Association between remnant cholesterol, metabolic syndrome, and cardiovascular disease: post hoc analysis of a prospective national cohort study

This study used data from the CHNS, an ongoing longitudinal community-based cohort study carried out by the national and local governments of China. The study includes data for more than 12,000 individuals across approximately nine provinces. Trained researchers conducted household surveys, using standard questionnaires and face-to-face interviews to obtain information about participants. Each participant provided written informed consent. Study details are described in our previous article [21], and the relevant protocol is published elsewhere [22]. The inclusion criteria were: (1) age ≥ 18 years old, (2) measurement of RC level. The exclusion criteria were: (1) pregnant women, (2) without sufficient information to diagnose MetS. We first removed 2466 participants that lacked fasting blood samples, then we excluded 1054 participants without RC measurement, 741 participants who were under 18 years of age, 57 pregnant women, 226 participants without sufficient information to diagnose MetS (Fig. 1). Based on the aforementioned criteria, 7471 eligible participants were identified. We also presented demographic information, health behaviors, health status and laboratory examinations of study participants. After six years follow-up, 205 patients diagnosed with CVD, with 1041 individuals loss follow-up.

This manuscript was written in strict accordance with the STROBE statement [23].

RC (mmol/L) was calculated as total cholesterol (TC) (mmol/L) minus LDL-C (mmol/L) minus HDL-C (mmol/L) and blood sample were obtained in the fasting state [24]. MetS was diagnosed according to the International Diabetes Federation as individuals with central obesity (waist circumference ≥ 90 cm in men or ≥ 80 cm in women) plus any two of the following: (1) raised TG (> 1.7 mmol/L) or specific treatment for TG abnormality; (2) reduced HDL-C in men (< 1.03 mmol/L) and in women (< 1.29 mmol/L) or specific treatment for TG abnormality; (3) raised blood pressure (SBP ≥ 130 mm Hg or DBP ≥ 85 mmHg) or treatment of previously diagnosed hypertension; (4) raised fasting plasma glucose (fasting plasma glucose ≥ 5.6 mmol/L) or previously diagnosed type 2 DM [25]. CVD was defined as myocardial infarction or stroke and was derived by medical diagnosis [26]from doctors.

Height and weight were measured while the subjects were wearing light clothing and standing without shoes. We calculated body mass index (BMI) as weight (kg)/height (m)². Health behaviors (smoking and alcohol consumption), education background and residence were self-reported. Smoking was defined as any previous smoking (yes/no), and alcohol consumption was defined as imbibition greater than three times per week (yes/no). Renal function was presented as eGFR using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation [27]. Nutrition intake was assessed through questionnaire that included 24 h diet recalls about the food on the same 3 days (2 week days and 1 weekend day). Energy intake, carbohydrate intake, fat intake and protein intake were calculated by multiplying the intake of each food from the average dietary intake for 3 days by the standard serving size (100 g). Physical activity was defined as more than 150 min of moderate exercise or more than 75 min of vigorous exercise per week [28]. Creatinine, uric acid, fasting blood glucose, HDL-C, LDL-C, TG, TC,were measured using a Hitachi 7600 machine (Randox, UK and Kyowa, Japan). HbA1c was detected by HLC-723 G7/D10/PDQ A1c (Tosoh, Japan/Bio-Rad, USA/Primus, USA). Insulin was detected by Gamma counter XH-6020,China (North Institute of Bio-Tech,China). High-sensitivity C-reactive protein (Hs-CRP) was measured by Hitachi 7600 machine (Denka Seiken, Japan).

Homeostatic model assessment of insulin resistance (HOMA-IR) was calculated by: fasting insulin (µIU/mL) × fasting glucose (mmol/L)/22.5.

Participant characteristics were described based on tertiles of RC. Continuous variables are expressed as means ± standard deviation for normal distributions or medians and interquartile range (25% to 75%) for skewed distributions. Categorical variables are presented as relative frequencies (percentages). Each tertile of RC was taken as a unit and p-values for trends were calculated using linear-regression analyses for continuous variables and Cochran-Armitage test for categorical variables. We used an upset and correlation plot to show the distribution of different combinations of MetS components and the relationships between the MetS components. Generalized additive models were then used to identify relationships between RC and the prevalence of each MetS component, since RC was a continuous variable. The independent association of RC with MetS and CVD were evaluated using logistic models with odds ratios (ORs) and 95% confidence intervals (CIs), and Cox regression models with hazard ratios (HRs) and 95% CIs. Potential covariates that were significant in the baseline comparison, or that we considered to be of clinical importance were included in the multivariate models. We established two main models for covariate adjustment: crude model; adjusted model for age, sex, education, residence, smoking, alcohol consumption, protein intake, carbohydrate intake, fat intake, daily energy intake and physical activity. Subgroup analyses and effect modification were performed considering age (< 60/ ≥ 60 years old), sex (male/female), education levels (middle school and below/high school and above), alcohol consumption (yes/no), smoking(yes/no) and residence(urban/rural). Finally, given that the MetS has been identified as a way through which RC may affect the CVD, we investigated whether the association between RC and CVD was mediated by MetS. The bootstrap test was used to assess the effects of these mediators [29].

The sensitivity analysis was conducted by adopting the WHO criterion to diagnose MetS [26]. The proportion of missing data in the analytic sample did not exceed 2%. Missing data were interpolated using the method of last observation carried forward or using the means and medians for continuous variables and skewed variables. Comparisons where P was < 0.05 (two-sided) were considered to be statistically significant. We performed all analyses with Stata 15.0, R (version 3.4.3) and EmpowerStats (http://​www.​empowerstats.​com, X and Y Solutions, Inc., Boston, MA).

Demographic characteristics of the study population are summarized in Tables 1 and 2. Among the 7471 participants, there were 1848 individuals (24.74%) who had MetS and 205 individuals (2.74%) with CVD. Urban residents accounted for the majority of subjects across the different RC groups. Subjects with MetS exhibited statistically higher RC levels and had significantly higher BMI, uric acid, creatinine, LDL-C, TG, TC, HOMA-IR, Hb1Ac, fasting blood glucose, and insulin levels. Conversely eGFR and HDL-C levels were significantly lower within the elevated RC tertile. Compared with lower-level RC group, subjects within higher tertiles of RC were older, more educated, and engaged in smoking and alcohol consumption more frequently, and their protein intake was also higher. Participants were spilt into two groups, according to whether MetS was present. Female and urban residents accounted for the majority of subjects in the group with MetS. These subjects were statistically significant higher in age, BMI, uric acid, Hs-CRP, LDL-C, TG, TC, RC, HOMA-IR, HbA1c, fasting blood glucose, carbohydrate intake, daily energy intake and insulin levels. Subjects with MetS also had statistically significantly lower HDL-C and eGFR when compared to participants without MetS.

By analyzing the upset plot (Fig. 2), we observed that among the general metabolic disorders, individuals with expanded waist circumference were the most prevalent in the general population, and participants with low HDL-C were relatively rare by contrast. Corresponding to the diagnosis of MetS, the population with all five factors coexisting accounted for the majority of MetS cases, while the proportion of the participants with the combination of expanded waist circumference, high blood glucose and low HDL-C was comparatively small.

Table 3 shows the association between RC tertiles and prevalence of MetS. By using logistic regression models, per-tertile-increase in RC were associated with 317% increasing risk for MetS (OR: 3.17, 95% CI 2.93–3.43, P for trend < 0.001). After adjusting for age, sex, education, residence, alcohol consumption, protein intake, carbohydrate intake, fat intake, daily energy intake and physical activity, the result remained significant (OR: 3.49, 95% CI 3.21–3.79, P for trend < 0.001). As shown in Fig. 3, with rising concentration of RC, the prevalence of MetS, abdominal obesity, elevated triglycerides, reduced HDL-C, elevated blood pressure, and abnormal glucose metabolism simultaneously increased. After analyzing the correlation between RC and each component of MetS in Fig. 4, we found out that RC level had a positive correlation with waist circumference, abnormal glucose metabolism, raised blood pressure and triglyceride levels, with the triglycerides serving as the most relevant factor for RC. All components were correlated with each other positively except for reduced HDL-C, which has negative correlation with RC, triglyceride, abnormal glucose metabolism and waist circumference. Though, HDL-C is irrelevant to raised blood pressure.

After a median follow-up of 6 years, we identified 205 (2.74%) subjects who developed CVD. By using the Cox regression analysis in Table 4, RC had a significant correlation with risk of CVD in crude model (HR: 1.22, 95% CI 1.03–1.45, P for trend = 0.021). After adjusting for age, sex, education, residence, alcohol consumption, protein intake, carbohydrate intake, fat intake, daily energy intake and physical activity, the result remained significant (HR: 1.26, 95% CI 1.06–1.50, P for trend = 0.008).

Subgroup analyses of the asocciation between RC, MetS and CVD was listed in Table 5 and Table 6. There was an interaction between age and RC for MetS, as people under 60 years old at high RC levels were significantly associated with MetS (OR: 3.84, 95%CI 3.45–4.29, P for interaction < 0.001). However, other stratification factors such as sex, education, alcohol consumption, smoking and residence had no interaction with RC regarding the incidence of metabolic syndrome. No interaction were found of all subgroups in the relationship between RC and CVD.

Potential mediation effects of RC in the correlation between MetS and CVD are presented in Fig. 5. We observed significant indirect effects of RC between MetS and CVD with mediation at 48.46% of the association.

To consolidate our result, we conducted the sensitivity analysis base on the different diagnostic criterion of MetS (WHO standard) and the results were listed below in Table 7 and Fig. 6. We re-did the mediation analysis and logistic regression. The results were similar to the main results.