Introduction
The prevalence of cardiometabolic diseases (CMDs) is gradually increasing, with the affected population becoming younger, which is expected to significantly increase the medical burden, functional decline, and mortality risk among patients [1]. CMDs primarily refer to metabolic abnormalities that have a clear causal relationship with cardiovascular damage, including hypertension, type 2 diabetes, coronary heart disease, stroke, and other related conditions [2]. The coexistence of 2 or more CMDs, known as cardiometabolic multimorbidity (CMM), represents more than a simple superposition of multiple diseases.
The interaction between the CMDs accelerates disease progression and leads to worse clinical outcomes [3]. To reduce the risk of cardiovascular diseases, particularly the occurrence of ischemic complications, and to thereby prolong the life span and improve the patients’ quality of life, the American Heart Association has established comprehensive management and treatment guidelines for metabolic syndrome and cardiovascular diseases [4,5]. Various risk factors affect the occurrence of cardiometabolic diseases, including glucose metabolism disorders, dyslipidemia, and poor lifestyle habits [6].
Dyslipidemia is one of the most important risk factors for coronary heart disease, especially atherosclerotic cardiovascular disease [7]. Previously, low-density lipoprotein cholesterol (LDL-C) was considered the primary target for atherosclerotic cardiovascular disease risk prediction and treatment. However, studies have demonstrated that even when LDL-C levels are optimally managed with medication, a significant proportion of patients continue to experience cardiovascular events, suggesting the presence of residual risk [8].
Remnant cholesterol (RC), reflecting cholesterol carried in triglyceride-rich lipoproteins and their remnants, has been increasingly recognized as a contributor to atherosclerosis and cardiometabolic risk [9,10]. These remnant lipoproteins are highly atherogenic and can directly penetrate the arterial wall, where they are more readily taken up by macrophages to form foam cells, compared with low-density lipoprotein alone, thereby accelerating progression, releasing free fatty acids and monoacylglycerol, and triggering an inflammatory response [10–12]. In addition, several studies have demonstrated associations between RC and the risk of developing type 2 diabetes and chronic kidney disease [13–15].
Inflammation is also a critical factor in the progression of cardiovascular disease and metabolic disorders. C-reactive protein (CRP) and high-sensitivity CRP (hsCRP), as well-established markers of inflammation, have been identified as independent predictors of major cardiovascular events and are closely related to the severity of tissue damage and inflammatory response [16,17]. Since lipid-related residual risk and inflammation may act through overlapping and mutually reinforcing mechanisms, an index that integrates RC and hsCRP can provide additional information on cardiometabolic risk.
The RC inflammatory index (RCII) combines estimated RC and hsCRP to reflect joint exposure to remnant lipid burden and inflammation. However, evidence regarding the association between RCII and incident CMM remains limited. Using data from the China Health and Retirement Longitudinal Study (CHARLS), this study examined whether baseline RCII was associated with the risk of developing CMM among middle-aged and older Chinese individuals.
Material and Methods
Study Design and Population
The cohort data used in this study were derived from the CHARLS (https://charls.pku.edu.cn). This survey assesses the health, nutrition, and socioeconomic status of families and individuals aged 45 years and older in China. CHARLS follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for observational studies, and was approved by the Biomedical Ethics Review Board of Peking University (IRB 00001052–11015).
Informed written consent was obtained from all participants before their inclusion in the study [18].
The CHARLS National Baseline Survey began from June 2011 to March 2012, with participants undergoing regular follow-up visits every 2 years through face-to-face questionnaires. The questionnaires were conducted by trained interviewers using computer-aided technology to ensure standardized data collection, and routine follow-up surveys were conducted in 2013, 2015, 2018, and 2020. Only the first (2011) and third (2015) rounds included blood sample data; only in those years were blood samples collected for laboratory testing [19].
Participants interviewed between 2011 and 2012 were considered as baseline cohorts, and blood sample data from the first round of CHARLS (2011) were used to measure total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), LDL-C, and hsCRP.
The inclusion and exclusion criteria for this study are illustrated in a flowchart (Figure 1). Exclusion criteria included: (1) baseline age less than 45 years; (2) missing CMM diagnosis or presence of CMM at baseline; (3) missing data on hypertension, diabetes, cardiovascular disease, and stroke at baseline; (4) missing baseline data for TC, HDL-C, LDL-C, or hsCRP; (5) missing or unavailable CMM diagnostic data during follow-up. The final analysis included 6173 participants from the CHARLS cohort.
Assessment and Definition of Cardiometabolic Multimorbidity
The primary endpoint of this study is the incidence of CMM, defined as the coexistence of 2 or more of the following CMDs: hypertension, diabetes, heart disease, or stroke [20]. Diagnostic criteria for hypertension included office blood pressure ≥ 140/90 mm Hg or home blood pressure ≥ 135/85 mm Hg in the absence of antihypertensive medications, or self-reported and physician-diagnosed, current use of antihypertensive medications (“Have you been diagnosed with hypertension by your doctor?” and “Do you take any medications for high blood pressure?”). Criteria for diagnosing diabetes were as follows: fasting blood glucose ≥ 126 mg/dL, glycosylated hemoglobin (HbA1c) level ≥ 6.5%, current use of antidiabetic medications, or self-reported diagnosis (“Have you been diagnosed by your doctor with diabetes or hyperglycemia?” and “Do you take any medications for diabetes or hyperglycemia?”).
Identification of participants’ history of heart disease or stroke was primarily based on self-reported data collected in the baseline questionnaire and follow-up surveys (“Have you ever been diagnosed by a physician with coronary heart disease, angina, congestive heart failure, or any other heart problem?” and “Have you ever been diagnosed with a stroke by a doctor?”).
Calculation of Remnant Cholesterol and Remnant Cholesterol Inflammatory Index
In the CHARLS cohort, hsCRP (mg/L) was used, and RC (mg/dL) was calculated as: RC = TC − (HDL-C + LDL-C). RCII is calculated as RCII = RC × hsCRP/10 [21]. Division by 10 was used as a scaling factor to obtain a more convenient numerical range and does not indicate a clinically validated threshold.
Due to the skewed distribution of RCII, natural logarithmic transformation was applied to RCII values to yield lnRCII for statistical analysis. All included participants had positive RCII values.
Collection of Covariates
Covariates collected included age, sex, education, place of residence, body mass index (BMI), marital status, alcohol consumption, smoking status, nighttime sleep duration, napping habits, physical activity levels, mental health assessment, and clinical comorbidities (dyslipidemia, gastric disease, and kidney disease, among others). Comorbidity data were obtained via self-report questionnaires. Smoking status was assessed through the question “Are you smoking now?” and “Have you ever smoked?” Alcohol consumption status was determined by the question: “Have you ever consumed alcohol before?” BMI was calculated as weight (kg) divided by height (m) squared.
Statistical Analysis
Descriptive analysis and difference comparisons were performed on baseline characteristics of eligible participants. Normally distributed quantitative variables were described using means and standard deviations (SD), and analysis of variance (ANOVA) was used to compare the differences between groups. Quantitative variables with skewed distributions were described using medians and interquartile ranges and compared using the Kruskal-Wallis H test.
Categorical variables were described using frequencies and percentages and compared using the chi-square test, while ordinal variables were analyzed using the Kruskal-Wallis H test. Univariate modified Poisson regression analysis was used for all eligible participants for the association with the risk of CMM development. Next, variables with P < 0.05 from between-group comparisons, as well as covariates, were included as confounding factors in a multivariate modified Poisson regression.
The receiver operating characteristic (ROC) curve was constructed according to the modified Poisson regression model to evaluate the predictive ability and accuracy of lnRCII on the risk of CMM in patients, and compared with the ROC curves for RC and hsCRP. To compare the area under the curve (AUC) values among lnRCII, RC, and hsCRP, a post-hoc DeLong test was performed as a supplementary exploratory analysis. This test was not prespecified in the original statistical analysis plan but was conducted to provide a formal statistical comparison of the discriminatory abilities of the 3 biomarkers.
Additionally, the generalized variance inflation factor (VIF) for covariates in the multivariate modified Poisson regression was calculated to test for multicollinearity. Variables such as education level, marital status, hyperlipidemia, and nighttime sleep duration had missing values. Missing data were handled using Knearest neighbors (KNN) imputation implemented with the VIM package in R.
The optimal number of neighbors (k = 5) was selected via fold crossvalidation, minimizing the average rootmeansquare error. Since the proportion of missing data on exercise at the baseline survey was > 40%, it was not considered as a covariate, and sensitivity analyses were performed by excluding participants with any missing data in the variables. All eligible participants were analyzed in subgroups according to sex, age, type of residence, marital status, education level, alcohol consumption, BMI, napping, and hyperlipidemia.
Sensitivity analysis was performed by excluding participants with hyperlipidemia, cancer, or CMD at baseline. Finally, multivariate binary logistic regression was conducted, and nonlinear relationships were assessed using restricted cubic spline (RCS) curves. All statistical analyses were performed using R version 4.4.2, including the R packages VIM, dplyr, sandwich, and forestploter.
Results
Basic Characteristics of Participants
This study included 6173 participants in the CHARLS survey from 2011 to 2020. The mean age was 58.14±8.62 years, 55.8% of the participants were female, and the mean lnRCII value was 0.73±1.47 (Table 1). In the cohort, older age, higher BMI, and urban residence were associated with higher lnRCII levels (P < 0.001).
There were no significant differences in sex, education level, and marital status among the groups. The prevalence of hypertension and hyperlipidemia was higher in the group falling in the lowest quartile range of lnRCII levels (P < 0.001), but the prevalence of digestive disorders was highest in the lowest RCII group. Systolic blood pressure, diastolic blood pressure, triglycerides, blood glucose, and uric acid all increased significantly with increased RCII (P < 0.001).
Serum urea nitrogen, HDL-C, and estimated glomerular filtration rate (eGFR) decreased with increased RCII (P < 0.001). Values were missing for some variables (Table 2).
Association Between Natural Log of Remnant Cholesterol Inflammatory Index and Cardiometabolic Multimorbidity
In the CHARLS cohort, a total of 1832 cases (29.68%) of CMM occurred, and the incidence of CMM was significantly higher in the higher lnRCII quartile groups compared with the lower lnRCII quartile groups (P < 0.001). Model 1 was unadjusted. Model 2 was adjusted for some demographic characteristics and activities of daily living.
Model 3 was further adjusted for treatment for hypertension, dyslipidemia, eGFR, alcohol consumption, napping, and psychiatric problems. As shown in model 1 (Table 3), the relative risk ratios (RR) of each quartile compared with quartile 1 (Q1) were 1.36 (95% CI: 1.17~1.53) for quartile 2 (Q2), 1.70 (95% CI: 1.50~1.92) for quartile 3 (Q3), and 2.01 (95% CI: 1.79~2.26) for quartile 4 (Q4). In model 2, the association between lnRCII and the risk of CMM remained significant.
In model 3, RR values for each quartile were: 1.26 (95% CI: 1.10~1.43) for Q2, 1.54 (95% CI: 1.36~1.74) for Q3, and 1.69 (95% CI: 1.50~1.90) for Q4. Furthermore, the risk of CMM increased by 13% for each unit increase in lnRCII (all P < 0.001). In contrast with RCII, the RR values for RC and hsCRP were both 1.01 in model 3 (all P < 0.001) (Table 4).
According to the multivariate modified Poisson regression model, the ROC and AUC were plotted and calculated (Figure 2). The AUCs were 0.609 (95% CI: 0.594–0.624) for lnRCII, 0.577 (95% CI: 0.561–0.592) for RC, and 0.590 (95% CI: 0.574–0.605) for hsCRP, suggesting that lnRCII had modest discriminatory ability for incident CMM. The DeLong test revealed that lnRCII exhibited a significantly higher AUC than RC or hsCRP alone (all P<0.001).
All VIF values were below 5.
Nonlinear Relationship Between Natural Log of Remnant Cholesterol Inflammatory Index and Cardiometabolic Multimorbidity
The RCS models, fully adjusted for covariates (Model 3), revealed a linear association between lnRCII and the incidence of CMM (P for nonlinearity = 0.135, Figure 3).
Subgroup Analysis and Interaction Analysis
To further explore the relationship between RCII and CMM risk, subgroup analyses were performed across various demographic and clinical characteristics. The P value for interaction was used to test for heterogeneity across subgroups. Because each interaction test corresponds to a single hypothesis, no adjustment for multiple testing was applied.
The results demonstrated that the association between elevated lnRCII levels and CMM risk was consistent across different subgroups (Figure 4). The relationship between lnRCII and CMM risk was stronger in middle-aged participants younger than 60 years (P for interaction < 0.001).
Sensitivity Analysis
Modified Poisson regression was performed on complete cases with missing data removed, and the results were consistent with the data imputed by KNN (Table 5). The results of the fully adjusted logistic regression model (Table 6) showed that the odds ratio (OR) for lnRCII was 1.21 (95% CI: 1.16–1.26). The ORs for each quartile were 1.40 (95% CI: 1.17–1.67) for Q2, 1.85 (95% CI: 1.56–2.20) for Q3, and 2.28 (95% CI: 1.92–2.71) for Q4, compared with Q1.
These logistic regression results demonstrate the robustness of the primary analysis findings. Additionally, after excluding participants with hyperlipidemia, cancer, or CMD at baseline, multivariate Poisson regression analyses were performed again, and the results remained consistent with the primary analysis (Table 7), further confirming the reliability of our findings.
Discussion
This study is based on a prospective cohort study in a Chinese population aged 45 years and older to evaluate the association between RCII and CMM. Modified Poisson regression found that lnRCII was positively and independently associated with the risk of CMM. This analysis provided a more comprehensive risk assessment than the assessment of RC and hsCRP alone.
Subgroup analyses showed a stronger association between lnRCII and the risk of CMM in middle-aged participants. The results of the sensitivity analysis also demonstrated the robustness of the relationship between lnRCII and CMM risk.
The construction of the RCII is based on the synergistic pathogenic mechanism of RC and inflammatory markers (CRP/hsCRP); that is, lipid metabolism dysfunction and inflammation are interrelated in the pathogenesis and comorbidity of chronic diseases [21–24]. The core of lipid metabolism dysfunction is triglycerides and LDL-C. Statins are mainly used to reduce LDL-C levels to improve dyslipidemia [8,25].
However, residual cardiovascular risk persists despite high intensity statin therapy, as inflammation, atherosclerosis, and thrombosis can still be induced. Existing studies have shown that high lipoprotein(a) and triglycerides contribute to this residual risk, and current treatments are not sufficiently effective at lowering lipoprotein(a) [25,26].
The remnant risk value of RC has not been consistent across studies. Li et al found that the greater the variability of RC, the higher the risk of ischemic stroke, whereas Yun et al found that RC was positively associated with an increased risk of prognosis in patients with coronary heart disease, but that this increased risk was not statistically significant in terms of risk of all-cause mortality. Moreover, Chen et al found a strong association between RC and cancer mortality but not cardiovascular mortality in the National Health and Nutrition Examination Survey [27–29].
Some studies have shown that RC enhances the immune response and may have a protective effect. In the present study, RC showed a positive correlation with CMM risk after being classified by quartiles.
In the present study, urban dwellers, current or former smokers, individuals with high Center for Epidemiologic Studies Depression Scale-10 (CESD-10) scores, and those with abnormal body weight all increased the risk of CMM, as did participants with low eGFR, consistent with previous studies [15]. The inverted L-shaped association between RCII calculated using hsCRP and the incidence of CMM was transformed into a linear positive correlation after natural logarithmic transformation, and the risk of CMM increased by 13% for each unit increase in lnRCII.
RCII can cover both abnormal lipid metabolism and inflammatory responses, enhancing the identification of risk indicators. RC promotes lipid accumulation through triglycerides and low-density lipoprotein and triggers oxidative stress and endothelial dysfunction leading to atherosclerosis, while CRP/hsCRP amplifies these effects by promoting vascular injury, monocyte recruitment, and inflammation-driven angiogenesis. In the present study, the RR values of lnRCII transformed by natural logarithmic transformation were always higher than those of RC and hsCRP, regardless of whether they were continuous or categorical.
In the present study, data from 6173 participants in the CHARLS were analyzed. The results of the study showed a significant correlation between elevated RCII levels and the prevalence of CMM, and RCII was better able to identify CMM risk than RC or hsCRP alone. This study also showed that the association between RCII and CMM risk was broadly consistent between sexes, with slightly higher RR in men after adjusted by covariates.
Previous research has shown that men generally exhibit a greater risk of cardiometabolic disease than women [19]. In addition, middle-aged people exhibit higher RR values compared with older adults, so it is critical to pay attention to RCII levels in this population.
Previous studies have focused on indicators and stroke risk and mortality risk [28–30], whereas evidence regarding CMM remains limited. The present study extended its application to the risk of cardiometabolic disease and comorbidities, confirming that higher baseline RCII levels are associated with an increased risk of CMM.
Limitations
Several limitations should be acknowledged. First, due to differences in the definition of CMM, with some studies defining CMM as 2 or more conditions in combination with type 2 diabetes, ischemic heart disease, stroke, or hypertension [31–33], the incidence of CMM in different studies has varied with these different definitions. Second, although the confidence in self-reported disease events in the CHARLS questionnaire has been high [34], self-reporting has still led to possible misclassification of outcomes.
The lack of classification of cardiovascular disease and stroke subtypes may mask the pathophysiology and risk factors between different subtypes (ischemic and non-ischemic). Third, RC was estimated using TC, HDL-C, and LDL-C rather than directly measured remnant lipoprotein cholesterol, and this derived measure may be affected by measurement error. Fourth, RCII was measured only at baseline, preventing assessment of longitudinal changes in RCII.
In addition, although confounders have been adjusted for in studies, confounder bias cannot be avoided and needs to be verified by randomized controlled trials in the future. Finally, ROC analysis showed only modest discrimination, and no calibration or reclassification metrics were evaluated. Therefore, it is worth further expanding the validation in future studies.
Conclusions
The results of this study demonstrate that RCII is significantly associated with CMM incidence. By combining assessment of lipid metabolism and inflammatory markers, RCII provides a more comprehensive assessment of chronic comorbidity risk in middle-aged and advanced-age Chinese adults and enables identification of high-risk groups for cardiometabolic comorbidities. However, this study considered only baseline RCII levels, and its clinical utility and practical application require further validation in larger, multiethnic, long-term prospective cohort studies.