Abstract
Objective
To investigate the associations of blood inflammatory biomarkers with all-cause and cardiovascular disease (CVD) mortality in individuals with self-reported obstructive sleep apnea (OSA) symptoms.
Methods
This retrospective cohort study included 10,230 adults aged ≥18 years with self-reported OSA symptoms from the NHANES database.Participants were followed from baseline through December 31, 2019. Kaplan-Meier analysis, multivariable Cox proportional hazards models, restricted cubic spline (RCS), segmented regression and sensitivity analyses were employed to evaluate the associations of inflammatory biomarkers, including red blood cell distribution width (RDW), RDW-to-albumin ratio (RAR), neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), and monocyte-to-lymphocyte ratio (MLR) with all-cause and CVD mortality.
Results
Kaplan-Meier analysis showed that survival rates were significantly lower among individuals with self-reported OSA symptoms who had the highest levels of RDW, RAR, NLR, and MLR. In the fully adjusted Cox model, the highest quartile (Q4) of these biomarkers was associated with significantly increased risks of all-cause mortality compared with the lowest quartile (Q1): RDW (HR=3.70; 95% CI:2.14-6.42), RAR (HR=2.93; 95% CI:2.20-3.91), NLR (HR=1.65; 95% CI: 1.32-2.07), and MLR (HR=1.68; 95% CI:1.22-2.31). For CVD mortality, the corresponding HRs (Q4 vs. Q1) were: RDW (HR=3.07; 95% CI:1.61-5.85), RAR (HR=2.99; 95% CI:1.57-5.68), NLR (HR=2.66; 95% CI:1.51-4.68), and MLR (HR=1.87; 95% CI:1.07-3.29). RCS model demonstrated that there was a nonlinear association between RDW, RAR, NLR and the both mortality endpoints, while MLR did not show a significant nonlinear relationship with mortality. Segmented regression further identified data-driven statistical thresholds. These exploratory, data-driven thresholds have not been clinically validated and should not be directly applied to clinical decision-making. Sensitivity analyses yielded consistent results.
Conclusions
Blood inflammatory biomarkers (RDW, RAR, NLR, MLR) are significantly associated with all-cause and CVD mortality in individuals with self-reported OSA symptoms. Given the observational design, these biomarkers should be regarded as associative rather than prognostic, pending future confirmation.
1. Introduction
Obstructive sleep apnea (OSA) is a common sleep-disordered breathing condition characterized by recurrent upper airway obstruction during sleep, which leads to chronic intermittent hypoxia and sleep fragmentation. 1 Epidemiological studies have reported that the prevalence of OSA among adults ranges from 9% to 38%, 2 with an estimated 936 million adults aged 30-69 years affected by mild-to-severe OSA worldwide. 3 The prevalence of this condition continues to rise with advancing age and the escalating global obesity epidemic. 4 Accumulating evidence has demonstrated that OSA significantly increases the risk of various comorbidities, particularly cardiovascular diseases, including hypertension, coronary heart disease, atrial fibrillation, heart failure, diabetes mellitus, and stroke. Furthermore, OSA is associated with an elevated risk of mortality attributable to these complications.5–8 Despite the substantial mortality and cardiovascular burden associated with OSA, simple, convenient, and widely applicable tools for clinical risk stratification remain lacking, representing a critical unmet need in current OSA management. Therefore, identifying and screening OSA patients at high risk of mortality is essential for preventing adverse health outcomes.
In recent years, the inflammatory response has been increasingly recognized as a core mechanism underlying the pathogenesis and progression of OSA. 9 Both low-grade systemic inflammation and a hypercoagulable state are closely linked to OSA and its associated cardiovascular diseases. 10 Several novel immune-inflammatory markers derived from complete blood cell counts have been shown to effectively reflect the balance between inflammatory and immune responses.11–13 Notably, markers such as the neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), monocyte-to-lymphocyte ratio (MLR), and red blood cell distribution width (RDW) have been utilized to assess systemic inflammation, disease severity, and cardiovascular risk in patients with OSA.14–17 As biomarkers derived from routine blood tests, these indicators are cost-effective and readily accessible, potentially addressing the aforementioned unmet need for OSA risk stratification. The red blood cell distribution width-to-albumin ratio (RAR), a composite indicator reflecting both inflammatory status and nutritional status, may also serve as a potential prognostic marker warranting further investigation.18–20
However, the associations between these inflammatory markers and mortality in individuals with self-reported OSA symptoms have not been fully elucidated. Utilizing data from the National Health and Nutrition Examination Survey (NHANES), the present study aims to investigate the relationships between blood inflammatory markers and mortality trends among individuals with self-reported OSA symptoms. We sought to identify potential inflammatory biomarkers associated with mortality risk, thereby contributing to the evolving understanding of inflammatory profiles in this population.
2. Methods
2.1. Study design and population
The data of this study were derived from the NHANES database in the United States, which employs a multistage stratified cluster random sampling design to enroll nationally representative participants. This database has been approved for use by the National Health Statistics Research Ethics Review Committee, and all participants gave informed consent. Research design and data about NHANES details, please see the website: https://www.cdc.gov/nchs/nhanes/. This study was conducted in accordance with the Helsinki Declaration of 1975, as revised in 2024. This retrospective cohort analysis used publicly available, de-identified NHANES data with mortality follow-up; therefore, no additional ethical review was required. The reporting of this study conforms to STROBE guidelines. 21
We combined survey data from NHANES 2005–2008 and 2015–2018 cycles. No substantial temporal changes were identified in the diagnostic criteria, case definitions, or clinical management of OSA-related symptoms between these periods, and the assessment of self-reported OSA symptoms remained consistent across cycles. Pooling these cycles was conducted to increase the sample size and improve the statistical power and representativeness of the study. A total of 39,722 participants from four investigation periods of NHANES (2005-2008 and 2015-2018) were initially included. The exclusion criteria are as follows: (1) Not meeting the diagnostic criteria for OSA-related symptoms; (2) Under the age of 18; (3) No follow-up data; (4) Any of the RDW, RAR, NLR, PLR, or MLR indicator data is missing. After screening, a total of 10,230 participants were finally included in the analysis (Figure 1). Participant selection and study flow diagram.
2.2. Diagnosis of self-reported OSA/OSA-related symptoms
The diagnostic basis for self-reported OSA symptoms is the participants’ affirmative response to at least one of the following three NHANES questions related to OSA-related symptoms 22 : (1) Despite sleeping for no less than 7 hours each night, excessive daytime sleepiness still occurs (reported at a frequency of 16-30 times) per month; (2) At least 3 episodes of snorting, gasping, or stopping breathing occur per week; (3) Snoring ≥3 times a week.
2.3. Definitions of inflammatory markers
Blood samples were collected and processed by qualified blood collectors in accordance with NHANES standard procedures. The calculation method of inflammatory markers is as follows: The RDW percentage was detected by the Coulter analyzer in the mobile physical examination center for peripheral blood samples; RAR is the ratio of RDW to serum albumin concentration; NLR is derived by dividing the neutrophil count by the lymphocyte count. PLR is the ratio of platelet count to lymphocyte count. MLR is calculated by dividing the monocyte count by the lymphocyte count.
2.4. Mortality assessment
Death information is obtained through the Associated National Death Index (NDI), a database that provides detailed data on causes of death. The follow-up period was calculated from the baseline interview date and ended on the date of death or December 31, 2019. The primary outcomes of this study were all-cause mortality and cardiovascular disease mortality. All-cause mortality encompasses all causes of death, while cardiovascular disease mortality specifically refers to death resulting from heart disease (I00-I09, I11, I13, I20-I51) or cerebrovascular disease (I60-I69).
2.5. Assessment of covariates
The study collected baseline data of the participants through questionnaires and laboratory tests, including age, sex, marital status, race, education level and body mass index (BMI, divided into four categories: <18.5, 18.5-25, 25-30 and ≥30). The socio-economic status is evaluated by the poverty income ratio (PIR), which is classified into three categories: ≤1.3, 1.3-3.5, and >3.5. Smoking status can be classified into three categories: never smokers, former smoking, and current smoking. 23 The metabolic equivalent (MET) score is calculated based on activity type/intensity, average duration and frequency to obtain the 30-day MET-minute value corresponding to each activity. According to the national guidelines, 24 participants were divided into the low physical activity group (<500 MET/week) and the high physical activity group (≥500 MET/week) before the analysis. Hypertension was defined as self-report or a measured blood pressure of ≥140/90 mmHg. Diabetes was diagnosed based on self-report, the use of insulin or glucose-lowering medications, fasting blood glucose levels of ≥7 mmol/L, glycated hemoglobin A1c levels of ≥6.5%, or 2-hour post-load glucose levels of ≥11.1 mmol/L. 25 Hyperlipidemia was identified in participants who were either taking lipid-lowering medications or had triglyceride levels of ≥1.7 mmol/L or total cholesterol levels of ≥5.2 mmol/L, low-density lipoprotein cholesterol (LDL-C) levels of ≥3.4 mmol/L and/or high-density lipoprotein cholesterol (HDL-C) levels of ≤1.0 mmol/L for males or ≤1.3 mmol/L for females were considered. 26 The presence of CVD was assessed through self-reported histories of congestive heart failure, coronary heart disease, angina, myocardial infarction, or stroke. Cancer diagnosis data were also self-reported: participants were queried about whether a physician or other healthcare professional had ever informed them of a cancer or malignancy diagnosis, and affirmative responses were classified as indicative of cancer. 27
2.6. Declaration of generative AI and AI-assisted technologies in the writing process
During the preparation of this work the authors used ChatGPT to improve readability and language. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.
2.7. Statistical analysis
To accurately represent the representativeness between the selected samples and the actual population, and to mitigate the effects of missing samples, oversampling, and differences in sample selection on the overall analysis, we employed a complex sampling analysis method, and applied weights to the samples. Missing quantitative data were imputed using predictive mean matching (PMM), while missing categorical data were handled using logistic regression. Continuous variables are represented by the mean and its standard error (SE), while categorical variables are described by frequency and composition ratio. The t-test was used for the comparison between groups of continuous variables, and the chi-square test was used for the comparison between groups of categorical variables.
Participants were divided into four groups (Q1-Q4) according to quartiles of each inflammatory indicator, with Q1 serving as the reference group. The mortality outcomes were analyzed using the Kaplan-Meier curve and log-rank test, and the association between systemic inflammatory indicators and the risk of mortality was estimated using the Cox proportional hazards model. All Cox proportional hazards models accounted for the complex NHANES survey design, including sampling weights, strata, and primary sampling units.
Three models were constructed in the research: Model 1 was uncorrected; Model 2 adjusted age, sex and race; Model 3 further adjusted for marital status, educational level, poverty income ratio, smoking status, body mass index, physical activity, hypertension, diabetes, hyperlipidemia, cardiovascular disease and cancer. The analysis of mortality from cardiovascular diseases was conducted using the same method. To explore the possible nonlinear relationship between death and inflammatory biomarkers in participants with self-reported OSA symptoms, RCS were used for analysis. Segmented models were constructed using the R segmented software package, which estimates breakpoints and fits separate linear models for each segment defined by these breakpoints.We conducted sensitivity analyses to compare imputed and raw dataset results, repeating the main analyses with raw data. To assess whether the association between inflammatory biomarkers and mortality varied according to the specificity of self-reported OSA symptoms, and to evaluate the robustness of our findings against potential symptom misclassification, we performed subgroup analyses in two mutually exclusive groups: (1) a more OSA-specific subgroup (participants with witnessed breathing disturbances, i.e., snorting, gasping, or stopping breathing during sleep), and (2) a less specific subgroup (participants with snoring only). To assess whether the association was confounded by underlying respiratory diseases, we performed a sensitivity analysis excluding participants with self-reported asthma or COPD. Data processing and analysis utilized R version 4.4.0.
3. Results
3.1. Baseline characteristics of the participants
Throughout the follow-up period involving 10,230 participants with self-reported OSA symptoms, there were 1,026 deaths from various causes, with 287
Characteristics of individuals with self-reported OSA symptoms stratified by all-cause and cardiovascular mortality status.
CVD, cardiovascular disease; BMI,body mass index; RDW, red cell distribution width; RAR,red cell distribution width-to-albumin ratio; NLR, neutrophil-to-lymphocyte ratio; PLR,platelet-to-lymphocyte ratio; MLR, monocyte-to-lymphocyte ratio. The values are presented as the means [standard errors (SEs)] or n (%). Bold values indicate statistical significance. A p value <0.05 indicated a significant difference.
3.2. Association between hematologic inflammatory markers and mortality
Kaplan-Meier curves demonstrated significant survival differences across quartiles of RDW, RAR, NLR, and MLR, with a log-rank p-value of less than 0.001 (Figure 2, Figure 3). Mortality rates were highest in the fourth quartile for these indices. K–M Survival curves for all-cause mortality among hematologic inflammatory markers. [(A) RDW, (B) RAR, (C) NLR, (D)PLR,(E)MLR]. K–M Survival curves for cardiovascular mortality among hematologic inflammatory markers. [(A) RDW, (B) RAR, (C) NLR, (D)PLR,(E)MLR.
Cox regression models for the associations between hematologic inflammatory markers and all-cause mortality in individuals with self-reported OSA symptoms.
HR: Hazard Ratio, CI: Confidence Interval.
Q1–Q4 represent quartiles of the respective biomarker, with Q1 (the lowest quartile of the biomarker) as the reference category.
Bold values indicate statistical significance; a p value <0.05 indicated a significant difference.
Model 1: No covariates were adjusted for.
Model 2: Adjusted for age, sex, race.
Model 3: Adjusted for age, sex, race, marital status, education level, PIR, smoking status, BMI, physical activity, hypertension, diabetes, hyperlipidemia, CVD and cancer.
Cox regression models for the associations between hematologic inflammatory markers and cardiovascular-disease mortality in individuals with self-reported OSA symptoms.
CVD:cardiovascular-disease, HR: Hazard Ratio, CI: Confidence Interval.
Q1–Q4 represent quartiles of the respective biomarker, with Q1 (the lowest quartile of the biomarker) as the reference category.
Bold values indicate statistical significance; a p value <0.05 indicated a significant difference.
Model 1: No covariates were adjusted for.
Model 2: Adjusted for age, sex, race.
Model 3: Adjusted for age, sex, race, marital status, education level, PIR, smoking status, BMI, physical activity, hypertension, diabetes, hyperlipidemia, and cancer.
3.4. Non-linear relationships between hematologic inflammatory markers and mortality
The RCS analysis (Figure 4, Figure 5) revealed threshold-based nonlinear associations between RDW, RAR, NLR and all-cause/CVD mortality. To further characterize these non-linear patterns, we performed segmented regression and threshold analysis. The identified breakpoints, along with the HRs for the associations below and above each breakpoint, are summarised in Table 4. Key quantitative findings are as follows: Restricted cubic spline (RCS) illustrating the correlation between indicators and all-cause mortality among participants with self-reported OSA symptoms. [(A) RDW, (B) RAR, (C) NLR, (D)MLR]. Restricted cubic spline (RCS) illustrating the correlation between indicators and cardiovascular mortality among participants with self-reported OSA symptoms. [(A) RDW, (B) RAR, (C) NLR, (D)MLR]. Threshold effect analysis of hematologic inflammatory markers on all-cause/CVD mortality in individuals with self-reported OSA symptoms. HR: Hazard Ratio, CI: Confidence Interval. Bold values indicate statistical significance; a p value <0.05 indicated a significant difference.
RDW: The breakpoint was 15.8 for both mortality endpoints. For all-cause mortality, each 1-unit increase in RDW below the breakpoint was associated with an HR of 1.37 (95% CI: 1.27-1.47, p<0.001); above the breakpoint, the HR was 1.04 (95% CI: 0.95-1.14, p=0.425). For CVD mortality, the HR below the breakpoint was 1.41 (95% CI: 1.31-1.52, p<0.001), and above the breakpoint it was 1.03 (95% CI: 0.94-1.13, p=0.512). The likelihood ratio test for a non-linear model was significant for both outcomes (p< 0.001), indicating a threshold effect. This suggests a nonlinear association between RDW and mortality, with risk increasing only up to a threshold of 15.8; beyond this point, further elevations in RDW do not confer additional risk.
RAR: The breakpoint was 4.2. For all-cause mortality, below the breakpoint HR =2.59 (95% CI: 2.14-3.14, p<0.001); above the breakpoint HR=1.48 (95% CI: 1.15-1.91, p=0.002). For CVD mortality, below breakpoint HR=2.95 (95% CI: 2.07-4.20, p<0.001); above breakpoint HR=0.35 (95% CI: 0.11-1.10, p=0.072). The likelihood ratio test was significant (p<0.001 for both).This suggests a pronounced threshold effect, where the steepest increase in mortality risk occurs below a RAR value of 4.2; above this breakpoint, the risk for all-cause mortality remains elevated but plateaus, while the association with CVD mortality loses statistical significance.
NLR: The breakpoint was 4.1. For all-cause mortality, below breakpoint HR= 1.22 (95% CI: 1.12-1.32, p<0.001); above breakpoint HR=1.04 (95% CI: 0.95-1.13, p=0.427). For CVD mortality, below breakpoint HR=1.44 (95% CI: 1.23-1.69, p<0.001); above breakpoint HR=1.02 (95% CI: 0.87-1.20, p=0.784). Likelihood ratio test p=0.002 (all-cause) and 0.003 (CVD). This indicates a threshold effect for NLR similar to RDW, wherein rising NLR values are associated with increased mortality risk only up to the breakpoint of 4.1, with no significant additional hazard observed beyond that level.
Notably, these statistically derived thresholds are exploratory in nature and should not be regarded as validated clinical cut-off values.
3.5. Sensitivity analyses for hematologic inflammatory markers and mortality
Sensitivity analysis was conducted using raw data. Similar associations between RDW, RAR, NLR and MLR and all-cause/CVD mortality were observed before imputation (Tables S1 and S2).
To assess the robustness of our findings across different definitions of self-reported OSA-related symptoms, we performed subgroup analyses in two mutually exclusive groups: snoring-only (less specific, n=7,862) and witnessed breathing disturbances (more specific, n=2,368). Supplementary tables S3–S6 present the fully adjusted hazard ratios for all-cause and CVD mortality, respectively. For all-cause mortality (Tables S3 and S5 in supplement, summarized below): In the snoring-only subgroup, the corresponding HRs (Q4 vs. Q1) were: RDW (HR=2.21, 95% CI: 1.70-2.88, p<0.001), RAR (HR=2.57, 95% CI: 2.03-3.25, p<0.001), NLR (HR=1.50, 95% CI: 1.25-1.79, p<0.001), MLR (HR=1.44, 95% CI: 1.11-1.86, p=0.005). In the witnessed breathing disturbances subgroup, the corresponding HRs (Q4 vs. Q1) were: RDW (HR=3.80, 95% CI: 2.13-6.77, p<0.001), RAR (HR=4.28, 95% CI: 2.60-7.05, p<0.001), NLR (HR=1.56, 95% CI: 1.05-2.31, p=0.028), MLR (HR=1.92, 95% CI: 1.27-2.91, p=0.002).
For CVD mortality (Tables S4 and S6 in supplement, summarised below): In the snoring-only subgroup, the corresponding HRs (Q4 vs. Q1) were: RDW (HR = 2.56, 95% CI: 1.58-4.16, p<0.001), RAR (HR=3.65, 95% CI: 2.27-5.87, p<0.001), NLR (HR=2.64, 95% CI: 1.77–3.94, p<0.001), MLR (HR=1.69, 95% CI: 1.20-2.39, p=0.003). In the witnessed breathing disturbances subgroup, the corresponding HRs (Q4 vs. Q1) were: RDW (HR=3.65, 95% CI: 1.20-11.09, p= 0.022), RAR (HR=3.18, 95% CI: 1.40-7.21, p=0.006), MLR (HR=5.30, 95% CI: 2.71-10.33, p<0.001). NLR showed a similar direction but did not reach statistical significance (HR=2.67, 95% CI: 0.96-7.42, p=0.060).
To assess whether the association was confounded by underlying respiratory diseases, we performed a sensitivity analysis excluding participants with self-reported asthma or COPD. The results remained consistent with the primary analysis (data shown in Tables S7 and S8), indicating robustness of our findings.
4. Discussion
This study systematically evaluated the associations of multiple circulating inflammatory markers in individuals with self-reported OSA symptoms and characterized their nonlinear associations with mortality risk. After full adjustment for potential confounders, the highest quartiles of RDW, RAR, NLR, and MLR but not PLR were independently associated with significantly elevated risks of all-cause and CVD mortality. Restricted cubic spline analyses further confirmed nonlinear relationships of RDW, RAR, and NLR with both mortality endpoints, whereas no significant nonlinear association was detected for MLR. The inflection points identified in threshold analysis represent data-driven statistical estimates specific to this cohort rather than clinically validated cut-offs; accordingly, these findings should be regarded as exploratory and associative, and require external validation in future research. Without such validation, these biomarkers should not be interpreted as clinically actionable prognostic tools.
RDW is a routinely available hematological parameter that reflects erythropoietic function, inflammatory status, oxidative stress, and nutritional metabolism. During chronic inflammation, proinflammatory cytokines may elevate RDW by disrupting erythroid differentiation and iron metabolism. 28 Oxidative stress can damage erythrocyte membranes and perturb the hematopoietic microenvironment, both of which may contribute to increased heterogeneity in red blood cell size. 29 Nutritional imbalance may also directly impair erythrocyte maturation through insufficient supply of essential hematopoietic substrates. 30 Together, these mechanisms provide a biological basis for the association between RDW and mortality across diverse diseases. Nonetheless, the relatively strong associations observed in the present study warrant cautious interpretation. Elevated RDW may reflect overall systemic disease burden, chronic comorbidities, malignancy, nutritional deficiencies, or other nonspecific inflammatory states, rather than OSA-specific inflammatory processes alone. Although comprehensive confounder adjustment was applied, residual confounding from unmeasured factors cannot be entirely excluded. Furthermore, reverse causality remains possible, as elevated RDW may be a consequence of preexisting systemic disorders rather than a direct result of OSA pathogenesis. RAR is a composite biomarker that integrates RDW and albumin, enabling simultaneous assessment of inflammatory and nutritional status. Since both components may decrease in malnutrition or catabolic states, RAR has emerged as a potential prognostic indicator in various disease settings.31–33 Elevated RAR levels may signify systemic disturbances in inflammation, oxygenation, and nutrition pathophysiological processes particularly relevant to OSA.
NLR is a well-established marker of systemic inflammation that reflects the balance between neutrophil-driven innate immune activation and lymphocyte-mediated adaptive immune regulation. Elevated NLR is closely linked to oxidative stress, endothelial dysfunction, and hypoxia-induced inflammation, which are core components of OSA pathophysiology. 34 Peripheral blood monocyte counts are significantly elevated in patients with OSA and have been recognized as inflammatory biomarkers. 35 Enhanced inflammatory reactivity of monocytes and macrophages may contribute to chronic inflammatory disorders, and the association of MLR with mortality may reflect pathological processes related to monocyte/macrophage activation and lymphocyte exhaustion. The inflection points identified for these biomarkers are data-driven estimates derived from this specific cohort and should be interpreted with caution. Their divergence from conventional reference ranges may simply reflect population-specific variation rather than clinically meaningful cut-offs. These exploratory observations are offered primarily as hypothesis-generating information and are not positioned as validated thresholds for risk stratification or clinical decision-making. In contrast, PLR did not show independent predictive value in this study. This may be attributable to the susceptibility of platelet counts to numerous confounding factors, including medications and comorbidities. Within the specific pathophysiological context of OSA, this association may be obscured by more dominant inflammatory and metabolic pathways.
Collectively, these biomarkers (RDW, RAR, NLR, MLR) may reflect chronic low-grade systemic inflammation, increased oxidative stress, immune dysregulation, and nutritional-metabolic disturbances associated with OSA. Intermittent hypoxia-reoxygenation, a hallmark pathological feature of OSA, activates inflammatory pathways such as NF-κB and impairs bone marrow function, hepatic albumin synthesis, and immune cell homeostasis. 36 Increased levels of these inflammatory markers may exacerbate endothelial dysfunction, atherosclerosis, and insulin resistance, thereby increasing cardiovascular and mortality risk.37–39 RAR particularly underscores the potential synergistic role of malnutrition and hypercatabolism in this process.The identified nonlinear relationships suggest these biomarkers may reflect exploratory population-level statistical trends, and these inflection points represent hypothesis-generating signals only, not clinically actionable or applicable targets. As this was an observational study, all findings demonstrate associations rather than causal relationships, and reverse causality cannot be excluded.
To verify the robustness of our findings, we performed sensitivity analyses stratified by self-reported OSA symptom specificity. In the snoring-only subgroup, associations were generally consistent with the main analysis. In the more OSA-specific subgroup (participants with witnessed breathing pauses), elevated RDW, RAR, and MLR remained significantly associated with increased risks of both all-cause and CVD mortality. The association between NLR and CVD mortality was no longer statistically significant in this smaller subgroup, likely due to reduced statistical power rather than a genuine lack of association. Furthermore,to assess whether the association was confounded by underlying respiratory diseases, we performed a sensitivity analysis excluding participants with self-reported asthma or COPD. The results remained consistent with the primary analysis. Overall, consistency across different symptom definitions supports the reliability of our main findings.
This study is based on the large-scale national cohort of NHANES, with strong sample representativeness and reliable mortality data. It also employs multiple methods to enhance robustness of the results, including complex weighting, multi-model correction, and nonlinear analysis. However, the research also has certain limitations. Firstly, there may be a classification bias in the definition of OSA. In this study, the OSA status was determined based on the self-reported questionnaire data from the NHANES, rather than objective diagnostic tools such as polysomnography or home sleep apnea tests. The definition based on the questionnaire may lead to a serious misclassification of OSA and exclude the assessment of severity indicators of the disease, such as the apnea-hypopnea index and the decline in nocturnal blood oxygen saturation.This non-differential misclassification of exposure factors is highly likely to bias the observed association towards the null, consequently attenuating the magnitude of the relationship among OSA, inflammatory burden, and mortality. In other words, the true associations may be stronger than those estimated in the present study. Although misclassification may theoretically introduce residual confounders, the main effect of using self-reported rather than objectively OSA is expected to be attenuation of effect estimates rather than spurious exaggerated associations. Given that the severity of OSA is closely related to the burden of systemic inflammation and the risk of death, this bias may weaken the observed association and limit the precision of the effect size. Therefore, this limitation should be carefully considered for the validity and potential impact on the interpretation of our research results. Future studies that objectively diagnose OSA are needed to verify our findings.Secondly, the NHANES database does not collect information on specific treatments for OSA, including continuous positive airway pressure (CPAP), oral appliances, surgical interventions, and other therapies. Therefore, we are unable to adjust for or analyze the effects of these treatments in this study. As the gold standard treatment for OSA, CPAP can effectively alleviate chronic intermittent hypoxia, reduce systemic inflammation, and improve cardiovascular outcomes and mortality risk. The lack of treatment data may result in residual confounding, as differences in the treatment status of obstructive sleep apnea could simultaneously affect levels of inflammatory biomarkers and mortality outcomes. This limitation prevents us from evaluating the potential moderating effect of OSA treatment on the observed associations, and should be taken into account when interpreting the results. Thirdly, the study population was limited to U.S. adults from the NHANES database. Therefore, the generalizability of the findings to other ethnicities, geographic regions, and sociodemographic populations is restricted, and extreme caution should be made when extrapolating the present conclusions to non-U.S. populations or other ethnic/geographic groups. Despite extensive multivariable adjustment, the possibility of residual confounding cannot be entirely excluded. Fourth, the relatively large hazard ratios observed, particularly for RDW, might partly reflect unmeasured or imprecisely measured factors such as the severity of underlying chronic diseases, frailty, or other socioeconomic and behavioral determinants of health that are associated with both elevated inflammatory markers and mortality. Future prospective studies with objectively diagnosed OSA are warranted to validate these findings and explore the potential impact of OSA treatments on inflammatory biomarkers and mortality.
5. Conclusion
In conclusion, this population-based analysis demonstrates that blood inflammatory biomarkers (RDW, RAR, NLR, and MLR) are associated with all-cause and CVD mortality in individuals with self-reported OSA symptoms.These associative and exploratory findings are not intended for direct clinical application, risk stratification, or patient-level decision-making. Our results may provide supportive population-level evidence to inform future investigations of inflammatory profiles in OSA.
Supplemental material
Supplemental material - Association between blood inflammatory biomarkers and mortality in individuals with self-reported obstructive sleep apnea symptoms: A nationwide population-based retrospective cohort study
Supplemental material for Association between blood inflammatory biomarkers and mortality in individuals with self-reported obstructive sleep apnea symptoms: A nationwide population-based retrospective cohort study by Anyuan Zhong, Yongjian Pei, Zengli Zhang and Rui Chen in Science Progress.
Footnotes
Acknowledgments
We extend our sincere gratitude to all individuals who contributed to and participated in the National Health and Nutrition Examination Survey.We also acknowledge the use of ChatGPT for language improvement during manuscript preparation, and all authors take full responsibility for the final content.
Ethical considerations
Our institution does not require ethical approval for secondary analysis of publicly available, de-identified data.
Consent for publication
Informed consent for patient information to be published in this article was not obtained because the data are derived from a publicly available, de-identified database (NHANES) for which informed consent was obtained at the time of original data collection.
Author contributions
Anyuan Zhong and Yongjian Pei are designated as co-first authors of this study. AZ, and RC designed the experiments. AZ, and YP analyzed the data and wrote the manuscript. ZZ and YP provided helpful discussion and reviewed the manuscript. AZ, and RC reviewed the results and revised the manuscript. All authors read and approved the fnal manuscript.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Science and Education Strengthening HealthProject of Suzhou, China (grant no. QNXM2024019) and the National Natural Science Foundation of China (NSFC, grant no. 82070095, 82470094).
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
This study employed publicly accessible datasets for analysis. The original contributions to the research are detailed within the article and supplementary materials; for further inquiries, please contact the corresponding author.
Supplemental material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
