Overall Mortality and Comorbidities in Obstructive Sleep Apnea in Poland

Introduction

Obstructive sleep apnea (OSA) is characterized by the recurrent cessation of breathing (apneas) or partial upper airway obstruction (hypopneas) during sleep, and affects from 10% to 30% of the population [1,2]. OSA increases the risk of cardiovascular and cerebrovascular morbidity and mortality and has been recognized as a source of attributable risk in patients with arterial hypertension, obesity, diabetes, and cardiovascular disease that leads to premature death [3,4]. Polysomnography is considered the gold standard in OSA diagnostics. The apnea-hypopnea index (AHI) is a well-known, widely used metric for indicating the severity of OSA [5]. There is growing evidence that AHI does not accurately estimate the risk of death, even though AHI is used to diagnose and evaluate the severity of OSA [6]. Depending on the number of apneas and hypopneas per hour, OSA can be classified as mild (AHI ≥5 and <15), moderate (AHI ≥15 and <30) or severe (AHI ≥30) [7]. Reported AHI values typically refer to AHI for the total sleep time (TST). However, it is important to remember that AHI can be based on the time spent sleeping in the back or side positions, referred to as AHIback and AHIside, respectively. Additionally, the AHI can be calculated for rapid eye movement (REM) sleep and non-rapid eye movement (NREM) sleep, referred to as AHIREM and AHINREM, respectively. Therefore, categorization relying solely on the AHI related to TST alone proves inadequate. Consequently, extensive analyses have been conducted, leading to the identification of distinct phenotypic manifestations of OSA [8,9]. These distinctions are based on the severity of clinical symptoms, with a specific focus on daytime sleepiness and comorbidities to identify cohorts of patients who would derive maximum benefit from continuous positive airway pressure (CPAP) therapy. However, the impact of CPAP on overall mortality in the case of OSA is a subject of discussion [10–12]. Most of these analyses consider AHI as the sole predictor of the severity of disease, often neglecting critical data such as the predominant type of respiratory events or OSA subgroups related to sleep positions or sleep stages. We hypothesize that individuals with OSA, have diverse predictors of all-cause mortality and OSA-related mortality, beyond AHI. Although the impact on morbidity and mortality and the precise relationship between OSA and mortality remains complex and multifaceted, the mortality among OSA patients can be attributed either to OSA-related causes – conditions directly linked to the pathophysiology of the dis order – or to all-cause mortality, which considers all causes of death, including those seemingly not related to OSA [13,14].

The primary objective of this study was to assess the mortality risk of patients diagnosed with OSA based on a set of variables, such as parameters measured during polysomnography, comorbidity history, medications taken regularly, and symptoms reported at admission. The secondary objective of this study was to compare each of the particular risk factors in the study groups during 5, 10, and 15 years of observation. By exploring this distinction, we aimed to deepen the understanding of OSA’s role in overall mortality and provide valuable insights into targeted prevention strategies for this vulnerable population.

Material and Methods

STUDY DESIGN:

This is a retrospective, observational cohort, single-center study conducted at the Department of Sleep Medicine and Metabolic Disorders, Medical University of Łódź, Poland. Patients were admitted between 2005 and 2019, and all of them underwent overnight polysomnography, which was one of inclusion criteria. Patients who were referred with a suspected diagnosis of OSA were admitted to the sleep clinic at 8: 00 pm and underwent physical examination, including measurement of body mass, height, calculated body mass index (BMI), heart rate, and blood pressure. All patients completed a questionnaire regarding comorbidities, with the attending physician. These comorbidities included arterial hypertension, diabetes, atrial fibrillation, dyslipidemia, depression, stroke, myocardial infarction, and drugs taken, including statins, oral hypoglycemic drugs, insulin, acetylsalicylic acid, and anticoagulants. Signs and symptoms of OSA were self-reported by the patients during the admission process, with special interest in self-reported sleepiness by the patients and measured on the Epworth Sleepiness Scale (ESS). In case of any doubts, the attending physician clarified the questions and verified the list of medications taken. Other symptoms, including morning fatigue, snoring, or morning headaches, were self-reported by the participants. Patients were classified as having arterial hypertension if: 1) they self-reported a history of arterial hypertension; 2) they self-reported medications used in the treatment of arterial hypertension; or 3) their mean blood pressure exceeded 140/90 mmHg during 1 of 3 measurements (first consultation prior to the polysomnographic examination, or blood pressure measurement before or immediately after the polysomnographic examination). Dyslipidemia was defined as self-reported dyslipidemia by the patient or the presence of statins in the list of medications taken regularly. Diabetes was defined as self-reported diabetes or the presence of hypoglycemic agents including insulin.

OUTCOME VARIABLES:

Patients were followed up for up to 15 years for all-cause mortality and plausibly OSA-related mortality (Figure 1). The cause of death was verified by the Ministry of Digital Affairs and the Central Statistical Office in accordance with International Classification of Diseases (ICD-10). Mortality data were recorded up to the 24th of February 2024. Loss to follow-up was negligible, as mortality status was obtained for all patients through comprehensive national registries. The group of patients classified under OSA-related mortality included individuals whose cause of death, as determined by the ICD-10 classification, was categorized as cardiovascular diseases, respiratory system diseases, or other conditions related to OSA, such as obesity, traffic accidents, and suicides.

POLYSOMNOGRAPHY:

Polysomnography started at 10 pm and continued until 7 am the next day, and somnographic findings were collected from inpatient reports. A standard nocturnal, single-night polysomnography was performed by recording the following channels: electroencephalography (C4\A1, C3\A2, F3\O1, F4\O2), chin muscles and anterior tibialis electromyography, electrooculography, measurements of oronasal air flow (a nasal canula), snoring, body position (via a gravitational gauge placed on the sternum), respiratory movements of the chest and abdomen (piezoelectric gauges), unipolar electrocardiogram, average oxygen saturation, and time with oxygen saturation <90% (SpO2 <90), as measured by pulse oximetry (Sleep Lab, Jaeger - Viasys, Hoechberg, Germany). Sleep stages were scored manually by 3 somnologists with European Sleep Research Society certification, according to American Academy of Sleep Medicine (AASM) criteria based on 30-second epoch standards. Apnea was indicated when the reduction of airflow resulted in less than 10% of the baseline airflow for at least 10 seconds. Hypopnea was defined as at least 30% reduction of airflow for at least 10 seconds, accompanied by a 3% or greater decrease in SpO2 or an arousal. TST was defined as the sum of all sleep stages (NREM and REM) expressed in hours, and AHI_TST was calculated as the total number of apneas and hypopneas divided by TST.

Electroencephalography arousals were scored according to AASM guidelines. The diagnosis of OSA was made according to the International Classification of Sleep Disorders 2 [15]. T90 (time with desaturation below 90%) was defined as the cumulative percentage of total sleep time spent with oxygen saturation below 90%, as measured by pulse oximetry.

STATISTICAL ANALYSIS:

R Studio software (version 2023.12.1) was used for data analysis. We used the survival R package for Kaplan-Meier curves and Cox models. We used Cox regression analysis to identify prognostic factors and chose those with P-values <0.05 for the analysis. We calculated the hazard ratio (HR) with its 95% confidence interval (CI) for each predictor variable using Cox proportional hazard regression. Data distribution was tested with the Shapiro-Wilk test. Results were presented as mean (±standard deviation) and median (min-max). To compare categorical data, exact and chi-square tests were used. Additionally, we developed proportional hazard regression models to calculate relative HRs for all-cause mortality and OSA-related mortality. We performed dynamic analysis of hazard regression models in the study time frame to evaluate 5-, 10-, and 15-year survival using the log rank test and Kaplan-Meier survival curves. We performed multivariate modeling, in univariate analysis and in intergroup comparisons, to determine risk factors for 5-, 10-, and 15-year mortality and overall mortality in multivariate Cox regression analysis. All assessed clinical variables and HRs in the mild, moderate, and severe OSA groups were referenced against a control group of patients in whom OSA was excluded. Detailed data are presented in Table 1 and Figures 2 and 3. To perform multivariate modeling, we separated all study variables into 3 groups: 1) clinical signs and symptoms, weight, height, BMI, and blood pressure; 2) comorbidities and medications; and 3) polysomnographic variables.

This study was conducted in compliance with the amended Declaration of Helsinki, and the Ethics Committee of the Medical University of Łódź, Poland approved the study protocol (RNN/23/15/KE; RNN/393/19/KE). The study was neither funded by an institutional grant nor pharmaceutical industry or medical company. This manuscript was developed with the support of a medical research agency as part of an educational program: Polish Clinical Scholars Research Training, managed by Harvard Medical School Postgraduate Medical Education, Boston, Massachusetts. The manuscript is the authors’ work and is not affiliated with Harvard Medical School or Medical Research Agency.

Results

A total of 4644 patients met the inclusion criteria (3157 men, 72.1%). Of these, 474 patients were excluded from the analysis: 220 patients had a total sleep time below 150 minutes, 292 patients had crucial variables of interest missing, 74 patients were excluded due to time of observation over 15 years, 35 patients were excluded due to BMI below 18 or over 60 kg/m2, or age below 18 (Figure 1). A final study group of 4023 patients was used for the statistical analysis. Among these 4023 patients, 982 (24.4%) did not meet the diagnostic criteria for OSA (AHI <5) and served as the control group. During 15 years of observation, 3170 patients (78.80%) remained alive and 853 (21.20%) died; clinical characteristics of the group are listed in Table 1. The 5-year survival rate was 89.7%, 10-year survival rate was 81.9%, and 15-year survival rate was 78.8%. The Kaplan-Meier survival curves for OSA-related mortality and all-cause mortality over the 15-year follow-up period illustrate significant differences (P<0.001) in survival probabilities based on severity of the OSA (Figures 2, 3), and the Kaplan-Meier curves demonstrated a clear dose-response relationship between OSA severity and survival probability. Patients with severe OSA had consistently lower survival over 5, 10, and 15 years, whereas patients with mild OSA showed survival patterns similar to those without OSA, emphasizing the prognostic importance of OSA severity.

The predominant causes of death over the 15 years of follow-up were cardiovascular (284, 33.3%), cancer (226, 26.5%), and pulmonary (102, 12%) (Figure 1). For OSA-related deaths, ie the cause of death was associated with OSA based on the ICD-10, the main causes were chronic heart failure I.50 (97, 21.1%), myocardial infarction I.21 – I.25 (72, 15.6%), stroke/intracerebral hemorrhage I.60 – I.64 (62, 13.5%), and chronic obstructive pulmonary disease J.44 (52, 11.3%) (Figure 1, Table 2).

The prevalence of OSA in the cohort, defined as AHI ≥5 and accompanied by typical symptoms, was high – 75.6%. This included 929 patients with mild OSA (23.1%), 770 with moderate OSA (19.1%), and 1343 with severe OSA (33.4%).

The results of the Cox regression analyses revealed factors associated with both all-cause and OSA-related mortality over 3 time periods (up to 5, 10, and 15 years). Univariate analysis showed significant associations with both all-cause and OSA-related mortality across all timeframes among the study variables. Some of the variables had no impact on mortality risk among the study groups. These included clinical symptoms (snoring, morning headaches) and polysomnography findings (AHI<5) (Table 3).

In the multivariate analyses among clinical signs and symptoms, sleepiness assessed by ESS was associated significantly with all-cause mortality, (0–5 years, HR 1.28, 95% CI: 1.16–1.42; 0–10 years, HR 1.20, 95% CI: 1.11–1.30; 0–15 years, HR 1.17, 95% CI: 1.09–1.26) and OSA-related mortality (0–5 years, HR 1.38, 95% CI: 1.22–1.57; 0–10 years, HR 1.23, 95% CI: 1.11–1.37; 0–15 years, HR 1.21, 95% CI: 1.10–1.33) (Table 4). With regard to comorbidities, episodes of stroke were associated with all-cause mortality (0–5 years, HR 2.41, 95% CI: 1.66–3.49; 0–10 years, HR 1.53, 95% CI: 1.17–1.99; 0–15 years, HR 1.77, 95% CI: 1.38–2.28) and OSA-related mortality (0–5 years, HR 2.84, 95% CI: 1.77–4.57; 0–10 years, HR 1.73, 95% CI: 1.24–2.41; 0–15 years, HR 2.04, 95% CI: 1.49–2.80). With regard to medications, use of new oral anticoagulants was associated with all-cause mortality (0–5 years, HR 2.56, 95% CI: 1.67–3.94; 0–10 years, HR 2.07, 95% CI: 1.44–2.98; 0–15 years, HR 1.71, 95% CI: 1.21–2.43) and OSA-related mortality (0–5 years, HR 3.20, 95% CI: 1.91–5.537; 0–10 years, HR 2.63, 95% CI: 1.70–4.07; 0–15 years, HR 2.08, 95% CI: 1.36–3.17). OSA severity was also associated with mortality in the third analyzed group of polysomnographic findings. AHIREM was associated with significantly higher all-cause mortality (0–5 years, HR 1.38, 95% CI: 1.18–1.62; 0–10 years, HR 1.26, 95% CI: 1.12–1.41; 0–15 years, HR 1.23, 95% CI: 1.11–1.37) and OSA-related mortality (0–5 years, HR 1.42, 95% CI: 1.16–1.74; 0–10 years, HR 1.31, 95% CI: 1.12–1.52; 0–15 years, HR 1.26, 95% CI: 1.10–1.44), but there was no significant association between AHINREM and mortality (Table 3, Figure 4).

The dynamic curve of AHIREM HR with 95% CI was related to time with desaturation below 90%, while that of AHINREM was related to minimal SpO2. The AHITST curve was associated with average SpO2 in all-cause mortality and OSA-related mortality groups. A similar analysis was performed to evaluate dynamic HR with respiratory events: obstructive, mixed, and central apneas and hypopnea during NREM and REM sleep.

Discussion

DIFFERENCES BETWEEN ALL-CAUSE MORTALITY AND OSA-RELATED MORTALITY:

The role of excessive daytime sleepiness, as a predictor of OSA-related and all-cause mortality underscores the significance of sleep fragmentation and daytime symptoms in OSA prognosis. Similar findings have been reported by Randerath et al [19], who demonstrated a link between sleepiness and cardiovascular outcomes in OSA patients. Excessive daytime sleepiness is considered as a crucial symptom of OSA diagnosis that is associated with higher mortality independently of OSA severity and therapy [20,21]. Insulin use, as a marker of diabetes severity, was strongly associated with OSA-related mortality. This aligns with evidence showing that OSA exacerbates glucose intolerance and insulin resistance, further compounding the risk of adverse outcomes [22]. The protective association of depression against all-cause mortality may reflect increased healthcare engagement among patients diagnosed with depression, as suggested in other studies [23]. However, this finding warrants further exploration to rule out residual confounding factors.

The highest short-term mortality risks were associated with OSA severity, use of Vitamin K antagonists, and strokes. This likely reflects the high baseline cardiovascular risk of anticoagulated patients, and aligns with prior findings that suggest a strong association between thromboembolic events, OSA, and mortality [6,24] The association of hyperthyroidism with OSA-related mortality is novel and warrants further investigation. It may reflect the impact of thyroid hormone dysregulation on cardiac function, as previously described [25]. The observed association between hypertension and OSA mortality highlights the complex role of blood pressure in OSA patients. Recent studies have proposed that isolated systolic hypertension may be a more significant predictor of adverse outcomes in OSA patients [9,21,26]. The protective association of statins with all-cause mortality underscores the importance of lipid-lowering therapy in OSA patients, given the known links between dyslipidemia, systemic inflammation, and cardiovascular risk [16,27]. Over the 15-year follow-up period, the persistence of diabetes, systolic blood pressure, and new oral anticoagulants as predictors of all-cause mortality emphasizes the importance of addressing metabolic and thrombotic risk in long-term management. These findings are consistent with longitudinal studies showing that diabetes amplifies the cardiovascular burden of OSA [28]. The role of hyperthyroidism as the sole predictor of OSA-related mortality over this period highlights the need for ongoing endocrine evaluation in OSA patients, especially given the potential cardiovascular effects of untreated hyperthyroidism [25].

POLYSOMNOGRAPHIC FINDINGS:

The association of AHIREM with mortality in early follow-up (0–5 years) and AHINREM with mortality in longer-term follow-up suggests that apneas occurring in different sleep stages may have distinct physiological impacts. This aligns with evidence that REM apneas are associated with greater hypoxemia and sympathetic activation, whereas NREM apneas contribute to chronic volume overload and cardiovascular strain [20,29]. The persistence of AHITST as a predictor of mortality at 0–15 years underscores the cumulative impact of overall apnea burden on long-term outcomes. These findings highlight the utility of polysomnography not only for diagnosis but also for prognostic stratification in OSA patients.

The survival curves demonstrate a clear dose-response relationship between OSA severity and both OSA-related and all-cause mortality. Patients with severe OSA consistently exhibited the worst outcomes, a finding which aligns with prior research linking untreated or poorly managed OSA to cardiovascular disease, systemic inflammation, and metabolic dysfunction [18]. Patients with mild OSA had survival probabilities similar to those without OSA, suggesting that mild OSA may have a relatively limited impact on long-term mortality. Moderate OSA, however, marked the transition to more significant mortality risks, emphasizing the importance of early intervention to prevent progression to severe OSA [14]. The patterns for OSA-related mortality were similar to those for all-cause mortality, indicating that OSA’s systemic effects contribute broadly to both direct (eg, respiratory failure) and indirect (eg, cardiovascular disease) mortality risks. This underscores the need to address comorbid conditions, such as hypertension, diabetes, and dyslipidemia, in OSA management. Our analysis, in contrast to the work of Azarian et al, focused on the impact of comorbidities, medications, and medical history rather than age as an independent prognostic factor. The results of our study provide a broader perspective on the complexity of OSA [13].

LIMITATIONS:

Despite the robust findings, this study has certain limitations. Being a single-center cohort study, it is subject to inherent biases related to data collection and patient selection. The single-center nature of the study may also limit the generalizability of the results. Future research should aim to validate these findings in larger, multi-center cohorts and explore the integration of other emerging biomarkers and technological advancements in the assessment of OSA.

Further research should focus on:

Conclusions

Our study confirms the necessity of a multidisciplinary approach for patients admitted to sleep centers with different sleep disorders. The association between depression and an increased risk of suicide highlights the need to pay particular attention to sleep disorders beyond OSA. Moreover, it underscores the importance of establishing dedicated sleep research laboratories, staffed by qualified professionals specializing in sleep disorders, which is not yet a standard in all countries, including Poland.

The study has several important clinical implications: