Obstructive sleep apnea (OSA) is characterized by repetitive upper airway collapse resulting in intermittent hypoxemia and hypercapnia, large intrathoracic pressure swings, and cortical arousals. The rate of apneas and hypopneas observed during sleep, the apnea-hypopnea index (AHI), has been used for decades to diagnose OSA and to classify its severity. Despite the wide acceptance of this metric by the sleep medicine community, clinical research has found poor correlations between the AHI- and OSA-related complications or symptoms. We have come to learn that the AHI is an oversimplification of a complex and diverse disease process. (Punjabi. Chest.).
The most important features of a disease metric are reliability, and the ability to predict clinically relevant outcomes. The reliability of the AHI has been in question due to substantial night-to-night variability that can lead to missed diagnosis and disease severity misclassification (Dzierzewski et al. J Clin Sleep Med.
Over the past decade, our understanding of the different pathophysiological mechanisms leading to OSA has grown substantially, suggesting the need for a phenotype-specific treatment approach (Zinchuk, Yaggi. Chest.). The reliance on a single metric that does not capture this heterogeneity may prove detrimental to our therapeutic efforts. One extremely important dimension that is missed by the AHI is the patient. Individual response to airway obstruction varies with age, genetics, gender, and comorbidities, among other things. This may explain the difference in symptoms and outcomes experienced by patients with the same AHI. During the era of precision medicine, the concept of defining a clinical condition by a single test result, without regard to patient characteristics, is antiquated.
Several studies have attempted to propose complementary metrics that may better characterize OSA and predict outcomes. The hypoxic burden has gained a lot of attention as it is generally felt that hypoxemia is a major factor contributing to the pathogenesis of OSA-related comorbidities. Azarbarzin, et al. reported a hypoxic burden metric by measuring the area under the oxygen desaturation curve during a respiratory event (Azarbarzin et al. Eur Heart J.). It factors the length and depth of the desaturations into a single value that expresses the average desaturation burden per hour of sleep time. The hypoxic burden was independently predictive of cardiovascular mortality in two large cohorts. Interestingly, the AHI did not have such an association. Similarly, another novel proposed parameter, the oxygen desaturation rate (ODR), outperformed the AHI in predicting cardiovascular outcomes in severe OSA patients (Wang et al. J Clin Sleep Med. ). The ODR measures the speed of an oxygen desaturation during an apnea event. Subjects with a faster ODR were found to have higher blood pressure values and variability. The authors hypothesized that slower desaturations generate hypoxemia-conditioning that may protect from exaggerated hemodynamic changes. These findings of novel hypoxemia metrics, albeit having their own limitations, recapitulate the need to move beyond the AHI to characterize OSA.
The apnea-hypopnea event duration is another overlooked feature that may impact OSA outcomes. Butler, et al. demonstrated that shorter event duration predicted a higher all-cause mortality over and beyond that predicted by AHI (Butler et al. Am J Respir Crit Care Med.). These results contrast views that early arousals in response to respiratory events may improve outcomes as they reflect a protective mechanism to prevent further hypoxemia and sympatho-excitation. For example, Ma, et al. found that higher percentage of total sleep time spent in apnea/hypopnea (AHT%) predicted worse daytime sleepiness to a higher degree than standard AHI (Ma et al. Sci Rep. ). However, shorter event duration may represent lower arousal thresholds (increased excitability), and ventilatory control instability (higher loop gain), predisposing patients to augmented sympathetic activity. Along similar lines, the intensity of respiratory-related arousals (as measured by EEG wavelet transformation) was found to be independent of preceding respiratory stimulus, with higher arousal intensity levels correlating with higher respiratory and heart rate responses (Amatoury et al. Sleep. ). The contribution of arousals to OSA morbidity is of particular importance for women in whom long-term outcomes of elevated AHI are poorly understood. Bearing in mind the differences in the metrics used, these results underscore the role of event duration and arousability in the pathogenesis of OSA-related morbidity.
The AHI is certainly an important piece of data that is informative and somewhat predictive. However, when used as a sole disease-defining metric, it has yielded disappointing results, especially after OSA treatment trials failed to show cardiovascular benefits despite therapies achieving a low residual AHI. As we aim to achieve a more personalized approach for diagnosing and treating OSA, we need to explore beyond the concept of a single metric to define a heterogenous and complex disorder. Instead of relying on the frequency of respiratory events, it is time to use complementary polysomnographic data that better reflect the origin and systemic effects of these disturbances. Machine-learning methods may offer sophisticated approaches to identifying polysomnographic patterns for future research. Clinical characteristics will also likely need to be considered in OSA severity scales. The identification of symptom subtypes or blood biomarkers may help identify patient groups who may be impacted differently by OSA, and consequently have a different treatment response (Malhotra et al. Sleep.).
Almost half a century has lapsed since the original descriptions of OSA. Since then, our understanding of the disorder has improved greatly, with much still to be discovered, but our method of disease capture is unwavering. Future research requires a focus on novel measures aimed at identifying OSA endophenotypes, which will transform our understanding of disease traits and propel us into personalized therapies.
Dr. Mansour is Assistant Professor of Medicine, Division of Pulmonary, Allergy, and Critical Care Medicine, Duke University School of Medicine, Durham, North Carolina. Dr. Won is Associate Professor of Medicine, Section of Pulmonary, Critical Care, and Sleep Medicine, Yale University School of Medicine; and VA Connecticut Healthcare System, West Haven, Connecticut.