The authors, led by Peter C. Gay of the Mayo Clinic’s Division of Pulmonary, Critical Care, Sleep and Allergy Medicine, together with Luigi Taranto-Montemurro, Lenise J. Kim, and Ali Azarbarzin, point out that AHI correlates poorly with patient-reported symptoms and inconsistently with cardiovascular outcomes. Several limitations of AHI need to be considered. As a frequency metric, it does not account for the contributions of different pathophysiological traits of OSA, including the duration and depth of oxygen desaturation or the degree of sleep fragmentation. Similarly, the traditional single-metric definition of treatment success (AHI below 5 events per hour or a 50 percent or greater AHI reduction) may not fully reflect meaningful improvements in pathophysiology or patient-centered outcomes. Current payer coverage is dependent on AHI alone, which may limit access to treatments in patients with different presentations of OSA.

The Agency for Healthcare Research and Quality published a highly critical report of the sleep health community in 2020, calling upon sleep researchers to move beyond the AHI as a single descriptor of disease and to better justify continuous positive airway pressure (CPAP) treatment as a therapy for clinically meaningful outcomes. Recent efforts, such as the Baveno classification, aim to move beyond AHI in OSA by evaluating both symptoms and comorbidities to better characterize disease burden. The Baveno system grades OSA based on a combination of symptom intensity and the presence of clinically significant comorbidities, such as resistant or uncontrolled hypertension, atrial fibrillation, heart failure, stroke, and diabetes. Patients are assigned to one of four groups (A through D), and these groups appear to differ in hypoxic load and patient-reported outcomes. A modified Baveno classification has more recently reincorporated a high-AHI threshold of 30 events per hour or greater and formal cardiovascular risk estimation to refine treatment indication.

The proposed multidimensional evaluation integrates objective measures of pathophysiologic burden with subjective measures of patient-reported symptomatic burden. On the objective side, the authors propose that AHI should be considered a starting point rather than the single endpoint. Observational studies have demonstrated associations between higher AHI and increased risk of cardiometabolic sequelae, including hypertension, heart failure, and stroke. However, large randomized controlled clinical trials have not consistently shown a reduction in cardiovascular events among patients using CPAP, with positive outcomes only shown in post-hoc analyses including more adherent patients. The relationship between AHI and clinical outcomes is further complicated by the heterogeneity of respiratory events during different sleep stages. Obstructive events occurring predominantly during rapid eye movement (REM) sleep are independently associated with higher incidence of cardiovascular events, insulin resistance, and type-2 diabetes, despite patients with REM-related OSA often having relatively low AHI. Limited nightly CPAP use in the early portion of sleep may reduce measured AHI but still fail to prevent REM-related disease burden and its downstream consequences.

Furthermore, reductions in AHI do not necessarily capture other dimensions of OSA pathophysiology. Residual hypoxemia and persistent symptoms may remain despite improvement in AHI. Conversely, a clinically meaningful therapeutic effect does not necessarily require complete normalization of AHI, as observational data suggest a progressive relationship between OSA severity and cardiovascular risk. The effectiveness of CPAP therapy is also limited by poor adherence, with clinical practice data suggesting that up to 61 percent of patients discontinue CPAP within the first year of treatment. Even among patients who remain on therapy, adherence is often sub-optimal, with many patients using CPAP for 4 hours per night or less. Consequently, improvements in AHI measures during CPAP use may not fully reflect the residual burden of untreated disease through the entire sleep period. Alternative therapies, such as oral appliances, upper-airway surgery, hypoglossal nerve stimulation, and weight reduction provide clinical benefits to selected subgroups of patients, but their effects on long-term cardiovascular risk remain unclear.

As a complementary objective metric, the authors advocate for hypoxic burden (HB), a robust marker of chronic intermittent hypoxia. In contrast to metrics based on time and frequency, such as time spent below 90 percent saturation (T90) and the oxygen desaturation index (ODI), HB integrates the frequency, depth, and duration of desaturations as the area under the desaturation curve. This integrative formulation allows area-based metrics to discriminate among patients with different nocturnal hypoxia profiles. HB has been shown to predict cardiovascular outcomes and all-cause mortality more strongly than AHI, with patients in the highest HB categories showing increased risk of cardiovascular-related mortality. In addition, HB appears to strongly predict symptoms related to OSA, with higher HB independently associated with increased odds of excessive daytime sleepiness defined by an Epworth Sleepiness Scale score above 10. Multiple studies have demonstrated that subjective sleepiness assessed by the Epworth scale poorly correlates with AHI, highlighting the symptom-severity dissociation with this OSA metric.

On the subjective side, the proposed framework incorporates nighttime symptoms, including snoring and subjective sleep quality, and daytime symptoms, including sleepiness, fatigue, and quality of life. Snoring is one of the most common and often earliest nighttime manifestations of OSA, and heavy or habitual snoring has been associated with carotid atherosclerosis independent of apnea severity. Snoring also imposes a considerable psychosocial burden on patients and their bed partners, and has been associated with marital problems and, in some studies, with higher levels of depressive and anxiety symptoms independently of OSA severity. Sleep fragmentation is an important contributor to symptoms and impaired daytime functioning in OSA, but arousal-related indices have shown limited incremental value over AHI for cardiovascular risk prediction in most analyses. Insomnia symptoms are reported by roughly 22 to 36 percent of patients with OSA, and the comorbid insomnia and sleep apnea (COMISA) phenotype has been associated with a higher prevalence of cardiovascular disease, increased risk of all-cause mortality, and a greater burden of depressive symptoms and psychological distress.

Among patients with similar AHI-defined disease severity, cluster-based analyses have consistently identified distinct symptom phenotypes: an excessively sleepy phenotype marked by high levels of daytime sleepiness and impaired alertness; a disturbed sleep or insomnia-like phenotype characterized by frequent awakenings, poor sleep quality, and fatigue; and a minimally symptomatic phenotype with limited perceived symptoms despite significant physiological disease burden. These phenotypes differ not only in symptom expression but also in demographic and clinical characteristics, and most importantly in clinical outcomes and treatment response. The excessively sleepy phenotype has been associated with a higher risk of adverse cardiovascular outcomes compared with other phenotypes, suggesting that symptom manifestation may capture clinically relevant disease burden not reflected by AHI alone. Treatment adherence also varies by phenotype, with patients experiencing excessive daytime sleepiness more likely to adhere to CPAP therapy, whereas minimally symptomatic or COMISA patients demonstrate lower adherence and may derive less perceived benefit from treatment. Notably, sex differences are observed, with women typically presenting lower AHI despite reporting greater impairments in quality of life than men, as well as higher rates of depression and anxiety.

The authors also draw parallels with other heterogeneous chronic diseases, such as chronic obstructive pulmonary disease (COPD), multiple sclerosis, and rheumatoid arthritis, in which clinical assessment has evolved from reliance on a single metric toward integrated models incorporating multiple domains of disease pathophysiology and patient outcomes. In COPD, for example, the Global Initiative for Chronic Obstructive Lung Disease (GOLD) classification integrates symptom burden with objective measures such as airflow limitation and exacerbation risk, enabling more accurate risk stratification and treatment personalization. Methodologically, development of a multidimensional disease-severity index for a heterogeneous chronic disease would typically involve defining domains through literature review and expert consensus, selecting and weighting candidate measures within each domain, and prospectively testing the composite for discrimination, reliability, construct validity, responsiveness to treatment, and prediction of clinically meaningful outcomes. The authors also highlight exposure-adjusted metrics such as the mean disease alleviation (MDA), calculated as the product of therapeutic efficacy and objective adherence, with reference values suggesting a mean MDA of approximately 54 percent after 6 months of CPAP and an MDA of at least 40 percent associated with clinically meaningful improvements in daytime sleepiness. An extended three-dimensional MDA (3D-MDA) has been proposed to integrate the symptoms domain, consistent with the multidimensional approach advocated here.

Looking ahead, the authors acknowledge that a future framework grounded in multidimensional evaluation of OSA will require thorough validation and ongoing discussion before it can be widely adopted in clinical practice. Key questions remain regarding the integration of proposed metrics for grading OSA risk and guiding treatment responses, including the refinement of clinically meaningful thresholds for measures such as HB. Studies have identified HB above approximately 60 percent-minutes per hour as a candidate threshold associated with increased cardiovascular morbidity and mortality, but reported cut-offs have varied across cohorts, and a large European cohort study showed a dose-dependent cardiovascular risk increase without a clear inflection point or safe threshold. Standardized definitions and analytical procedures across sleep centers have not been established, and feasibility is an equally important consideration, as comprehensive symptom assessment, sleep studies, and phenotyping require additional time, expertise, and infrastructure that may be limited in resource-constrained healthcare systems. Despite these remaining challenges, the authors argue that meaningful treatment benefit should not solely rely on normalization of AHI, but rather combine AHI reductions with improvements in the physiological and symptomatic domains most relevant to the individual patient. Defining clinical meaningfulness in OSA, they conclude, requires moving beyond AHI to a comprehensive, multidimensional approach that reflects the totality of disease biology, patient experience, and treatment exposure.

Subject of Research: Health & Medicine

Article Title: Beyond the apnea-hypopnea index: a call to action for a multidimensional evaluation in obstructive sleep apnea

Article References: Gay, P. C., Taranto-Montemurro, L., Kim, L. J., & Azarbarzin, A. (2026). Beyond the apnea-hypopnea index: a call to action for a multidimensional evaluation in obstructive sleep apnea. Frontiers in Sleep, 5, Article 1881431. https://doi.org/10.3389/frsle.2026.1881431

Image Credits: AI Generated

DOI: 10.3389/frsle.2026.1881431

Keywords: obstructive sleep apnea, apnea-hypopnea index, hypoxic burden, CPAP, sleep medicine, cardiovascular risk, symptom phenotypes, multidimensional evaluation, chronic intermittent hypoxia, Frontiers in Sleep, Beyond, apnea-hypopnea