Since the first report over four decades ago on the use of continuous positive airway pressure (CPAP) via a nasal mask to eliminate obstructive apneas, there have been few major advances in the treatment of obstructive sleep apnea (OSA) [1]. CPAP remains the first-line therapy for most patients with moderate-to-severe OSA who do not have correctable anatomical abnormalities. When properly used, it abolishes apneas and markedly reduces both daytime symptoms (mainly excessive sleepiness) and nocturnal symptoms (snoring, nocturia, awakenings) [2]. It also lowers blood pressure in hypertensive OSA patients, particularly those with resistant hypertension [3], and long-term observational data suggest potential reductions in cardiovascular morbidity and mortality [4]. However, several randomized controlled trials of 3–5 years in OSA patients with ischemic heart disease have failed to demonstrate benefit for secondary cardiovascular prevention [5–7], leaving the impact of CPAP on long-term survival uncertain.

In clinical practice, CPAP pressure is determined through manual or automatic titration, followed by periodic adjustment to minimize side effects and ensure adherence. The RICCADSA trial recently reported that CPAP pressures>7cm H2O were associated with increased circulating angiopoietin-2 (Ang-2)—a pro-inflammatory factor linked to endothelial activation—and reduced vascular endothelial growth factor (VEGF), a biomarker with potential cardioprotective properties [8]. This biomarker profile has been linked to increased cardiovascular risk [7,8]. Importantly, RICCADSA participants were survivors of myocardial infarction due to coronary artery disease (CAD), a group not representative of the broader OSA population. Given that most CPAP-treated patients with OSA receive pressures above 7 cmH2O, these findings raised concern that CPAP might induce a pro-inflammatory state even in most patients with OSA without established cardiovascular disease (CVD). Moreover, clinical trials in heart-failure populations have shown conflicting cardiovascular outcomes with CPAP or servo-ventilation [9,10], emphasizing the complexity of its vascular impact. Therefore, there is an unmet need to clarify the vascular biomarker response to CPAP in OSA patients without cardiovascular comorbidities.

To address this question, we analyzed longitudinal changes in Ang-2, VEGF, and related endothelial and inflammatory biomarkers at baseline, 1 year, and 5 years in the Epigenetics Modification in Obstructive Sleep Apnea (EPIOSA) study, an ongoing prospective cohort of OSA patients free of established cardiovascular disease and managed with or without CPAP according to current guidelines [11].

The EPIOSA study is a 5-year, prospective, non-interventional cohort study conducted at the Sleep Clinic of Hospital Universitario Miguel Servet, Zaragoza, Spain (ClinicalTrials.gov: NCT02131610) [11,12]. Its primary aim is to identify biomarkers and epigenetic factors associated with the prevalence and progression of subclinical atherosclerosis in OSA patients free from established cardiovascular and major metabolic comorbidities.

Consecutive participants aged 20–60 years were recruited during the first year. Inclusion criteria were: (1) Apnea–hypopnea index (AHI)≥15events/h (OSA group) or AHI<5events/h (control group); (2) willingness to participate and provide written informed consent; and (3) availability for 5 years of follow-up visits. For the purpose of this analysis, we excluded a previous diagnosis of diabetes, metabolic or endocrine disorders, cardiovascular or cerebrovascular diseases, autoimmune disease, malignancy, chronic inflammatory, infectious, or respiratory diseases; morbid obesity (BMI≥40kg/m2); sleep disorders other than OSA; prior upper airway surgery or previous CPAP therapy.

All participants received standardized lifestyle counseling, including recommendations for regular exercise, a healthy diet, and avoidance of smoking and alcohol. The study protocol was approved by the Regional Institutional Review Board of Aragón, Spain (IRB#03/2013, February 13, 2013). All participants provided written informed consent.

Of the 326 participants enrolled, 72 had AHI<5events/h (controls) and 195 had AHI>15events/h (OSA). Among patients with moderate-to-severe OSA, 118 received CPAP therapy and 77 were managed with usual care (Fig. 1).

Fig. 1.

Participant flow diagram.

Baseline characteristics are presented in Table 1. Compared with controls, both OSA groups were older and had higher BMI, neck circumference, waist-to-hip ratio, and blood pressure. As expected, OSA groups exhibited more severe sleep abnormalities and elevated hematocrit, fasting glucose, lipid profile parameters, uric acid, apolipoprotein B, and C-reactive protein. Medication use and baseline plasma biomarker concentrations are summarized in Tables S1 and S2. At baseline, E-selectin and VEGF were significantly higher in both OSA groups than in controls, while Ang-2 and Tie-2 showed no significant differences.

In OSA, baseline Ang-2 correlated positively with body weight (ρ=0.265, p<0.001), BMI (ρ=0.239, p=0.001), hypoxic burden (ρ=0.211, p=0.018), plasma sST2 (ρ=0.380, p<0.001), sRAGE (ρ=0.375, p<0.001) and Flt-1 (ρ=0.273, p=0.011). In multivariable models, only body weight (β=0.376, p=0.006) and plasma sST2 (β=0.666, p<0.001) remained independent predictors of baseline Ang-2 (Table S3 and Fig. S1). E-selectin correlated with age (ρ=0.178, p=0.013), body weight (ρ=0.181, p=0.011), BMI (ρ=0.250, p<0.001), waist-to-hip ratio (ρ=0.155, p=0.031), ESS score (ρ=0.150, p=0.039), mean nocturnal saturation (ρ=0.204, p=0.005), CT90.(ρ=0.180, p=0.014), and plasma concentrations of hemoglobin (ρ=0.155, p=0.032), triglycerides (ρ=0.204, p=0.004), sST2 (ρ=0.363, p<0.001) and IL-6 (ρ=0.387, p<0.001). Multivariable analysis identified body weight (β=0.274, p=0.003), sST2 (β=0.317, p=0.001), and IL-6 (β=0.378, p<0.001) as independent predictors (Table S4 and Fig. S2).

CPAP-treated patients used the device for an average of 6.7±1.1h at a mean pressure of 11±2cm H2O, with 112 patients (95%) meeting adherence criteria. Anthropometric variables differed between groups but remained stable over time (Table S5).

Plasma Ang-2 concentrations increased over time in patients with OSA irrespective of treatment allocation, whereas participants without OSA showed minimal longitudinal variation (Table 2). Formal interaction analyses showed no significant differences in Ang-2 trajectories between groups (group×visit interaction: F=1.118, p=0.347). Plasma VEGF concentrations in CPAP-treated patients were approximately 40–50% lower at follow-up compared with baseline values, whereas minimal longitudinal variation was observed in untreated OSA patients or controls. Formal interaction analyses confirmed significantly different VEGF trajectories between groups (group×visit interaction: F=3.154, p=0.012). No significant interaction effects were observed for Tie-2 or E-selectin.

No consistent dose–response relationship was observed between CPAP exposure variables and Ang-2 trajectories at either follow-up time point. In contrast, modest associations were identified between CPAP pressure and longitudinal VEGF changes at both follow-up visits. Fig. 2 illustrates the differential relationships between CPAP exposure and longitudinal changes in VEGF and Ang-2 concentrations.

Fig. 2.

Relationship between CPAP pressure and longitudinal changes in endothelial biomarkers in CPAP-treated OSA patients.

At 12 months, increases in Ang-2 correlated inversely with baseline Ang-2 (ρ=−0.420, p<0.001) and ApoA (ρ=−0.221, p=0.022), and positively with hematocrit (ρ=0.305, p=0.001). At 60.months, changes in Ang-2 correlated inversely with baseline Ang-2 (ρ=−0.171, p=0.020), mean nocturnal saturation (ρ=−0.311, p=0.001), and HDL cholesterol (ρ=−0.219, p=0.021), and positively with waist-to-hip ratio (ρ=0.280, p=0.003), CT90 (ρ=0.317, p<0.001), fasting glucose (ρ=0.148, p=0.046), ApoA (ρ=−0.202, p=0.003), and CCL2 (MCP-1) (ρ=0.303, p=0.029). Multivariable analysis retained baseline Ang-2 (β=−0.512, p<0.001) and hematocrit (β=0.426, p<0.001) as independent predictors of the 12-month change and mean nocturnal saturation (β=−0.229, p=0.020) for the 60-month change (Table S6 and Fig. S3).

For VEGF, 12-month changes correlated with baseline VEGF (ρ=−0.676, p<0.001), waist-to-hip ratio (ρ=0.191, p=0.038), mean nocturnal oxygen saturation (ρ=−0.189, p=0.043), and CT90 (ρ=0.200, p=0.035), while the 60-month changes correlated only with baseline VEGF (ρ=−0.549, p<0.001). Multivariable analysis identified baseline VEGF (β=−0.674, p<0.001) and waist-to-hip ratio (β=0.175, p=0.012) as independent predictors at 12 months, and baseline VEGF (β=−0.494, p<0.001) as the sole predictor at 60 months (Table S7 and Fig. S4). Additional mixed-effects models including baseline VEGF concentrations as a covariate showed materially unchanged longitudinal VEGF trajectories in the CPAP-treated group, suggesting that regression-to-the-mean effects alone are unlikely to explain the observed findings.

In this prospective cohort of moderate-to-severe OSA patients free of major comorbidities, long-term CPAP therapy was associated with a substantial and sustained reduction in VEGF concentrations, with median levels approximately halved during follow-up compared with baseline. Although the clinical implications of this magnitude of change remain uncertain, the consistency of the findings over 5 years supports a biologically relevant modulation of hypoxia-responsive pathways. These findings contrast with those reported in OSA patients with established CVD, where CPAP failed to consistently alter VEGF or even reduced potentially cardio protective biomarkers [6,18]. Our data therefore extend the previous evidence by demonstrating that, in otherwise healthy OSA patients, CPAP may be associated with favorable vascular effects, at least concerning VEGF biology.

VEGF is a potent endothelial mitogen essential for angiogenesis and vascular permeability [19]. In OSA, intermittent hypoxia stimulates VEGF production via hypoxia-inducible factor-1α (HIF-1α) activation [20], representing an adaptive mechanism to preserve microvascular integrity. The long-term reduction in VEGF observed here in CPAP-treated patients likely reflects suppression of nocturnal hypoxemia and associated oxidative stress. While the clinical impact of reduced VEGF remains uncertain, it could be associated with less maladaptive neovascularization and vascular remodeling associated with early atherosclerosis [21,22]. Further mechanistic studies should clarify whether VEGF downregulation under CPAP is accompanied by improvements in endothelial function and microvascular health.

In contrast to VEGF, Ang-2 trajectories appeared largely independent of CPAP exposure. The longitudinal increases observed were relatively modest in magnitude and were similarly present in untreated participants and controls, arguing against a clinically meaningful CPAP-related pro-inflammatory effect. This pattern suggests that Ang-2 dynamics may be influenced by non-OSA-specific or non-hypoxic factors. Ang-2, an endothelial-derived growth factor stored in Weibel–Palade bodies, antagonizes angiopoietin-1 at the Tie-2 receptor, promoting endothelial activation, inflammation, and vascular permeability [23]. Elevated circulating Ang-2 has been linked to adverse cardiovascular outcomes in diverse populations, including those with CAD [24], heart failure [25], and critical illness [26].

The RICCADSA trial reported higher Ang-2 levels in OSA patients with CAD treated with CPAP pressures>7cm H2O [8]. In contrast, in our cohort neither prescribed pressure nor nightly CPAP usage correlated with Ang-2 changes at 1 or 5 years. This divergence likely reflects differences in baseline vascular status; in CAD, the vascular endothelium is primed toward an activated, pro-inflammatory state [27], possibly rendering it more susceptible to mechanical or shear stress-related perturbations from CPAP therapy. By contrast, in otherwise healthy OSA cohort, endothelial homeostasis may buffer any pressure-related stress.

The progressive rise in Ang-2 in both OSA and control participants may be associated with age-related endothelial dysfunction and low-grade vascular inflammation [28], processes known to upregulate Ang-2 expression [29]. Metabolic factors (waist-to-hip ratio, glucose, HDL cholesterol, and hematocrit) were also associated with Ang-2 changes, supporting this hypothesis. The possibility that Ang-2 elevation reflects cumulative environmental or lifestyle exposures not captured in our protocol should also be considered. Other contributors, such as subclinical infection, oxidative stress, or intermittent sympathetic activation—can trigger Weibel–Palade body exocytosis and Ang-2 release [30].

The observed association between Ang-2 and baseline sST2 concentrations in OSA raises the possibility of coordinated vascular and myocardial inflammatory activation independent of hypoxia. Ang-2, a well-established marker of endothelial injury and vascular inflammation, remained elevated despite CPAP exposure, suggesting the involvement of alternative triggers such as alveolar stretch, systemic inflammation, or complement activation [31]. Likewise, sST2, a member of the interleukin-1 receptor family, reflects myocardial stress and fibrosis and is generally considered to be independent of hypoxia-driven signaling [32]. Although longitudinal sST2 measurements were not available, these findings suggest that endothelial and myocardial inflammatory pathways may coexist in OSA beyond hypoxemia-related mechanisms.

In CVD patients, Ang-2 is closely associated with vascular remodeling and adverse outcomes [33], but in our comorbidity-free cohort its long-term prognostic value remains unclear. The presence of similar trajectories in controls suggests that Ang-2 could reflect vascular aging rather than OSA-specific effects.

Overall, our findings suggest that inflammatory and angiogenic pathways activated in OSA may respond differently to CPAP exposure. While VEGF concentrations were lower during follow-up in CPAP-treated patients, Ang-2 trajectories did not significantly differ between study groups and showed no consistent relationship with CPAP exposure variables, underscoring their distinct regulatory mechanisms. VEGF is largely hypoxia-dependent through HIF-1α, whereas Ang-2 responds to cytokines (e.g., TNF-α, IL-1β) and hemodynamic stimuli [34,35]. These observations suggest that CPAP exposure may be associated with modulation of hypoxia-related signaling, whereas low-grade inflammatory activation may follow different longitudinal patterns. Given the observational design and non-randomized treatment allocation, these findings should be interpreted as longitudinal associations between CPAP exposure and vascular biomarker trajectories rather than evidence of direct causal effects.

Clinically, these results indicate that long-term CPAP is unlikely to exacerbate systemic inflammation via Ang-2 upregulation in patients without CVD, alleviating concerns raised by prior studies [8]. The absence of association between CPAP pressure and Ang-2 changes supports maintaining current titration practices. Furthermore, we have recently demonstrated that in real-world practice, in patients with OSA, with or without cardiovascular comorbidities, very long-term use of CPAP (14 years on average) is not only not associated with an increase in adverse cardiovascular effects, but also significantly reduces major cardiovascular events compared to OSA patients who were not treated with CPAP [36]. These observations are consistent with the long-term cardiovascular safety and potential benefit of CPAP use in patients with OSA.

This study has several limitations. First, biomarker measurements were limited to baseline, 1-year, and 5-year intervals. Second, the observational design precludes definitive causal inference, and treatment allocation was not randomized. Third, despite stringent exclusion criteria, residual confounding from unmeasured variables (e.g., dietary factors, subclinical infections) cannot be excluded. Fourth, most participants underwent unattended type 3 polygraphy rather than full polysomnography for OSA diagnosis. Although this approach may underestimate AHI and introduce some degree of OSA severity misclassification, the devices used were previously validated and type 3 polygraphy is widely accepted in routine clinical practice. Moreover, the use of conservative diagnostic thresholds likely minimized clinically meaningful misclassification between groups. Fifth, the ELISA assays used have intra- and inter-assay variability that, although within acceptable ranges, could influence the detection of small changes. Finally, while our cohort was well characterized, the sample size—particularly in the control group—limits the precision of subgroup analyses and the generalizability to other populations, including women and older adults.

In conclusion, in moderate-to-severe OSA patients without established cardiovascular disease, long-term CPAP exposure was associated with lower VEGF concentrations over follow-up, whereas Ang-2 trajectories appeared largely independent of CPAP exposure. Ang-2 increased over time in both OSA and non-OSA participants without a clear relationship to CPAP pressure or adherence. These findings support differential longitudinal behavior between hypoxia-related and inflammation-related angiogenic pathways in OSA. Future studies with randomized designs and more comprehensive sleep assessment are needed to clarify the causal and clinical significance of these biomarker trajectories.

DSR, CCZ, FGR, and JMM contributed to the study design. DSR, CCZ, MM-O, EDG, JR-S, PPM, CL-F, FGR, and JMM collected the data. CCZ and FGR performed the data analyses. DSR, CCZ, FGR, and JMM contributed to data interpretation. DSR, CCZ, FGR, and JMM drafted the first version of the manuscript. All authors reviewed, critically revised, and approved the final manuscript. DSR and CCZ contributed equally as co-primary authors; FGR and JMM contributed equally as co-senior authors.

Artificial intelligence tools were not used in the preparation of this manuscript.

Supported by Instituto de Salud Carlos III, Ministry of Health, Spain and the European Regional Development Fund (FEDER) (PI21/01954; CD22/00033), Department of Science and Universities, Government of Aragon, Spain (CUS/1638/2022; B22_23R), IIS Aragón, Spain (INTRAMURAL23/FSE-I/01) and Spanish Respiratory Society – SEPAR- (Research Grant 1584-2024).

The authors declare no competing interests.

De-identified individual participant data, along with the data dictionary and analytic code used in this study, are available upon reasonable request. Researchers interested in accessing these materials should submit a brief proposal outlining the planned analysis to the corresponding author. Data will be shared after review and approval of the proposal, provided that the intended use complies with ethical and legal requirements.