This study was conducted in the sleep center of Peking Union Medical College Hospital. Since in clinical practice, many more male patients were found seeking consultation than females, only male patients with OSA were evaluated for enrollment after completing the PSG in the sleep center to get a better analysis. The apnea-hypopnea index (AHI) ≥ 5/h with concomitant symptoms (such as snoring, sleepiness, or observable apnea), or the AHI ≥ 15/h without symptoms, was used to diagnose OSA [10]. Patients were excluded if they met the following criteria: (1) having had OSA treatment, such as continuous positive airway pressure (CPAP), before enrollment; (2) utilizing drugs to improve sleep; (3) the pulmonary function tests (PFT) were performed with a rate of the forced expiratory volume in 1s (FEV1) to the forced vital capacity (FVC) < 0.7; (4) suffering from severe neuromuscular disease and central nervous system diseases; (5) having total sleep time < 4 h; (6) refusing or being unable to complete the rebreathing test. The study was approved by the ethics committees of Peking Union Medical College Hospital (K3069) and was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants.

Basic demographic data, including age, body mass index (BMI), and neck circumference of each participant, were collected. The smoking status, alcohol consumption, habit of drinking tea or coffee, medication consumption, comorbidities, and blood pressure were collected before finishing PSG. All participants filled out the Epworth Sleepiness Scale (ESS) questionnaires, the Pittsburgh Sleep Quality Index (PSQI), the Basic Nordic Sleep Questionnaires (BNSQ), and the Berlin questionnaires, thus 20 clinical symptoms reflecting the complaints in the past one month were extracted from them. The data on fasting blood glucose (FBG) and fasting insulin were obtained in the early morning after patients fasted for more than 8 h. The homeostasis model assessment of insulin resistance (HOMA-IR) was counted as the product of fasting insulin (with the unit µIU/mL) and fasting glucose (with the unit mmol/L) divided by 22.5. The estimated glomerular filtration rate based on creatinine level (eGFRcr) was calculated using the 2021 Chronic Kidney Disease Epidemiology (CKD-EPI) equations [11].

All participants underwent the whole-night PSG (Embla N7000, Natus Medical Incorporated, Orlando, FL, USA) from 11 p.m. to 6 a.m. in the sleep center. Participants were not allowed to drink additional fluids before or during bedtime unless they had kept this habit for a long time. Data were recorded and analyzed by skilled sleep laboratory technicians following the standard criteria recommended by the American Academy of Sleep Medicine, version 2.5 [12]. AHI was defined as the number of apnea and hypopnea events per hour. The oxygen desaturation index (ODI) was defined as the number of desaturations per hour with the desaturation threshold set at 3%. The arousal index was defined as the number of arousals per hour. The total sleep time (TST), the events of apnea or hypopnea per hour, the events in the supine position and the rapid eye movement (REM) states per hour, the mean oxygen saturation by pulse oximetry (mSpO2), the lowest oxygen saturation by pulse oximetry (LSpO2), and the percent of time spent with SpO2 below 90% (T90) were also collected. The low arousal threshold (LAT) was calculated through the method proposed by Edwards et al. [13]: AHI < 30 events/h, LSpO2 > 82.5%, and the percentage of hypopneas > 58.3%. Each criterion earns 1 score, and a score of 2 or above predicts the existence of LAT.

The participants undertook the PFT and the rebreathing test in the early morning with fasted status. The cardiopulmonary exercise cart was used to assess the pulmonary function of participants, and data on the FEV1 and FEV1/FVC ratios were acquired. Then the participants finished the rebreathing test in accordance with Duffin’s method [9]. They were in a seated position, breathed with a nose peg and a mouthpiece connected to a closed one-way circuit with a 6 L plastic rebreathing bag containing a gas mixture of 4% O2 /6% CO2 /balanced nitrogen, and an adjustable concentration of O2 was fed into the rebreathing bag to maintain end-expiratory oxygen partial pressure at 50 mmHg during the test. The test lasted 3–5 min or was terminated when the oxygen saturation by pulse oximetry was lower than 88%, end-tidal carbon dioxide partial pressure (PETCO2) was higher than 60 mmHg, or minute ventilation was exceeded 100 L/min. The volume transducer and gas sampling port of the cardiopulmonary exercise cart (MasterScreen CPX; Jaeger, Hoechberg, Germany) were connected with the mouthpiece and rebreathing bag to collect the data. Breath-by-breath measurements monitored expired gas concentrations and minute ventilation. HCVR was evaluated by relating PETCO2 with minute ventilation during the rebreathing test using linear regression, representing the overall chemosensitivity. The detailed procedure has been described in a previous study [14].

A total of 7 variables were enrolled in the cluster analysis, including the symptom of nocturnal urination (defined as occurring at least once per week in the last month, according to PQSI), 2 demographic data (age and BMI), and 4 PSG indices (AHI, ODI, T90, and arousal index) which were most commonly evaluated in clinical works.

Categorical data were presented as numbers (in percentage). Continuous variables were shown as mean ± SD or median (interquartile range, 25–75%) depending on whether the data were normally distributed or not. Differences between groups were examined via chi-squared, independent-sample Tests, ANOVA, Mann–Whitney U-test, and Kruskal-Wallis test as appropriate. Univariate and multivariate linear regression were used to evaluate the association between indices and chemosensitivity. Two-step cluster in SPSS was used to cluster subjects into groups based on the symptom and PSG indices. SPSS version 26.0 (Chicago, IL, USA) was used for data analysis. A two-sided P-value < 0.05 was considered statistical significance.

Combining earlier research makes it challenging to determine the degree of chemosensitivity in OSA patients and the variations among populations because chemosensitivity is not a commonly used and evaluated metric. Based on the empirical estimation of the built model and the idea that there should be more sample sizes than 10 events per variable (EPV) [15], the regression model built for this study included 13 covariates in total. As a result, it would be better to include at least 130 subjects in the study. Although our study’s EPV was 8 (104 subjects with 13 covariates), we believe that the sample size may accurately represent the findings to some extent since it is as effective as EPV 10 according to a previous study [16].