Beyond the apnea-hypopnea index: Precision cardiovascular risk stratification in obstructive sleep apnea

Obstructive sleep apnea (OSA) has a complex and unresolved relationship with cardiovascular medicine. Observational work has consistently linked OSA and other sleep-disordered breathing disorders with adverse cardiovascular outcomes, including hypertension, arrythmia, stroke, coronary artery disease, and mortality. However, in randomized trials of the effects of continuous positive airway pressure (CPAP), results have generally not shown associated reductions in cardiovascular events, despite improvements in sleep-related symptoms and some physiological measures. One explanation for this discrepancy is that cardiovascular and sleep medicine have examined OSA as a relatively uniform cardiovascular exposure. This approach has prompted increasing attention to the clinical heterogeneity within the OSA phenotype, including the potential contributions of cardiovascular vulnerability, comorbid risk profiles, and biological sex.

In this issue of the American Journal of Preventive Cardiology, Kosasih and colleagues attempted to address this challenge by applying phenogroup-based risk stratification to patients with coronary artery disease after coronary revascularization [1]. Across two prospective cohorts of 2318 patients treated with either percutaneous coronary intervention or coronary artery bypass grafting, latent class analysis identified three phenogroups that were broadly characterized as: 1) younger sleepy individuals with greater nocturnal hypoxemia, a phenogroup that has been consistently reproduced in prior unsupervised clustering analyses of OSA [2]; 2) patients with greater cardiorenal-metabolic comorbidity burden; and 3) older, lean Chinese patients with lower comorbidity burden. While the prevalence of OSA (defined by apnea-hypopnea index ≥15) was similar across groups, the association between OSA and major adverse cardiac and cerebrovascular events (MACCE) appeared to differ within phenogroup. Although the study had limitations in the analytical approach, the findings emphasize growing evidence that the cardiovascular implications of OSA may depend less on the presence of OSA per se and more on the underlying clinical context and pathophysiological mechanisms involved.

The disconnect between observational associations and neutral cardiovascular outcome CPAP trials is unlikely to be explained by apnea-hypopnea index (AHI) alone. AHI remains an incomplete surrogate for the biological consequences of sleep-disordered breathing and does not adequately capture heterogeneity in hypoxic burden, autonomic activation, endothelial dysfunction, inflammatory signaling, sleep fragmentation, or symptom expression. Consequently, equivalent AHI may not reflect equivalent cardiovascular vulnerability or treatment responsiveness. Thus, phenogroups may capture differing susceptibility to the biological pathways through which sleep-disordered breathing contributes to cardiovascular injury [2,3].

Still, Kosasih et al’s findings also underscore how far the field remains from clinically actionable phenotyping. Although the authors identify potentially differing patterns of association between OSA and cardiovascular risk across phenogroups, the signal across analyses is less definitive than the subgroup labels themselves may imply. Phenogroup 2 carried the greatest cardiometabolic burden and highest prevalence of severe OSA but only demonstrated borderline associations with MACCE after adjustment. In contrast, the results from Phenogroup 1 suggested that the association between OSA and recurrent cardiovascular events may evolve over time. Importantly, Phenogroup 1 was also the youngest subgroup, suggesting that OSA-related cardiovascular vulnerability may be more prominent earlier in the disease trajectory, while competing cardiometabolic risk pathways may increasingly drive events over longer follow-up.

The current limitations of OSA phenotyping do not invalidate the approach. In other areas of cardiometabolic medicine, phenotype-driven approaches have meaningfully advanced both mechanistic understanding and therapeutic development. For example, heart failure with preserved ejection fraction (HFpEF) phenomapping studies have consistently identified obesity-related inflammatory phenotypes across cohorts, analytic methods, and variable sets [4]. Notably, these phenotypes became clinically meaningful because they aligned with underlying pathophysiology, therapeutic responsiveness, and patient-level outcome prediction rather than descriptive clustering alone [5,6]. The common denominator was that the variables driving cluster separation reflected underlying disease biology rather than broad demographic or comorbidity descriptors alone.

OSA phenotyping faces a similar challenge. Many current OSA phenogroups still cluster heavily around age, obesity, comorbidity burden, and symptom patterns rather than direct measures of the biological pathways presumably driving cardiovascular injury. Therefore, reproducibility alone is difficult. The larger challenge is determining if phenotypes identify treatment-responsive cardiovascular vulnerability, meaningfully improve patient-level prediction beyond traditional risk stratification approaches, and align with the specific cardiovascular outcomes most biologically linked to sleep-disordered breathing [7]. It is plausible that different OSA-related pathways could preferentially contribute to distinct cardiovascular phenotypes, such as atherothrombotic disease, heart failure progression, arrhythmia burden, or cardiorenal dysfunction.

Put differently, the field still lacks a coherent framework for identifying OSA-attributable cardiovascular risk at the individual patient level. Current phenotyping approaches move beyond traditional AHI-centered frameworks, but many remain phenotypically shallow, heavily dependent on demographic and comorbidity profiles, and insufficiently linked to biological mechanisms or treatment responsiveness. Critically, identifying patients at a high absolute cardiovascular risk is not synonymous with identifying patients whose cardiovascular risk is mechanistically driven by sleep-disordered breathing or modifiable through OSA-directed therapy.

The central challenge moving forward is not simply identifying phenogroups with differing event rates. Rather, it is determining which patients exhibit biologically actionable cardiovascular vulnerability related to sleep-disordered breathing and are therefore most likely to derive cardiovascular benefit from treatment [3]. Those are not necessarily equivalent missions. A patient could possess high absolute cardiovascular risk while deriving minimal benefit from OSA-directed therapy, while another with fewer traditional risk factors may exhibit disproportionate vulnerability to intermittent hypoxia, autonomic surges, endothelial dysfunction, or sleep fragmentation. Biological sex may further modify these pathways through differences in autonomic regulation, hormonal milieu, endothelial function, symptom expression, and cardiometabolic susceptibility.

Prior CPAP trials may have struggled to demonstrate cardiovascular benefit because OSA was treated as a uniform cardiovascular exposure rather than enriching the sample for patients with OSA-attributable cardiovascular vulnerability. In this framework, neutral trial results may reflect inadequate biological enrichment rather than absence of cardiovascular treatment effect. Most trials enrolled relatively asymptomatic patients or excluded those with marked daytime sleepiness, limiting generalizability to the symptomatic patients most often considered for CPAP in practice. Eligibility that was largely based on AHI thresholds likely captured many patients whose cardiovascular risk was driven primarily by non-OSA factors, while excluding some patients most likely to benefit symptomatically. Therefore, neutral results should not be read as definitive evidence that CPAP lacks cardiovascular effect. They may instead reflect a failure to identify the subgroup whose cardiovascular risk is meaningfully driven by sleep-disordered breathing. Consistent with this interpretation, recent post hoc analyses across randomized CPAP trials indicated that cardiovascular benefit may concentrate among patients with higher-risk physiological OSA profiles – greater hypoxemic burden or OSA-related pulse-rate response [3].

Future cardiovascular and sleep research must evolve beyond treating OSA as a binary cardiovascular exposure and instead focus on identifying when, for whom, and through which biological pathways sleep-disordered breathing meaningfully contributes to cardiovascular disease progression. Doing so will require the integration of OSA metrics beyond AHI, including hypoxic burden, autonomic activation, symptom burden, and treatment adherence, with cardiovascular vulnerability domains such as inflammatory signaling, endothelial dysfunction, metabolic dysfunction, cardiorenal comorbidity, and reduced vascular reserve. Candidate phenotypes should then be validated by demonstrating mechanistic coherence, incremental prediction beyond conventional cardiovascular risk factors, and evidence that OSA-directed therapy modifies outcomes in the patients identified as vulnerable.

The present study represents a nascent step in the direction of a broader shift within cardiovascular prevention away from disease labels alone and toward biologically informed risk stratification. The next phase of cardiovascular sleep research should not primarily focus on whether OSA universally increases cardiovascular risk, but rather on identifying biologically actionable cardiovascular vulnerability that is attributable to sleep-disordered breathing. The field’s central limitation may not be that CPAP lacks cardiovascular efficacy, but that existing frameworks inadequately distinguish patients whose cardiovascular disease is mechanistically driven by OSA from those for whom OSA is largely an epiphenomenon of broader cardiometabolic risk. Precision cardiovascular phenotyping may ultimately determine whether OSA-directed therapies can meaningfully alter cardiovascular outcomes.

CRediT authorship contribution statement

Allison E. Gaffey: Writing – review & editing, Writing – original draft, Conceptualization. Filipe A. Moura: Writing – review & editing, Writing – original draft, Conceptualization.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:

Alisson Gaffey reports financial support was provided by National Institutes of Health. Given his role as Associate Editor, Filipe A. Moura had no involvement in the peer review of this article and had no access to information regarding its peer review. Full responsibility for the editorial process for this article was delegated to another journal editor. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.