This cross-sectional study was approved by the Institutional Ethics Committee of Shanxi Medical University First Hospital. All patients were enrolled after providing written informed consent and all underwent physical examination, overnight polysomnography, blood biochemistry, and multimodal echocardiography.

Between January 2019 and December 2019, a total of 221 patients diagnosed with OSAS from Sleep Units in the First Hospital of Shanxi Medical University were enrolled (167 men, 54 women; age, 20–68 years) (Fig. 1).

The inclusion criteria were age > 18 years and patients with an apnea–hypopnea index (AHI) ≥ 5 events/hour. Briefly, AHI is the sum of the average number of apnea and hypopnea per hour [12]. Apnea is defined as a loss or significant decrease (≥ 90% from baseline) of oral and nasal airflow during sleep for ≥ 10 s. Hypopnea is defined as a 30% or more increase in oral-nasal airflow from baseline during sleep accompanied by a 3% or more decrease in blood oxygen saturation for a duration ≥ 10 s. The exclusion criteria included the following: myocardial infarction, heart failure, arrhythmia, coronary heart disease (CHD), moderate or severe valvular regurgitation or stenosis, left ventricular ejection fraction < 50% (modified Simpson method), moderate or severe hepatic and renal disease, chronic obstructive pulmonary disease (COPD), hypothyroidism or hyperthyroidism, and neoplastic disease. Moreover, patients who received continuous positive airway pressure (CPAP) or surgical treatment and had unsatisfactory echocardiographic images were excluded from the study.

Physical examination was performed in all patients. Blood pressure (systolic blood pressure [SBP] and diastolic blood pressure [DBP]) was measured using a mercury sphygmomanometer on the naked right arm with patients in a seated point position after 30 min of rest. Hypertension was defined as systolic blood pressure (BP) ≥ 140 mmHg or diastolic blood pressure ≥ 90 mmHg or being treated with hypertensive medication [13]. According to the 1999 WHO diagnostic criteria, diabetes was defined as FPG ≥ 7.0 mmol/L and/or blood glucose ≥ 11.1 mmol/L at 2 h postprandial or having been treated with diabetes medication. Obesity was defined as body mass index (BMI) ≥ 28 kg/m², and dyslipidemia was defined as total cholesterol (TC) ≥ 5.17 mmol/L, triglycerides (TG) ≥ 1.7 mmol/L, high-density lipoprotein (HDL) < 1.03 mmol/L, and low-density lipoprotein (LDL) ≥ 3.33 mmol/L [14]. Body mass index (BMI) was calculated as the weight divided by height squared.

Blood samples were drawn from the peripheral vein the morning after PSG. Total cholesterol (TC), triglycerides (TG), high-density lipoprotein (HDL), low-density lipoprotein (LDL), and fasting blood glucose (Glu) levels were measured using an automatic biochemical analyzer (OLYMPUSAU5400, Japan).

All patients underwent overnight polysomnography in sleep units using standard recording techniques (SOMNO screen V series; German). Sleep recordings, including the apnea–hypopnea index (AHI), mean oxygen saturation (Mean-SaO₂), lowest oxygen saturation (lowest SaO₂), and percentage of total sleep time when blood oxygen saturation was less than 90% (T90), were obtained by a skilled technician based on the American Academic Sleep Medicine (AASM) 2017 criteria [15]. The AHI was defined as the number of apnea and hypopnea events per hour of sleep, and more than five were required for diagnosis.

Based on the recommendations of the American Society of Echocardiography [16], all patients underwent transthoracic echocardiographic examinations, including two-dimensional echocardiography (2DE), two-dimensional speckle-tracking echocardiography (2D-STE), and three-dimensional echocardiography (3DE), using EPIQ 7C Color Doppler Ultrasound (PHILIPS), with a S5-1and X5-1 probe and a frequency of 1.0–5.0 MHz. All patients were positioned in the left lateral decubitus position and connected to an electrocardiogram (ECG).

Left ventricular end-diastolic diameter (LVEDD), interventricular septum thickness (IVST), and left ventricular posterior wall thickness (LVPWT) were measured on the parasternal long-axis view. The early diastolic peak flow velocity (E) and late diastolic peak flow velocity (A) were measured using conventional pulsed Doppler imaging in the apical four-chamber view. Tissue Doppler imaging on the apical four-chamber view was used to measure mitral annular early diastolic myocardial velocity (e). Left ventricular ejection fraction was calculated using Simpson’s biplane method; meanwhile, left ventricular mass (LVM) was calculated while the left ventricular mass index (LVMI) was indexed to the body surface area by using the following formula [16]: 0.8 × 1.04 [ (LVEDD + IVST + PWT) ³ -LVEDD³]) + 0.6. The cutoff values of LVMI were 95 g/m² for women and 115 g/m² for men. Relative wall thickness (RWT), with a cutoff value of 0.42 [13], was calculated using the following formula: 2 × LVPWT/LVEDD.

Left ventricular geometric patterns were defined based on LVMI and RWT [16]: normal geometry (NG, normal LVMI and RWT), concentric remodeling (CR, normal LVMI, and increased RWT), concentric hypertrophy (CH, increased LVMI and RWT), and eccentric hypertrophy (EH, increased LVMI, and normal RWT).

The LA volumes (LAVs), including left atrial maximum volume (LAVmax), left atrium minimum volume (LAVmin), and left atrial pre-contraction volume (LAVpre), were measured using a biplane method in four-chamber and two-chamber views. The LAVmax was measured just before the mitral valve opening (the terminal of the T wave on the ECG), LAVmin was measured just after mitral valve closure (the peak of the R wave on the ECG), and LAVpre was measured at the onset of atrial systole (the peak of the P wave on the ECG). All LA volumes were indexed to body surface area (BSA).

All dynamic images of the four-chamber and two-chamber views were obtained using conventional two-dimensional echocardiography for three consecutive cardiac cycles. All images were digitally stored, while offline images were analyzed using QLAB-Philips software (version 10; Philips Medical Systems).

The software automatically tracked the LA endocardium and manually adjusted the region of interest when speckle tracking was insufficient. The software then calculated six segments of the LA strain and its overall strain. Finally, the overall LA strain, including LA strain during systole (LAS-S) and LA strain during late diastole (LAS-A), combined the average of the values obtained for the four-chamber and two-chamber views. The LAS-S was measured when the aortic valve was closed, while LAS-A was measured at the beginning of the P wave of the ECG. The LA strain during early diastole (LAS-E) was defined as the difference between LAS-S and LAS-A [17].

An X5-1 matrix-array transducer was used to acquire the “full-volume” for 4 consecutive cardiac cycles, and all images were digitally stored; meanwhile, they were analyzed offline using the QLAB-Philips software (version 10; Philips Medical Systems). The LA volumes were analyzed using the following anatomical landmarks: the septal, lateral, anterior, and posterior points on the atrial surface of the mitral annulus and the higher point of the LA roof. The LA appendages and pulmonary veins were excluded. Subsequently, the software automatically tracked the LA endocardium and manually adjusted the region of interest when necessary. Finally, the volume-time curve of the LA was obtained [18]. Measurement of LAVs was similar to measurement of the two-dimensional echocardiography as described above.

Based on the above LAVs, the parameters of LA function were calculated as follows [19]: LA total emptying volume (LA TotEV): LAVmax- LAVmax; LA passive emptying volume (LA PassEV): LAVmax- LAVpre; LA active emptying volume (LA ActEV): LAVpre-LAVmin; LA total emptying fraction (LA TotEF): LA TotEV/LAVmax × 100; LA passive emptying fraction (LA PassEF): LA PassEV/LAVmax × 100; LA active emptying fraction (LA ActEF): LA ActEV/LAVpre × 100.

Continuous variables were evaluated for normal distribution using the one-way Kolmogorov–Smirnov test. Normally distributed continuous variables were presented as mean ± standard deviation and analyzed using one-way ANOVA with Bonferroni post hoc test. Categorical variables were presented as percentages and were analyzed using the chi-square test. Multiple linear regression analysis was used to establish the relationship between left ventricular geometry and LAVImax, as well as left ventricular phasic function after adjusting for age, gender, BMI, SBP, diabetes mellitus, dyslipidemia, AHI, and LAMI. A P-value < 0.05 (P < 0.05) was considered statistically significant. All data were analyzed using SPSS 22.0 for Windows (IBM, Armonk, NY, USA). The figures were generated using GraphPad Prism 8.

As shown in Table 1, we recruited 221 patients with OSAS, 110 patients with normal geometry, 56 patients with concentric remodeling, 32 patients with concentric hypertrophy, and 23 patients with eccentric hypertrophy. In contrast with normal geometry and concentric remodeling, patients with concentric hypertrophy and eccentric hypertrophy were associated with older age (P < 0.05). Patients with concentric hypertrophy and eccentric hypertrophy had higher values of SBP, DBP, hypertension, AHI, and T90, but had lower values of mean-SaO₂ than those with normal geometry (P < 0.05). Regarding sex, BMI, obesity, diabetes mellitus, dyslipidemia, medications (including ARBs/ACEIs, Beta-blockers, CCB, diuretics, statin and antidiabetic), heart rate, lowest- SaO_2, and blood biochemical values, no statistical difference was found among the four groups (Table 1).

As shown in Table 2, patients with concentric hypertrophy and eccentric hypertrophy demonstrated a significantly larger LA dimension than those with normal geometry and concentric remodeling (P < 0.05). Similarly, both had greater interventricular septal thickness, left ventricular posterior wall thickness, and left ventricular mass index (P < 0.05). Furthermore, patients with concentric remodeling and concentric hypertrophy exhibited higher relative wall thickness than those with normal geometry and eccentric hypertrophy (P < 0.05). The E/A ratio gradually decreased, while the E/e ratio gradually increased from normal geometry to concentric hypertrophy. Regarding LVEF, no statistical difference was found among the four groups (Table 2).

As demonstrated in Table 3, 2DE- and 3DE LA volumes and indices (maximum, minimum, pre-contraction) gradually increased from normal geometry to concentric hypertrophy. The 3DE-LAVmax showed a statistically significant difference among the four groups, while 2DE-LAVmax was statistically significant among all four groups, except for concentric hypertrophy and eccentric hypertrophy. In addition, 2DE- and 3DE-LAVImax revealed a statistically significant difference among the other left ventricular geometries, except for normal geometry and concentric remodeling (Figs. 2 and 3). Patients with concentric hypertrophy had higher 3DE-LAVmin and 3DE-LAVImin than those with normal geometry, concentric remodeling, and eccentric hypertrophy, as well as higher 2DE- LAVmin and 2DE-LAVImin than those with normal geometry and concentric remodeling (P < 0.05) (Figs. 2 and 3). Additionally, 2DE-and 3DE-LAVpre and LAVIpre were significantly different in patients with left ventricular geometry (P < 0.05) (Figs. 2 and 3).

Notably, 2DE-LA TotEV was greater in patients with concentric hypertrophy than in those with normal geometry and concentric remodeling; 3DE-LA TotEV was greater in patients with concentric hypertrophy than in those with normal geometry, concentric remodeling, and eccentric hypertrophy; 2DE- and 3DE-LA TotEV were greater in eccentric hypertrophy than in normal geometry. The 3DE-LA PassEV was lower in patients with concentric hypertrophy than in those with normal geometry and concentric remodeling. However, 3DE-LA PassEV showed no differences among the left ventricular geometric patterns. Patients with concentric hypertrophy and eccentric hypertrophy demonstrated a significantly larger 2DE-LA ActEV compared to those with normal geometry and concentric remodeling. Furthermore, patients with concentric hypertrophy revealed a significantly larger 3DE-LA ActEV compared to any of the left ventricular geometric patterns, and larger eccentric hypertrophy compared to those with normal geometry.

The trend of change in left ventricular strains was similar to the change observed in LA volumes. The LAS-S and LAS-E values progressively decreased from normal geometry to concentric hypertrophy; however, LAS-A gradually increased. Patients with concentric hypertrophy and eccentric hypertrophy had higher LAS-S and LAS-E than those with normal geometry and concentric remodeling. The LAS-A was higher among patients with concentric hypertrophy than in those presenting any left ventricular geometric pattern (P < 0.05) (Fig. 4).

As shown in Table 4, the LA reservoir, estimated by LA total emptying fraction (LA TotEF) and conduit function, estimated by LA passive emptying fraction (LA PassEF), gradually decreased from normal geometry to concentric hypertrophy. Nevertheless, LA booster function, estimated by LA active emptying fraction (LA ActEF), gradually increased in a similar direction. Patients with concentric hypertrophy had lower 2DE-LA TotEF compared to those with normal geometry and lower 3DE-LA TotEF compared to those with normal geometry and concentric remodeling (Figs. 5 and 6). Patients with concentric hypertrophy and eccentric hypertrophy had lower 2DE-LA PassEF compared to those with normal geometry and concentric remodeling. In addition, patients with concentric hypertrophy had lower 3DE-LA PassEF compared to those with normal geometry and concentric remodeling, and lower 3DE-LA PassEF in patients with eccentric hypertrophy compared to those with normal geometry (Figs. 5 and 6). Patients with concentric hypertrophy had higher 3DE-LA ActEF values than those with normal geometry. However, 2DE-LA ActEF showed no difference among the left ventricular geometric patterns (Figs. 5 and 6).

In order to clarify the independent influence of the left ventricular geometric patterns on LAVImax and LA phasic function, we performed multiple linear regression analyses, in which age, gender, BMI, SBP, diabetes mellitus, dyslipidemia, AHI, and LVMI were included in the model. As shown in Table 5, concentric hypertrophy was significantly associated with 2DE-LAVImax (β = 0.449, 95%CI: 3.592–7.436; P < 0.001), while concentric and eccentric hypertrophy were associated with 3DE-LAVImax (β = 0.539, 95%CI: 5.946–9.761; P < 0.001 and β = 0.137, 95%CI: 0.290–4.306; P = 0.025, respectively). Subsequently, concentric hypertrophy was associated with 2DE- and 3DE-LA TotEF (β = -0.219, 95%CI: -0.074- (-0.001); P = 0.046 and β = -0.276, 95%CI: -0.073- (-0.010); P = 0.010, respectively). Moreover, concentric hypertrophy and eccentric hypertrophy were associated with 2DE-LA PassEF (β = -0.422, 95%CI: -0.148- (-0.052); P < 0.001 and β = -0.221, 95%CI: -0.111- (-0.010); P = 0.019, respectively), while concentric hypertrophy, eccentric hypertrophy, and concentric remodeling were associated with 3DE-LA PassEF (β = -0.589, 95%CI: -0.126- (-0.070); P < 0.001, β = -0.382, 95%CI: -0.103- (-0.044); P < 0.001, and β = -0.138, 95%CI: -0.034- (-0.003); P = 0.018, respectively). Only concentric hypertrophy was associated with 3DE-LA ActEF (β = 0.239, 95%CI: 0.005–0.144; P = 0.035). Furthermore, concentric and eccentric hypertrophy were associated with 3DE-LAS-S, 3DE-LAS-E, and 3DE-LAS-A (all, P < 0.05).