Associations of the Cardiometabolic Index with the Risk of Cardiovascular Disease in Patients with Hypertension and Obstructive Sleep Apnea: Results of a Longitudinal Cohort Study

2.1. Study Population

All data were obtained from the UROSAH study, and a detailed description of the UROSAH study has been reported elsewhere [33]. In brief, UROSAH is a single-center, observational, retrospective cohort study designed to investigate the relationship between long-term cardiovascular outcomes in patients with hypertension and OSA. In this study, the study population was hypertensive patients diagnosed with OSA at the Hypertension Center of Xinjiang Uygur Autonomous Region People's Hospital between January 2011 and December 2013. Patients with a definite prior diagnosis of hypertension or with a newly diagnosed hypertension at admission, criteria of which were referred to guidelines from the Chinese Hypertension League, were considered to have hypertension. All patients were diagnosed with OSA after admission by laboratory nocturnal polysomnography (PSG), which was defined in detail following the relevant guidelines. Other inclusions and exclusion criteria are detailed in the flow chart (Figure 1). Overall, 2067 participants were ultimately included in this analysis after strict adherence to the inclusion and exclusion criteria.

2.2. Ethical Approval

The ethics application was approved by the Ethics Committee of the Xinjiang Uygur Autonomous Region People's Hospital (reference: 2019030662). All patients gave informed consent according to the Declaration of Helsinki.

2.3. Assessment of Covariates

Baseline data collection was expected to be completed by professional research staff following the operational procedures mentioned in previous studies [33]. One-on-one interviews were held with each participant using a standardized electronic medical record. Diabetes was defined as fasting glucose ≥ 7.0 mmol/L, use of any glucose-lowering medication, or self-reported history of diabetes. Regular continuous positive airway pressure (CPAP) treatment was defined as the use of CPAP therapy for more than 70% of nights throughout the follow-up period and no less than 4 hours per night, or an average of ≥4 hours per night [34, 35]. Similar to the definition of CPAP treatment use, subjects with at least 4 hours of use per night and at least 70% of nights of continuous use were considered to be regular oral appliance treatment users [36].

2.4. Laboratory Assays

On the morning of the survey, venous blood samples were harvested after an overnight fast (>8 hours) and sent to the central laboratory for analysis shortly after sampling. All biochemical measurements including total cholesterol (TC), TG, HDL-C, low-density lipoprotein cholesterol (LDL-C), and fasting plasma glucose (FPG) were measured on a fully automated biochemical analyzer (Hitachi 7600; Japan). The estimated glomerular filtration rate (eGFR) was derived by using the equations of the Chronic Kidney Disease Epidemiology Collaborative Group [37].

2.5. Calculation of CMI

The TG/HDL-C ratio was computed by dividing the serum concentration of TG (mmol/L) by HDL-C (mmol/L). WHtR was derived by dividing WC (cm) by height (cm). CMI was computed as TG/HDL‐C × WHtR according to the formula in previously published papers [26].

2.6. OSA Evaluation

All participants underwent nocturnal PSG monitoring in the sleep laboratory. All sleep monitoring results were analyzed by specialized sleep technicians according to criteria [38, 39]. Details of the PSG and scoring criteria applied in this research are available in the Supplementary Material. OSA was defined as an apnea-hypopnea index (AHI) of ≥5 events/h. The severity of OSA was classified as normal (<5), mild (5-14.9), moderate (15-29.9), and severe (≥30) according to the AHI [39, 40].

2.7. Endpoints and Follow-Up

In this study, the primary outcome was new-onset cardiovascular disease (CVD), including stroke and coronary heart disease (CHD). Cardiologists from the Clinical Events Committee of a tertiary care hospital diagnose CVD after reviewing all medical records. CHD was defined as myocardial infarction (fatal or nonfatal), coronary revascularization (percutaneous transluminal coronary angioplasty or coronary artery bypass grafting), and angina pectoris (unstable or stable), which were diagnosed by electrocardiogram, signs or symptoms of ischemia, or diagnostic enzymes. A stroke was defined as a sudden or rapid onset of focal neurological deficit lasting longer than 24 hours or until death. The diagnosis was confirmed by computed tomography scan, magnetic resonance imaging, or the patient's clinical symptoms. Stroke includes both ischemic and hemorrhagic strokes. After entry into this cohort, subjects were followed up for the first occurrence of CVD. Follow-up results were available through outpatient interviews, telephone interviews, and inpatient medical records. Deaths due to CVD were verified by hospital death certificates or hospitalization data or by the local police system. The follow-up period began with the initial visit and continued until the participant had a new onset of CVD or died (whichever occurred first), otherwise until the end of the follow-up period in January 2021.

2.8. Statistical Analysis

Means ± standard deviation (SD), medians (Q25, Q75), or proportions were illustrated for characteristics of the participants by CMI tertiles. Differences in group characteristics were compared by using Fisher's exact test, ANOVA test, Kruskal-Wallis test, or χ² test, accordingly. Kaplan-Meier curves and log-rank tests were used to present the unadjusted cumulative hazards across the tertiles of CMI.

Multicollinearity among the variables was tested by calculating the variance inflation factor (VIF). Variables with VIF > 5 were considered to have severe multicollinearity and could not be incorporated in the multivariate regression analysis. Hazard ratios (HRs) and 95% confidence intervals (CIs) of new-onset CVD or its subtypes (CHD and stroke) in relation to CMI (continuous, tertiles, and categories) were calculated through the use of Cox proportional hazards models. Model 1 was adjusted for age and sex; Model 2 was further adjusted for drinking status, history of diabetes, smoking status, DBP, SBP, and BMI; Model 3 was further adjusted for all noncollinear variables. After entering the median CMI values of each tertile into the model and treating them as continuous variables, the regression model was tested for trend. To test the robustness of our results, we performed several sensitivity analyses. First, we excluded participants who were current smokers and alcohol drinkers and repeated the analysis. Second, we excluded participants who were regularly receiving CPAP treatment. Third, we excluded events that occurred one year prior to follow-up to minimize potential reverse causes. Finally, non-CVD deaths were considered a competing risk, and we performed a competing risk regression analysis based on the methods reported by Fine and Gray. Furthermore, we explored the relationship between the CMI and new-onset CVD, CHD, and stroke using a generalized additive model and smoothed curve fitting.

Stratified analyses and interactions were separately performed by age (<60 or ≥60years), sex (female or male), SBP (<140 or ≥140 mmHg), BMI (<28 or ≥28 kg/m²), smoking (never, past, or current), DBP (<90 or ≥90 mmHg), drinking (never, past, or current), and AHI (≥30, ≥15 to <30, or ≥5 to <15 events/h).

Receiver operating characteristic (ROC) curves were used to estimate the predictive values of possible predictors for endpoints (CVD, CHD, and stroke). The C-index, continuous net reclassification improvement (NRI), and integrated discrimination index (IDI) were used to evaluate the additional predictive value of the CMI beyond established risk factors. Confidence intervals for C-index, continuous NRI, and IDI were estimated using 1000 bootstrap procedures.

All analyses were undertaken with R software (version 4.0.1). The threshold of significance was fixed at 5% (P < 0.05 was significant).

3.1. Baseline Characteristics of Participants

A total of 2067 eligible subjects with a mean age of 49.51 ± 10.73 years were included in this study, of which 68.92% were male. Baseline characteristics of subjects according to CMI tertile are presented in Table 1. Participants in the other groups were more likely to be younger, male, more current smokers, and drinkers; have a higher prevalence of diabetes; and have a higher BMI, eGFR, TC, TG, FPG, and AHI levels and a lower HDL-C level compared to the tertile 1 group.

3.2. Association between CMI and Risk of CVD and Its Subtypes

During a median follow-up of 6.83 years (interquartile range: 5.92-8.00 years), 326 (14.19%) incident CVD were identified, including 121 (5.27%) incident stroke and 205 (8.92%) incident CHD. Furthermore, cumulative hazard curves of cardiovascular events across CMI tertiles are shown in Figure 2 (log-rank test: P < 0.001 for CVD and CHD, respectively; P = 0.084 for stroke).

In total, there was a positive relation between CMI and the risk of new-onset CVD (Figure 3(a)), new-onset CHD (Figure 3(b)), and new-onset stroke (Figure 3(c)). As presented in Table 2, the HR (95% CI) for CVD in the fully adjusted regression model was 1.31 (1.20, 1.43) for per SD increase in CMI. Fully adjusted HRs (95% CIs) for CVD were 1.60 (1.18, 2.16) and 2.17 (1.63, 2.90) for subjects in tertile 2 and tertile 3, respectively. The risk of CVD increased substantially from quartile 1 to quartile 3 (P for trend < 0.001). Similarly, there was a meaningful association between CMI and CHD and stroke in the fully adjusted regression model (Table 2). For each SD increase in the CMI, the HR (95% CI) of the CHD rate was 1.33 (1.20, 1.48) in the fully adjusted regression model. The fully adjusted HR (95% CI) for CHD in tertile 3 was 2.62 (1.79, 3.83). The increased risk of CHD from tertile 1 to tertile 3 was statistically significant (P for trend < 0.001). Also, a remarkable positive association was shown between CMI and stroke (per SD increment; HR: 1.27; 95% CI: 1.10, 1.47). In the fully adjusted regression model, the HRs (95% CIs) for stroke were 1.23 (0.77, 1.98) and 1.64 (1.05, 2.57) for participants in tertile 2 and tertile 3, respectively, compared with participants in tertile 1 (P for trend = 0.029). In addition, the risk of new-onset CVD (HR: 1.68; 95% CI: 1.35, 2.09), CHD (HR: 1.81; 95% CI: 1.37, 2.39), and stroke (HR: 1.47; 95% CI: 1.02, 2.12) was significantly higher in fully adjusted regression models for participants in tertile 3 of CMI level (≥1.21) compared with those in tertile 1-2 (<1.21) (Table 2).

3.3. Sensitivity Analysis

In sensitivity analyses, the association between CMI and the risk of CVD, CHD, and stroke did not change substantially after excluding participants who were regularly treated with CPAP (Supplemental Table 2), or after excluding participants who were current smokers and drinkers (Supplemental Table 3 and Supplemental Table 4), or after excluding participants who developed CVD within the first year of follow-up (Supplemental Table 5). In addition, a sensitivity analysis was performed using the Fine-Gray model to reduce the effect of competing mortality risks. The results of the competing risks model were similar to those of the Cox model (Supplemental Table 6).

3.4. Subgroup Analysis

We further performed a stratified analysis to assess the association of CMI (per SD increment) with risk of CVD (Figure 4(a)), risk of CHD (Figure 4(b)), and risk of stroke (Figure 4(c)). None of the variables, including age (<60 vs. ≥60 years), sex (female vs. male), SBP (<140 vs. ≥140 mmHg), BMI (<28 vs. ≥28 kg/m²), smoking (never vs. past vs. current), DBP (<90 vs. ≥90 mmHg), drinking (never vs. past vs. current), and AHI (≥ 30 vs. ≥ 15 to <30 vs. ≥ 5 to <15 events/h), were found to modify the association between CMI (per SD increment) and the risk of CVD (all P interactions > 0.05). Similar outcomes were observed for CHD and stroke.

3.5. ROC Curve Analysis of Different Markers

The results of the ROC analysis for CMI and other predictors are shown in Table 3 and Supplementary Figure 1. Compared to TC, TG, HDL-C, TG/HDL-C, LDL-C, FPG, and WHtR, CMI was the strongest predictor of CVD (AUC = 0.617), with a sensitivity of 0.675, a specificity of 0.509, and the best threshold of 1.006 (Table 3). Similar outcomes were observed for CHD and stroke.

3.6. Additive Effect of CMI to Established Risk Factors

The additional value of CMI for risk prediction of CVD, CHD, and stroke was evaluated based on discrimination (C-index), continuous NRI, and IDI (Table 4). The addition of CMI significantly improved the reclassification and discrimination for CVD prediction over the baseline risk model with a continuous NRI of 0.179 (0.114, 0.248) and an IDI of 0.013 (0.005, 0.025) (both P < 0.001). Similarly, the reclassification and discrimination for CHD and stroke prediction were also significantly improved with the addition of CMI. The C-index for predicting CVD, CHD, and stroke without CMI in the established risk factors were 0.600 (0.567, 0.633), 0.639 (0.602, 0.677), and 0.540 (0.484, 0.596), respectively. However, the addition of CMI to the established risk factors significantly improved the prediction of new-onset CVD, new-onset CHD and new-onset stroke with C-index of 0.648 (0.615, 0.681), 0.675 (0.638, 0.712), and 0.596 (0.541, 0.652), respectively.