: identification of novel therapeutic targets. Am J Respir Crit Care Med 2013;188:996–1004.|
|31.||Sands SA, Eckert DJ, Jordan AS, Edwards BA, Owens RL, Butler JP, et al. Enhanced upper-airway muscle responsiveness is a distinct feature of overweight/obese individuals without sleep apnea. Am J Respir Crit Care Med 2014;190:930–937.|
|32.||Owens RL, Edwards BA, Eckert DJ, Jordan AS, Sands SA, Malhotra A, et al. An integrative model of physiological traits can be used to predict obstructive sleep apnea and response to non positive airway pressure therapy. Sleep (Basel) 2015;38:961–970.|
|33.||Azarbarzin A, Sands SA, Taranto-Montemurro L, Oliveira Marques MD, Genta PR, Edwards BA, et al. Estimation of pharyngeal collapsibility during sleep by peak inspiratory airflow. Sleep (Basel) 2017;40:zsw005.|
|34.||Edwards BA, Andara C, Landry S, Sands SA, Joosten SA, Owens RL, et al. Upper-airway collapsibility and loop gain predict the response to oral appliance therapy in patients with obstructive sleep apnea. Am J Respir Crit Care Med 2016;194:1413–1422.|
|35.||Joosten SA, Leong P, Landry SA, Sands SA, Terrill PI, Mann D, et al. Loop gain predicts the response to upper airway surgery in patients with obstructive sleep apnea. Sleep (Basel) 2017;40:zsx094.|
|36.||Li Y, Ye J, Han D, Cao X, Ding X, Zhang Y, et al. Physiology-based modeling may predict surgical treatment outcome for obstructive sleep apnea. J Clin Sleep Med 2017;13:1029–1037.|
|37.||Wellman A, Malhotra A, Jordan AS, Stevenson KE, Gautam S, White DP. Effect of oxygen in obstructive sleep apnea: role of loop gain. Respir Physiol Neurobiol 2008;162:144–151.|
|38.||Sands SA, Edwards BA, Terrill PI, Butler JP, Owens RL, Taranto-Montemurro L, et al. Identifying obstructive sleep apnoea patients responsive to supplemental oxygen therapy. Eur Respir J 2018;52:1800674.|
|39.||Xie A, Teodorescu M, Pegelow DF, Teodorescu MC, Gong Y, Fedie JE, et al. Effects of stabilizing or increasing respiratory motor outputs on obstructive sleep apnea. J Appl Physiol (1985) 2013;115:22–33.|
|40.||Taranto-Montemurro L, Sands SA, Edwards BA, Azarbarzin A, Marques M, de Melo C, et al. Desipramine improves upper airway collapsibility and reduces OSA severity in patients with minimal muscle compensation. Eur Respir J 2016;48:1340–1350.|
|41.||Wellman A, Eckert DJ, Jordan AS, Edwards BA, Passaglia CL, Jackson AC, et al. A method for measuring and modeling the physiological traits causing obstructive sleep apnea. J Appl Physiol (1985) 2011;110:1627–1637.|
|42.||Wellman A, Edwards BA, Sands SA, Owens RL, Nemati S, Butler J, et al. A simplified method for determining phenotypic traits in patients with obstructive sleep apnea. J Appl Physiol (1985) 2013;114:911–922.|
|43.||Landry SA, Joosten SA, Sands SA, White DP, Malhotra A, Wellman A, et al. Response to a combination of oxygen and a hypnotic as treatment for obstructive sleep apnoea is predicted by a patient's therapeutic CPAP requirement. Respirology 2017;22:1219–1224.|
|44.||Landry SA, Joosten SA, Eckert DJ, Jordan AS, Sands SA, White DP, et al. Therapeutic CPAP level predicts upper airway collapsibility in patients with obstructive sleep apnea. Sleep (Basel) 2017;40:zsx056.|
|45.||Edwards BA, Eckert DJ, McSharry DG, Sands SA, Desai A, Kehlmann G, et al. Clinical predictors of the respiratory arousal threshold in patients with obstructive sleep apnea. Am J Respir Crit Care Med 2014;190:1293–1300.|
|46.||Zinchuk A, Edwards BA, Jeon S, Koo BB, Concato J, Sands S, et al. Prevalence, associated clinical features, and impact on continuous positive airway pressure use of a low respiratory arousal threshold among male United States veterans with obstructive sleep apnea. J Clin Sleep Med 2018;14:809–817.|
|47.||Lettieri CJ, Collen JF, Eliasson AH, Quast TM. Sedative use during continuous positive airway pressure titration improves subsequent compliance: a randomized, double-blind, placebo-controlled trial. Chest 2009;136:1263–1268.|
|48.||Lettieri CJ, Shah AA, Holley AB, Kelly WF, Chang AS, Roop SA; CPAP Promotion and Prognosis—The Army Sleep Apnea Program Trial. Effects of a short course of eszopiclone on continuous positive airway pressure adherence: a randomized trial. Ann Intern Med 2009;151:696–702.|
|49.||Terrill PI, Edwards BA, Nemati S, Butler JP, Owens RL, Eckert DJ, et al. Quantifying the ventilatory control contribution to sleep apnoea using polysomnography. Eur Respir J 2015;45:408–418.|
|50.||Sands SA, Edwards BA, Terrill PI, Taranto-Montemurro L, Azarbarzin A, Marques M, et al. Phenotyping pharyngeal pathophysiology using polysomnography in patients with obstructive sleep apnea. Am J Respir Crit Care Med 2018;197:1187–1197.|
|51.||Sands SA, Terrill PI, Edwards BA, Taranto Montemurro L, Azarbarzin A, Marques M, et al. Quantifying the arousal threshold using polysomnography in obstructive sleep apnea. Sleep (Basel) 2018;41:zsx183.|
|52.||Gooneratne NS, Vitiello MV. Sleep in older adults: normative changes, sleep disorders, and treatment options. Clin Geriatr Med 2014;30:591–627.|
|53.||Kobayashi M, Namba K, Tsuiki S, Matsuo A, Sugiura T, Inoue Y. Clinical characteristics in two subgroups of obstructive sleep apnea syndrome in the elderly: comparison between cases with elderly and middle-age onset. Chest 2010;137:1310–1315.|
|54.||Punjabi NM, Caffo BS, Goodwin JL, Gottlieb DJ, Newman AB, O’Connor GT, et al. Sleep-disordered breathing and mortality: a prospective cohort study. PLoS Med 2009;6:e1000132.|
|55.||Edwards BA, Wellman A, Sands SA, Owens RL, Eckert DJ, White DP, et al. Obstructive sleep apnea in older adults is a distinctly different physiological phenotype. Sleep (Basel) 2014;37:1227–1236.|
|56.||Edwards BA, O’Driscoll DM, Ali A, Jordan AS, Trinder J, Malhotra A. Aging and sleep: physiology and pathophysiology. Semin Respir Crit Care Med 2010;31:618–633.|
|57.||Zinchuk A, Edwards BA, Jeon S, Koo BB, Concato J, Sands S, et al. Prevalence, associated clinical features, and impact on continuous positive airway pressure use of a low respiratory arousal threshold among male United States veterans with obstructive sleep apnea. J Clin Sleep Med 2018;14:809–817.|
|58.||Yumino D, Redolfi S, Ruttanaumpawan P, Su MC, Smith S, Newton GE, et al. Nocturnal rostral fluid shift: a unifying concept for the pathogenesis of obstructive and central sleep apnea in men with heart failure. Circulation 2010;121:1598–1605.|
|59.||Deacon-Diaz NL, Sands SA, McEvoy RD, Catcheside PG. Daytime loop gain is elevated in obstructive sleep apnea but not reduced by CPAP treatment. J Appl Physiol (1985) 2018;125:1490–1497.|
|60.||Loewen A, Ostrowski M, Laprairie J, Atkar R, Gnitecki J, Hanly P, et al. Determinants of ventilatory instability in obstructive sleep apnea: inherent or acquired? Sleep 2009;32:1355–1365.|
|61.||Salloum A, Rowley JA, Mateika JH, Chowdhuri S, Omran Q, Badr MS. Increased propensity for central apnea in patients with obstructive sleep apnea: effect of nasal continuous positive airway pressure. Am J Respir Crit Care Med 2010;181:189–193.|
|62.||Haba-Rubio J, Sforza E, Weiss T, Schröder C, Krieger J. Effect of CPAP treatment on inspiratory arousal threshold during NREM sleep in OSAS. Sleep Breath 2005;9:12–19.|
|63.||Kaprio J, Koskenvuo M, Partinen M, Telakivi I. A twin study of snoring [abstract]. Sleep Res 1988;17:365.|
|64.||Jennum P, Hein HO, Suadicani P, Sørensen H, Gyntelberg F. Snoring, family history, and genetic markers in men: the Copenhagen Male Study. Chest 1995;107:1289–1293.|
|65.||Mathur R, Douglas NJ. Family studies in patients with the sleep apnea-hypopnea syndrome. Ann Intern Med 1995;122:174–178.|
|66.||Pillar G, Lavie P. Assessment of the role of inheritance in sleep apnea syndrome. Am J Respir Crit Care Med 1995;151:688–691.|
|67.||Redline S, Tishler PV, Tosteson TD, Williamson J, Kump K, Browner I, et al. The familial aggregation of obstructive sleep apnea. Am J Respir Crit Care Med 1995;151:682–687.|
|68.||Buxbaum SG, Elston RC, Tishler PV, Redline S. Genetics of the apnea hypopnea index in Caucasians and African Americans: I. Segregation analysis. Genet Epidemiol 2002;22:243–253.|
|69.||Liang J, Cade BE, Wang H, Chen H, Gleason KJ, Larkin EK, et al. Comparison of heritability estimation and linkage analysis for multiple traits using principal component analyses. Genet Epidemiol 2016;40:222–232.|
|70.||Cade BE, Chen H, Stilp AM, Gleason KJ, Sofer T, Ancoli-Israel S, et al. Genetic associations with obstructive sleep apnea traits in hispanic/latino Americans. Am J Respir Crit Care Med 2016;194:886–897.|
|71.||Chen H, Cade BE, Gleason KJ, Bjonnes AC, Stilp AM, Sofer T, et al. Multiethnic meta-analysis identifies RAI1 as a possible obstructive sleep apnea-related quantitative trait locus in men. Am J Respir Cell Mol Biol 2018;58:391–401.|
|72.||Wang H, Cade BE, Chen H, Gleason KJ, Saxena R, Feng T, et al. Variants in angiopoietin-2 (ANGPT2) contribute to variation in nocturnal oxyhaemoglobin saturation level. Hum Mol Genet 2016;25:5244–5253.|
|73.||Wang H, Cade BE, Sofer T, Sands SA, Chen H, Browning S, et al. Admixture mapping identifies novel loci for obstructive sleep apnea in hispanic/latino americans. Hum Mol Genet 2019;28:675–687.|
|74.||Cade BE, Chen H, Stilp AM, Louie T, Ancoli-Israel S, Arens R, et al. Associations of variants in the hexokinase 1 and interleukin 18 receptor regions with oxyhemoglobin saturation during sleep. PLoS Genet 2019;15:e1007739.|
|75.||Punjabi NM. COUNTERPOINT: is the apnea-hypopnea index the best way to quantify the severity of sleep-disordered breathing? No. Chest 2016;149:16–19.|
|76.||Stamatakis K, Sanders MH, Caffo B, Resnick HE, Gottlieb DJ, Mehra R, et al. Fasting glycemia in sleep disordered breathing: lowering the threshold on oxyhemoglobin desaturation. Sleep 2008;31:1018–1024.|
|77.||Rahangdale S, Yeh SY, Novack V, Stevenson K, Barnard MR, Furman MI, et al. The influence of intermittent hypoxemia on platelet activation in obese patients with obstructive sleep apnea. J Clin Sleep Med 2011;7:172–178.|
|78.||Djonlagic I, Saboisky J, Carusona A, Stickgold R, Malhotra A. Increased sleep fragmentation leads to impaired off-line consolidation of motor memories in humans. PLoS One 2012;7:e34106.|
|79.||Yue HJ, Bardwell W, Ancoli-Israel S, Loredo JS, Dimsdale JE. Arousal frequency is associated with increased fatigue in obstructive sleep apnea. Sleep Breath 2009;13:331–339.|
|80.||Patel SR, White DP, Malhotra A, Stanchina ML, Ayas NT. Continuous positive airway pressure therapy for treating sleepiness in a diverse population with obstructive sleep apnea: results of a meta-analysis. Arch Intern Med 2003;163:565–571.|
|81.||Terán-Santos J, Jiménez-Gómez A, Cordero-Guevara J; Cooperative Group Burgos-Santander. The association between sleep apnea and the risk of traffic accidents. N Engl J Med 1999;340:847–851.|
|82.||Schwartz AR, Gold AR, Schubert N, Stryzak A, Wise RA, Permutt S, et al. Effect of weight loss on upper airway collapsibility in obstructive sleep apnea. Am Rev Respir Dis 1991;144:494–498.|
|83.||Ng AT, Gotsopoulos H, Qian J, Cistulli PA. Effect of oral appliance therapy on upper airway collapsibility in obstructive sleep apnea. Am J Respir Crit Care Med 2003;168:238–241.|
|84.||Chan AS, Sutherland K, Schwab RJ, Zeng B, Petocz P, Lee RW, et al. The effect of mandibular advancement on upper airway structure in obstructive sleep apnoea. Thorax 2010;65:726–732.|
|85.||Sutherland K, Deane SA, Chan AS, Schwab RJ, Ng AT, Darendeliler MA, et al. Comparative effects of two oral appliances on upper airway structure in obstructive sleep apnea. Sleep (Basel) 2011;34:469–477.|
|86.||Schwartz AR, Schubert N, Rothman W, Godley F, Marsh B, Eisele D, et al. Effect of uvulopalatopharyngoplasty on upper airway collapsibility in obstructive sleep apnea. Am Rev Respir Dis 1992;145:527–532.|
|87.||Schwab RJ, Wang SH, Verbraecken J, Vanderveken OM, Van de Heyning P, Vos WG, et al. Anatomic predictors of response and mechanism of action of upper airway stimulation therapy in patients with obstructive sleep apnea. Sleep (Basel) 2018;41:zsy021.|
|88.||Vanderveken OM, Maurer JT, Hohenhorst W, Hamans E, Lin HS, Vroegop AV, et al. Evaluation of drug-induced sleep endoscopy as a patient selection tool for implanted upper airway stimulation for obstructive sleep apnea. J Clin Sleep Med 2013;9:433–438.|
|89.||Wang D, Wong KK, Rowsell L, Don GW, Yee BJ, Grunstein RR. Predicting response to oxygen therapy in obstructive sleep apnoea patients using a 10-minute daytime test. Eur Respir J 2018;51:1701587.|
|90.||Berry RB, Kouchi K, Bower J, Prosise G, Light RW. Triazolam in patients with obstructive sleep apnea. Am J Respir Crit Care Med 1995;151:450–454.|
|91.||Eckert DJ, Owens RL, Kehlmann GB, Wellman A, Rahangdale S, Yim-Yeh S, et al. Eszopiclone increases the respiratory arousal threshold and lowers the apnoea/hypopnoea index in obstructive sleep apnoea patients with a low arousal threshold. Clin Sci (Lond) 2011;120:505–514.|
|92.||Eckert DJ, Malhotra A, Wellman A, White DP. Trazodone increases the respiratory arousal threshold in patients with obstructive sleep apnea and a low arousal threshold. Sleep (Basel) 2014;37:811–819.|
|93.||Smales ET, Edwards BA, Deyoung PN, McSharry DG, Wellman A, Velasquez A, et al. Trazodone effects on obstructive sleep apnea and non-REM arousal threshold. Ann Am Thorac Soc 2015;12:758–764.|
|94.||Carter SG, Berger MS, Carberry JC, Bilston LE, Butler JE, Tong BK, et al. Zopiclone increases the arousal threshold without impairing genioglossus activity in obstructive sleep apnea. Sleep (Basel) 2016;39:757–766.|
|95.||Hirata RP, Schorr F, Kayamori F, Moriya HT, Romano S, Insalaco G, et al. Upper airway collapsibility assessed by negative expiratory pressure while awake is associated with upper airway anatomy. J Clin Sleep Med 2016;12:1339–1346.|
|96.||Azarbarzin A, Marques M, Sands SA, Op de Beeck S, Genta PR, Taranto-Montemurro L, et al. Predicting epiglottic collapse in patients with obstructive sleep apnoea. Eur Respir J 2017;50:1700345.|
|97.||Azarbarzin A, Sands SA, Marques M, Genta PR, Taranto-Montemurro L, Messineo L, et al. Palatal prolapse as a signature of expiratory flow limitation and inspiratory palatal collapse in patients with obstructive sleep apnoea. Eur Respir J 2018;51:1701419.|
|98.||Genta PR, Sands SA, Butler JP, Loring SH, Katz ES, Demko BG, et al. Airflow shape is associated with the pharyngeal structure causing OSA. Chest 2017;152:537–546.|
|99.||Islam S, Selbong U, Taylor CJ, Ormiston IW. Does a patient’s Mallampati score predict outcome after maxillomandibular advancement for obstructive sleep apnoea? Br J Oral Maxillofac Surg 2015;53:23–27.|
|100.||Li HY, Chen NH, Lee LA, Shu YH, Fang TJ, Wang PC. Use of morphological indicators to predict outcomes of palatopharyngeal surgery in patients with obstructive sleep apnea. ORL J Otorhinolaryngol Relat Spec 2004;66:119–123.|
|101.||Smith DF, Benke JR, Yaster S, Boss EF, Ishman SL. A pilot staging system to predict persistent obstructive sleep apnea in children following adenotonsillectomy. Laryngoscope 2013;123:1817–1822.|
|102.||Chi L, Comyn FL, Mitra N, Reilly MP, Wan F, Maislin G, et al. Identification of craniofacial risk factors for obstructive sleep apnoea using three-dimensional MRI. Eur Respir J 2011;38:348–358.|
|103.||Schwab RJ, Pasirstein M, Pierson R, Mackley A, Hachadoorian R, Arens R, et al. Identification of upper airway anatomic risk factors for obstructive sleep apnea with volumetric magnetic resonance imaging. Am J Respir Crit Care Med 2003;168:522–530.|
|104.||Sutherland K, Schwab RJ, Maislin G, Lee RW, Benedikstdsottir B, Pack AI, et al. Facial phenotyping by quantitative photography reflects craniofacial morphology measured on magnetic resonance imaging in Icelandic sleep apnea patients. Sleep (Basel) 2014;37:959–968.|
|105.||Schwab RJ, Leinwand SE, Bearn CB, Maislin G, Rao RB, Nagaraja A, et al. Digital morphometrics: a new upper airway phenotyping paradigm in OSA. Chest 2017;152:330–342.|
|106.||Trembach N, Zabolotskikh I. Breath-holding test in evaluation of peripheral chemoreflex sensitivity in healthy subjects. Respir Physiol Neurobiol 2017;235:79–82.|
|107.||Messineo L, Taranto-Montemurro L, Azarbarzin A, Oliveira Marques MD, Calianese N, White DP, et al. Breath-holding as a means to estimate the loop gain contribution to obstructive sleep apnoea. J Physiol 2018;596:4043–4056.|
Supported by a Heart Foundation of Australia Future Leader Fellowship (101167 [B.A.E.]); NIH grants R35 HL135818 and R01 HL113338 (S.R.); the American Heart Association (15SDG25890059), the American Thoracic Society Foundation, and NIH grant R01 HL102321 (S.A.S.); and NIH grant R01 HL142114 (R.L.O.).
Originally Published in Press as DOI: 10.1164/rccm.201901-0014TR on April 25, 2019