Figure 1 depicts a flowchart illustrating the study design and process. To ensure the validity of the assessment, the exclusion criteria were: age <18 years; ongoing OSA therapy; current use of anxiolytics, antidepressants, hypnotics, or antipsychotic drugs.

From January 2020 to May 2023, a total of 2,972 subjects were initially recruited. Among them, 147 subjects were excluded from formal enrollment due to age <18 years (n = 44), received OSA treatment (n = 50) or medication of anxiolytics, antidepressants, hypnotics, or antipsychotic drugs (n = 53). After undergoing overnight sleep apnea monitor, 600 subjects received laboratory tests, Echocardiography (ECHO) and Holter electrocardiography monitoring. Finally, this retrospective observational study included 600 subjects with clinical suspicion of OSA in Renmin Hospital of Wuhan University.

From the electronic medical records, we collected essential clinical information, including demographic and anthropometric data (such as sex, age, height, weight, systolic blood pressure, and diastolic blood pressure), as well as details related to cardiovascular risk factors (such as smoking, drinking, and diabetes) and cardiovascular diseases (such as hypertension, coronary heart disease, heart failure, and myocardial infarction).

All enrolled subjects signed written informed consent. The study has been approved by the ethics committee of Renmin Hospital of Wuhan University. All studies have been conducted according to the ethics guidelines laid down in the Helsinki Declaration.

All subjects had undergone fasting blood tests including blood glucose, low-density lipoprotein cholesterol (LDL), high-density lipoprotein cholesterol (HDL), triglyceride (TG), total cholesterol (TC), brain natriuretic peptide (BNP), Creatinine (Cr), K, Na, CI, Ca, Activated Partial Prothrombin Time (APPT), Prothrombin Time (PT), Fibrinogen and D-Dimer.

HRV measurement was used to assess cardiac autonomic dysfunction. We analyzed the frequency and time domain of HRV by using Holter electrocardiography monitoring (BI9800, Biomedical Instruments Co., Ltd., Shenzhen, Guangdong, China). The time domain analysis includes parameters such as: standard deviation of normal-to-normal intervals (SDNN); standard deviation of 5 min average NN intervals (SDNN Index); standard deviation of 5 min average NN intervals (SDANN); root mean square successive difference (RMSSD); percentage of adjacent NN intervals (PNN50); and trigonometric index (11). The frequency analysis includes parameters such as: power in high frequency spectrum (HF; 0.15–0.4 Hz), power in low frequency spectrum (LF; 0.04–0.15 Hz), power in very low frequency spectrum (VLF; 0.003–0.04 Hz) and the ratio of the two values (LF/HF) (12).

RMSSD and HF predominantly represent cardiac parasympathetic nerve activity, whereas LF is associated with cardiac sympathetic nerve activity. Consequently, the LF/HF ratio serves as an indicator of the balance between sympathetic and parasympathetic functions, with elevated LF/HF values suggesting heightened sympathetic nerve excitability (11, 12).

All subjects received transthoracic echocardiography with Color Doppler Echocardiography (ECHO) performed by an experienced ultrasound physician. The ascending aortic diameter (AAOD), main pulmonary artery diameter (MPAD), left atrial diameter (LAD), left ventricular diastolic diameter (LVDD), right atrial diameter (RAD), right ventricular diameter (RVD), left ventricular systolic diameter (LVSD) and left ventricular posterior wall thickness (LVPWD) were measured using standard M-mode ECHO. The aortic valve stroke volume (AVSV), mitral valve e peak velocity (MVE) and left ventricular ejection fraction (LVEF) values of each patient were obtained from the results of ECHO.

Each subject was monitored for at least 7 h of sleep using a portable and non-contact sleep apnea monitor (Megasens Technology Co. Ltd., Hangzhou, China), including an ultra-wideband sleep apnea monitor (OrbSense, ZG-S01D) and blood oxygen saturation monitor (ZG-P11F). All of the subjects underwent 7 hours nighttime monitoring with this device in bed. No ingest caffeine, nap or engage and protracted or tedious exercises, and so on were informed during the study. The study recorded the following parameters: (1) Apnea-Hypopnea Index (AHI), which is the average number of apnea and hypopnea episodes per hour of sleep; (2) An apnea, which is a reduction in nasal airflow of at least ≥10 s 80%; (3) A hypopnea, which is a decrease in nasal airflow ≥30% and a decrease in oxygen saturation ≥4%; (4) the lowest and mean arterial oxygen saturation (LSaO2 and MSaO2), estimated by pulse oximetry during sleep [%], (5) Oxygen Desaturation Index (ODI), which is the number of average hourly decreases in oxygen saturation of ≥4%; and (6) the total sleep time with oxygen saturation below 90% (TST90) [%] (8–10). Subjects were categorized into four groups based on AHI by using common clinical cutoff points: none (AHI < 5), mild (5 ≤ AHI < 15), moderate (15 ≤ AHI < 30), and severe (AHI ≥ 30).

Based on the specific distribution and type of data, the results are presented as either mean ± standard deviation or number (percentage). Analysis of variance (ANOVA) was used for the normal distribution of data, and the Welch’s test was used for non-normally distributed data. The chi-squared test was used to compare categorical data. The relationship between variables was tested by Pearson and Spearman correlations, and multiple linear regression analyses were performed to evaluate the associations between dependent and independent variables. Statistical analyses were performed using SPSS v26.0 (IBM Corp., Armonk, NY, United States) and R program (v4.0; R Foundation, Vienna, Austria). The significance level of p < 0.05 was considered statistically significant for the entire analysis.

Multivariate logistic regression analysis was conducted to test for the correlation between OSA severity with AF and NSVT. Multivariate linear regression analysis was utilized in determining associations between OSA severity and various measurements of HRV and ECHO. Age, sex, BMI, drinking, smoking, hypertension, diabetes and myocardial infarction were considered as covariates in the multivariable models. In this study, mediation analysis was utilized for test whether the correlations of severe OSA with AF and NVST were mediated by HRV and ECHO indicators.

From January 2020 to May 2023, after screening 2,972 subjects with clinical suspicion of OSA in Renmin Hospital of Wuhan University, 600 subjects fulfilled these inclusions and were included in this study (Figure 1).

Table 1 summarizes the baseline characteristics of the study population and the four subgroups based on OSA severity. Of 600 subjects, 522 (87.0%) were diagnosed with OSA, including 145 with mild, 157 with moderate, and 122 with severe OSA. As expected, OSA severity was found to be associated with BMI and male gender. The incidence of smoking, drinking, hypertension, diabetes mellitus and myocardial infarction tended to increase in subjects with more severe OSA. The distribution of coronary artery disease and heart failure did not vary significantly among patients with and without sleep apnea. AAOD, MPAD, LAD, LVDD, RAD, RVD, BNP, Cr and fasting glucose were significantly higher in subjects with OSA than in those without OSA (all p < 0.05). In contrast, subjects with OSA had a significantly lower LVEF in comparison to those without OSA (p < 0.05). Table 1 also summarizes the sleep study parameters.

Table 2 summarizes the types of arrhythmias in subgroups based on OSA severity. The prevalence of AF and ventricular arrhythmias increased with the worsening OSA severity. There were no significant differences between groups regarding sinus arrhythmias and conduction delay arrhythmias. In addition, as shown in Table 3, considering the impact of AF on HRV, HRV parameters were compared among the different groups that excluded atrial fibrillation. The 24 h mean heart rate (HR), LF and LF/HF increased significantly as the worsened OSA severity. In contrast, SDNN, RMSSD and PNN50 decreased significantly as the worsened OSA severity.

Table 4 demonstrates that unadjusted and adjusted logistic regression examined the association between OSA with AF and NSVT. Compared with the group with none OSA, subjects with mild OSA had three times the unadjusted odds of AF (odds ratio [OR] 3.309; 95% confidence interval [CI], 1.097–9.977; p = 0.034). Subjects with moderate OSA had three times the unadjusted odds of AF (OR 3.338; 95% CI, 1.116–9.989; p = 0.031). Subjects with severe OSA had four times the unadjusted odds of AF (OR 4.624; 95% CI, 1.604–13.336; p = 0.005) and NSVT (OR 4.653; 95% CI, 1.073–20.169; p = 0.040). After further adjustments for age, gender, BMI, smoking, drinking, hypertension, diabetes mellitus and myocardial infarction, subjects with moderate OSA had three times the odds of AF (OR 3.055; 95% CI, 1.002–9.316; p = 0.048). Subjects with severe OSA had three times the odds of AF (OR 3.881; 95% CI, 1.306–11.534; p = 0.015) and NSVT (OR 3.690; 95% CI, 0.809–16.036; p = 0.046). In the unadjusted and adjusted model, there were significant linear trends for the association between OSA severity with AF and NVST (P for linear trends <0.05).

On the basis of Tables 1, 3, the correlations of severe OSA with significant indicators of HRV and ECHO indicators are shown in Table 5. After full adjustment for confounders, multivariate linear regression model showed that increased LAD, LVDD, RAD and RVD were independently associated with severe OSA (β = 1.638, p = 0.004; β = 1.069, p = 0.032; β = 1.418, p = 0.003; β = 0.590, p = 0.012, respectively) after full adjustment for confounders. Furthermore, after removing the influence of patients with AF, multivariate analysis demonstrated that severe OSA (β = 151.424, p = 0.010 and β = 0.334; p = 0.036, respectively) remained significantly associated with increases in LF and LF/HF.

Based on multivariate linear regression analysis, we examined whether severe OSA and AF and NVST were mediated by the significant HRV and ECHO indicators. Table 6 shows each step of the mediation analysis. These analyses, adjusted for potential confounders, yielded a mediation effect ratio that means how much the mediation effect accounts for the total effect of the independent variable on the dependent variable. In the mediation analysis between severe OSA and AF, the decomposition of the total effect showed that 93.1% of the total effect was mediated by the LAD, 16.1% of the total effect was mediated by the LVDD, 71.4% of the total effect was mediated by the RAD and 18.4% of the total effect was mediated by the RVD. And none of the controlled direct effect (effect directly associated with severe OSA) were significant (all p > 0.05). The relationship between severe OSA severity and AF depends almost entirely on LAD, LVDD, RAD and RVD (Figures 2A, B). In the mediation analysis between severe OSA and NSVT, the decomposition of the total effect showed that although the direct effect of severe OSA on NSVT remained significant (all p < 0.05), 19.9% of the total effect was mediated by the LAD, 9.7% of the total effect was mediated by the LVDD, 16.6% of the total effect was mediated by the RAD, and 10.6% of the total effect was mediated by the RVD (Figures 2C, D).

Considering the impact of AF on HRV, we also explored the associations mediated by LF and LF/HF between severe OSA and NSVT in patients who excluded atrial fibrillation. Compared with LF which indirect effect was not significant (b = 0.000; p = 0.820), the decomposition of the total effect indicated that 3.3% of the total effect was mediated by the LF/HF (Figure 2E).