Obstructive sleep apnea (OSA) is a common condition in the general population, with prevalence estimates ranging from 9% to 38% depending on age, sex, diagnostic criteria, and apnea–hypopnea index (AHI) thresholds [1]. Large population-based studies using objective sleep recordings have reported even higher prevalences of elevated AHI in middle-aged and older adults, particularly among men and individuals with overweight or obesity [2]. Despite this high prevalence, OSA remains widely underdiagnosed, especially when symptoms are mild or absent, although it is well recognized to be associated with increased cardiovascular morbidity [3].

Venous thromboembolism (VTE), encompassing pulmonary embolism (PE) and deep vein thrombosis (DVT), remains a major public health problem. PE is the most frequent cause of cardiovascular death after myocardial infarction and stroke [4], and VTE is likely underestimated, as it may remain asymptomatic or be detected incidentally [5].

OSA and VTE share several risk factors and comorbidities, including older age, obesity, hypertension, diabetes, dyslipidemia, and metabolic syndrome. Metabolic syndrome, highly prevalent among patients with OSA, has also been associated with an increased risk of PE [6].

Moreover, from a pathophysiological perspective, recurrent obstructive apneas induce chronic intermittent hypoxia, leading to sympathetic activation, systemic inflammation, oxidative stress, and endothelial dysfunction [7,8]. These processes promote a prothrombotic state characterized by increased platelet activation, altered coagulation pathways, and impaired fibrinolysis. In parallel, hemodynamic changes associated with sleep-disordered breathing, including negative intrathoracic pressure swings, nocturnal fluid redistribution, and reduced venous return, may favor venous stasis [9]. These mechanisms can be further amplified in the context of acute pulmonary embolism, where hypoxemia, right ventricular strain, and inflammatory activation coexist. This biological framework provides a plausible basis for investigating the relationship between obstructive sleep apnea and venous thromboembolic disease.

Several observational studies have reported a high prevalence of OSA in patients with acute PE, even after adjustment for common confounders such as age and body mass index [10]. Additionally, available studies suggest that the presence of OSA may worsen the prognosis of acute PE [11].

However, the relationship between OSA and VTE remains complex. Recent large studies suggest that nocturnal hypoxemia metrics may be more strongly associated with thromboembolic risk than AHI alone, and that the independent role of AHI is less consistent after comprehensive adjustment for confounders [12,13]. The clinical relevance of elevated AHI in this setting remains debated.

Beyond the debated association between OSA and venous thromboembolism (VTE), the prevalence of OSA—particularly clinically relevant disease—among patients with acute VTE remains incompletely characterized. Recent large studies suggest that thromboembolic risk may be more strongly related to nocturnal hypoxemia than to the apnea–hypopnea index (AHI) per se, and evidence that CPAP modifies this risk is inconclusive. Nonetheless, the existing literature on sleep-disordered breathing in acute VTE is heterogeneous with respect to diagnostic strategies, symptom assessment, and the timing of sleep evaluation, making comparisons across studies challenging. Therefore, we conducted a systematic review and meta-analysis to estimate the prevalence of OSA in patients with acute VTE and to contextualize these findings within current epidemiological and clinical evidence.

We conducted a systematic review and meta-analysis following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) recommendations (Supplemental Table 1) [14]. The present study relies on aggregated data already published in Table 1, which are accessible from the electronic databases described below, and therefore research/ethics committee approval was deemed not required.

We used a PECOS (Population, Exposure, Comparison, Outcomes, Study design) framework to define the research query and search strategy (Supplemental Table 2). Articles were eligible if they reported the prevalence of OSA among patients with VTE, (PE and/or DVT), and reported on at least any two of the following: sample size, number of subjects with OSA, and/or OSA prevalence. Patients with superficial vein thrombosis, central sleep apnea and other sleep respiratory disorders, were excluded. Only primary research articles were included. Conference abstracts, review articles, case reports/case series were not included. Articles published in languages other than English, Spanish, French, or German were excluded.

We searched PubMed, Embase and Cochrane from inception to September-2024. Free key terms and MeSH/Emtree terms related to VTE, PE and DVT (population of interest) and OSA (exposure) were used (Supplemental Table 3). The search, screening and full text review were performed independently by three investigators, (any discrepancy resolved through discussions with two additional investigators). Fig. 1 shows the search and study selection process (details in figure legend). From the 5066 different articles retrieved, 25 studies were finally included in the systematic review [11,15–38].

Fig. 1.

Study selection flow diagram. After excluding the duplicates from the different databases, 5066 articles were retrieved. Through initial screening by title and abstract, 38 articles were selected for full text review. Of these, 13 articles were excluded as they did not fulfill the eligibility criteria for inclusion in the systematic review. Specifically, four were congress abstracts, two evaluated the prevalence of PE in patients with OSA instead of the reverse, three did not report OSA prevalence data, two were published in languages other than English, Spanish, French or German, one included patients with other sleep disorders without differentiating them from OSA in the prevalence estimates, and one did not establish an OSA diagnosis but instead assessed OSA risk using a questionnaire. Through checking the references from the selected articles and relevant review articles, no additional studies meeting the eligibility criteria were found. Therefore, finally, 25 studies were included in the systematic review (tables).

Data was extracted independently by two investigators and reviewed by one additional investigator (Tables 1–5). Where several estimates of OSA prevalence were reported (e.g. based on different diagnostic criteria/severity, all of them were recorded.

Quality assessment of individuals studies was performed independently by two investigators (any disagreements resolved through discussion with a third investigator) by applying an adapted quality assessment tool for assessing risk of bias in prevalence studies (Hoy et al.; Supplemental Table 4) [39].

Data from each individual study is presented in detail in Tables 1–5. Prevalence is presented as absolute numbers and percentages. For each study we took the prevalence as reported or, if not indicated specifically, this was calculated from the number of patients with OSA and the total sample (or subgroup of interest) of the study. The 95% confidence intervals (95%CI) for the prevalence in each study were estimated using the binomial exact method (Clopper-Pearson method) from the total sample (or subgroup of interest) and proportion of OSA cases.

A meta-analysis was conducted to estimate the pool prevalence across the studies. Heterogeneity was anticipated and therefore a random-effects model (inverse variance method) was used. Before meta-analysis, the Freeman–Tukey double arcsine transformation was applied to the primary study data. Final results were back-transformed for presentation of results. Between-study heterogeneity was assessed with the I2 statistic. If the same cohort of participants (in full or in part) was included in different articles, only one of them was included in each set of analysis to avoid entering the same participants twice in the same analysis. Analyses were performed based on different OSA diagnostic criteria/severity, AHI≥5, ≥15,≥30, for studies where all study participants were systematically assessed for the presence of OSA through sleep studies. Separately, analyses were conducted for studies assessing an already known diagnosis of OSA through screening of health records. Small study effects/publication bias were assessed graphically using funnel plots; asymmetry was further examined using Peter's regression test. Where applicable, we applied the Duval and Tweedie trim-and-fill method to adjust for funnel plot asymmetry. Assessment of potentially influential studies was performed by searching for outlier studies using the function “findoutliers” and conducting sensitivity analysis excluding these outliers. Sensitivity analyses excluding studies with sample sizes<30 or <50 were also performed. Where reported data allowed, we conducted subgroup meta-analysis and meta-regressions Analyses were conducted using R (v4.2.3) through RStudio (v2023.03.0).

Twenty-five studies were included in the systematic review. Some of them reported on the same cohort of participants (in full or partly) as shown in the tables. Table 1 shows the characteristics of these studies. Sample sizes ranged from 30 to 268 participants in the studies systematically testing all participants for a diagnosis of OSA (21 studies), and between 840 and 755,532 participants in the studies that assessed for an established OSA diagnosis through screening of health records/databases (4 studies). Table 2 describes the characteristics of the participants in each study. In most studies, participants were in their 50s and 60s. Proportion of women was 38–60%, except in one study where it was 85.7%. When reported, body mass index (BMI) was in the range of overweight or obesity. Only ten studies reported the proportion of smokers (20–47%). Comorbidities such as cardiovascular or lung diseases were infrequently reported.

The diagnosis and characteristics of VTE are reported in Table 3. Criteria for diagnosis were not reported in six studies. The criteria for a PE diagnosis varied among the studies. In 22 studies all patients had PE; in the other three studies, this figure ranged 55.5–85.3%. Where reported (10 studies), the proportion of patients with DVT varied from 14.7% to 73.8%.

Table 4 shows the diagnosis and characteristics of OSA patients and the prevalence of OSA among participants with VTE. Among studies testing systematically all participants for a diagnosis of OSA, 13 reported the OSA prevalence according to an AHI≥5, and it ranged 50.8–82.4%. OSA prevalence according to an AHI≥15 was reported in 16 studies, with results ranging 15.5–63.2%. The prevalence of severe OSA (AHI≥30) was reported by 10 studies, and it ranged 10.4–28.4%. Finally, 4 studies assessed the prevalence of a known diagnosis of OSA through screening of health records, and the prevalence varied between 3.5% and 15.5%. Table 5 shows the OSA prevalence in different subgroups. Overall, where data reported, there was a trend toward a higher prevalence of OSA among men and in those with obesity or a history of stroke or heart failure.

Quality assessment of the studies suggested an overall low risk of bias in dimensions related to internal validity of the studies, but there was frequent high risk of bias in aspects related to external validity (i.e., generalizability) (Supplemental Table 4).

Meta-analysis

Meta-analysis estimates of OSA prevalence according to different criteria are shown in Fig. 2. Ten studies, comprising 1175 individuals with VTE, reported on the prevalence of OSA according to an AHI≥5, resulting in a pooled prevalence of 70% (95%CI 63–76%; I2 80%); Fig. 2A. OSA prevalence according to an AHI ≥15 was reported in 9 studies (1067 VTE individuals), resulting in a prevalence of 41% (95%CI 32–50%; I2 85%); Fig. 2B. Seven studies (855 VTE individuals) reported on the prevalence of severe OSA (AHI≥30), leading to a pooled prevalence of 21% (95%CI 16–26%; I2 70%); Fig. 2C. In only one study in the meta-analyses not all participants had a PE (i.e., they could have a DVT only); excluding this study did not meaningfully change the results (Supplemental Fig. 1A, B). Funnel-plots are shown in Supplemental Fig. 2A–C. Graphically, it seemed there was some asymmetry particularly in the group of severe OSA; a formal statistical test was not significant in any case (Peter's regression tests, all p>0.65). Nevertheless, to adjust for potential funnel-plot asymmetry, we applied the trim-and-fill method. This procedure filled the meta-analysis with two, none and three additional studies in the groups of AHI≥5, ≥15 and ≥30, respectively. It did not change the estimate in case of AHI≥5 or ≥15 (Supplemental Fig. 3), but the estimate was higher in the meta-analysis of severe OSA (26%; 95%CI 19–32%; Supplemental Fig. 4), suggesting a potential underestimation of the prevalence of severe OSA by the original meta-analysis (21%). The assessment of potential influential studies found one and two outliers the in groups of AHI≥5 and ≥15, respectively; however, excluding these studies did not meaningfully change the pooled estimates (Supplemental Table 5). Finally, sensitivity analyses excluding studies with sample sizes<30 or <50 did not change the results either (Supplemental Table 6).

Fig. 2.

Prevalence of obstructive sleep apnea among patients with venous thromboembolism, according to an apnea–hypopnea index≥5, overall in all participants with venous thromboembolism (Panel 2A) and among patients with pulmonary embolism specifically (with or without deep vein thrombosis) (Panel 2B); and according to an apnea–hypopnea index≥15, overall in all participants with venous thromboembolism (Panel 2C) and among patients with pulmonary embolism specifically (with or without deep vein thrombosis) (Panel 2D). Prevalence (“proportion”) and 95%-CI estimates for individual studies and pooled prevalence in the forest plots are shown as number of cases in 1; multiply by 100 to get prevalence percent. 95%-CI: 95% confidence intervals.

Four studies (805,275 VTE individuals) assessed the prevalence of an already established diagnosis of OSA through screening of health records; this meta-analysis resulted in a prevalence of 8% (95%CI 4–13%; I2 100%); Fig. 2D. Meta-analysis considering separately whether all participants had a PE or if they could have a DVT without PE, did not show a significant difference between both groups (p=0.76); Supplemental Fig. 1C. Funnel plot suggested asymmetry (Supplemental Fig. 2D). However, the trim-and-fill method did not fill the meta-analysis with additional studies. On the other hand, the exclusion of one identified outlier study led to a lower OSA prevalence (5.6%; 95CI 3.3–8.4%) than in the original meta-analysis (Supplemental Table 5).

Conducting subgroup meta-analyses and meta-regressions was limited due to the small number of studies reporting the prevalence stratified by the variables of interest in each OSA criteria group. In the 4 studies assessing OSA through screening of health records, the prevalence was approximately 5 times higher among those with obesity (20% [95%CI 12–29%], vs no obesity: 4% [95%CI 3–5%], p<0.01); Supplemental Fig. 5. There was a statistically non-significant trend toward a higher prevalence of OSA by AHI≥15 in men than women (3 studies; men: 48% [95%CI 36–61%], women: 35% [95%CI 25–45%], p=0.11); Supplemental Fig. 6. Meta-regression analyses did not show any statistically significant association between the study sample sizes, year of publication, average age or BMI, or proportion of women in the studies, with the prevalence of OSA; Supplemental Table 7.

OSA is the most common sleep disorder in the adult population. However, the prevalence of OSA varies between different studies according to their methodology, diagnostic criteria, severity of OSA, presence of comorbidities and geographical region [3]. Benjafield et al. described a wide variability in the OSA prevalence, ranging 4–30% based on an AHI≥15, with higher frequencies for milder OSA and exceeding 50% in some countries [40,41].

In the present study, we observed that OSA was more prevalent among patients with VTE than in the general population (70% for mild OSA to 41% for moderate-to-severe OSA) with 1 in 5 VTE patients having severe OSA, altogether suggesting a high burden of OSA in VTE. There was a high heterogeneity among the studies, and the evaluation of the quality of studies suggested a potential limited external validity of the results for some individual studies. These circumstances should be considered for interpretation of results. The high heterogeneity suggests a wide dispersion of the effect sizes of individual studies, methodological differences and/or diversity of the participants and diagnostics tools among the studies. Therefore, results should be interpreted taking into account these circumstances and, beyond the prevalence estimated, we should look at the 95%CI to see the precision of this estimator. To explore potential sources of heterogeneity, we assessed small study effects/publication bias, potential influential studies, sensitivity analysis, and, as far as reported data allowed, subgroup meta-analysis and meta-regressions. No significant findings were observed apart from some differences in prevalence within some AHI subgroups based on some characteristics of the participants such as sex or obesity. We must mention, however, that many variables of interests (characteristics of participants, VTE or OSA; OSA prevalence based on certain characteristics) were very frequently not reported in the included studies (see Tables) and it precluded us for conducting further analyses Between-study heterogeneity may be related to differences in study design and source of participants, characteristics of the populations studied, the diagnostic criteria for VTE, sleep studies conducted for assessment of OSA or time of this assessment from the VTE, among other factors. Finally, concerns on external validity may be more likely related to the selection of the sample, not necessarily being representative of the whole VTE population, thus affecting the generalizability of the results.

An important finding of our study is the observation of a large difference between the prevalence of known OSA among patients with VTE and the real prevalence of OSA that is found when all VTE patients are systematically evaluated for this condition. In fact, the prevalence of known OSA among VTE patients was 8%, what substantially contrasts with the prevalence of 70% (for AHI≥5) or 41% (for AHI≥15) that was found when all VTE patients were systematically checked for OSA. Given the substantial between-study heterogeneity and limited external validity in some cohorts, these findings should be interpreted cautiously. On one hand, the marked contrast between AHI-based prevalence and the low proportion of previously diagnosed OSA should be interpreted with caution, as many AHI≥5 cases likely represent mild or incidental disease in an older, overweight population. In this context, estimates based on AHI≥15 may better reflect clinically relevant OSA. On the other hand, from a clinical perspective, these findings do not support universal screening, but rather argue for increased awareness of OSA in VTE and a targeted diagnostic approach. Evaluation may be particularly relevant in patients with a higher pre-test probability of clinically meaningful disease, such as those with suggestive symptoms, obesity or large neck circumference, cardiometabolic comorbidity, or evidence of significant nocturnal hypoxemia. In this context, simple validated tools such as STOP-Bang may help prioritize patients for sleep assessment, while greater emphasis on hypoxemia-related phenotypes rather than AHI alone is consistent with contemporary expert recommendations.

Beyond prevalence, the prognostic relevance of OSA in acute VTE/PE remains uncertain because outcome definitions, follow-up, and endpoint ascertainment varied across included studies, precluding quantitative synthesis. Complementary evidence comes from the RIETE registry analysis by Le Mao et al. [42], with more than 4000 acute PE patients, which was not eligible for meta-analysis because OSA was based on a prior clinical diagnosis. Previously recognized OSA was uncommon (5.8%), consistent with our ‘known OSA’ estimate, and was associated with higher PE-related mortality after adjustment (adjusted OR≈3, as reported). These results warrant caution due to likely misclassification, residual confounding, and differences in case-mix and management, but they reinforce the need for prospective studies using standardized sleep and hypoxemia related metrics beyond apnea–hypopnea index alone to clarify prognostic pathways and actionable risk stratification [43].

Overall, the high prevalence estimates and apparent under-recognition of OSA in VTE/PE indicate a relevant comorbidity burden and a potential gap in routine care. Future work should prioritize targeted case-finding strategies (symptoms plus clinical predictors and, where available, hypoxemia measures), identify the most prognostically informative sleep metrics, and test whether OSA identification and treatment modifies VTE/PE outcomes through randomized studies. One example is the ongoing HEMERA investigator-initiated randomized study.

The prevalence of OSA is described to peak at age 45–65 years in the general population [3]. Our evaluation of OSA prevalence by age was rather limited, since the prevalence of OSA among VTE patients stratified by age was not reported in any of the studies in the systematic review. Meta-regression analysis did not find any significant association between the average age of the participants in the studies and the prevalence of OSA; however, average age of participants in most of the studies was similar (patients in their 50s-60s), what may have limited finding any difference by age in this analysis. In the study by Kezban OS et al. [18] mean age of patients without any significant risk factor for PE was significantly higher (61 years) than those with any major risk factors (52 years). This observation may suggest that age is an important risk factor for PE; however, when multivariate analysis was performed, it was observed that the factor making the real impact was OSA, supporting the association between these two conditions.

OSA is more common in men [11,18,19], and this pattern was also observed in our report among VTE patients in most of the 10 studies that presented the data stratified by sex. The meta-analysis also confirmed this trend for OSA prevalence defined as AHI≥15 (48% in men and 35% in women), albeit non statistically significant (p=0.11, likely due to lack of statistical power since only 3 studies could be included in this subgroup meta-analysis). In these studies, mean age of participants was around 60 years. It is worth notice that OSA incidence in women is described to increase significantly after menopause, with OSA prevalence in men and postmenopausal women tending to be similar.

The presence of other lung or cardiovascular diseases may be important factors impacting the occurrence of OSA in VTE. However, the report of these comorbidities was scarce in the studies. Chronic obstructive pulmonary disease ranged 4–27% in the seven studies that mentioned lung diseases [15,18,26,32,34,36,38]. OSA is also described to be a common condition among patients with stroke and to also be a risk factor for stroke (with stroke incidence increasing with the severity of OSA) [44]. Some studies reported on the OSA prevalence based on the PE risk category or sPESI score; the limited number of these studies prevented us from conducting subgroup meta-analysis based on these variables; however, findings among these studies were not consistent: while some observed a trend toward a higher OSA prevalence among individuals with intermediate or high-risk PE (vs. low-risk) or with sPESI≥1, this was not observed in others. Finally, certain methodological variability among the different studies, including differences in OSA diagnostic modality (e.g. PSG vs. polygraphy) and criteria, or timing of sleep assessment, among others, may also affect the prevalence estimates; these circumstances are described in detail in Table 4, and should be also considered for interpreting the results.

We must acknowledge several limitations of this systematic review. For instance, the limited samples sizes of most studies; the large between-study heterogeneity (which was anticipated and therefore a random-effects model for meta-analysis was employed, as well as conducting bias-adjusted estimates and sensitivity, subgroups and meta-regression analyses where possible, to assess potential sources of heterogeneity); however, the high heterogeneity means we should interpret the results with caution and under the circumstances further explained above in the discussion. And the potential limitations with the external validity of some studies as suggested by the quality of evidence assessment (discussed above). There is also an absence of data from large world regions, which is relevant considering the wide variation in OSA prevalence in the general population reported among different countries.

The results from the present systematic review and meta-analysis suggest a high prevalence of OSA (including for moderate and severe OSA) among patients with VTE (up to 4–7 in 10 VTE patients), mostly PE (with or without DVT), which seems to be higher than the one reported in the general population, regardless of the severity of OSA. This burden of OSA among patients with VTE seems to be further compounded by a large under-diagnosis of OSA among these patients.

Our results support increased awareness of sleep-disordered breathing in patients with VTE—particularly those with symptoms or features suggestive of OSA—and consideration of diagnostic evaluation on an individualized basis, consistent with contemporary expert recommendations. Further studies are needed to confirm these prevalence estimates, to clarify which sleep-related metrics (including hypoxemia measures) are most relevant, and to determine whether targeted identification and treatment of OSA influence clinical outcomes in VTE. Further research is needed to confirm these prevalence estimates, to clarify which sleep-related metrics (including hypoxemia measures) are most relevant, and to determine whether diagnosing and treating OSA (e.g., CPAP) modifies VTE/PE outcomes; prospective, protocolized studies, including ongoing randomized interventions will be critical to address this question.

• •

  CCC, AVV, CCE, ROC: have design the work, performed the analysis, have performed the analysis and they have interpreted the data. They have draft the work and review critically.

• •

  VSL, EMZ, ABA, CLR, MAC: they have performed the data acquisition and help with the design.

• •

  FJM, ECS, ASA: help to design the work and critically review the paper.

• •

  All the author have approve the final version of this paper.

The present study relies on aggregated data already published in previous publications which are accessible from the electronic databases described, and therefore research/ethics committee approval and inform consent was deemed not required.

During the preparation of this work the authors used DALL-E in order to prepare the graphical abstract. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the published article.

Beca SEPAR (Sociedad española de enfermedades respiratorias) para proyectos de investigación: Estudio clínico-traslacional del impacto de la Apnea obstructiva del sueño en la persistencia de los defectos de perfusión tras una Embolia de pulmón. “NIX study”. Financiación:18.000€.

PI21 – Proyectos de investigación en salud (AES 2021). Modalidad proyectos de investigación en salud. Instituto de Salud Carlos III, de la convocatoria 2021 de la Acción Estratégica en Salud 2017–2020. Ministerio de ciencia e innovación, Spanish government: “Estudio clínico-traslacional del impacto del síndrome de apneas nocturnas en la restauración de la perfusión tras una embolia pulmonar. NIX study.”

The authors declare not to have any conflicts of interest that may be considered to influence directly or indirectly the content of the manuscript.