Abstract
Background
Non-traditional lipid indices are diabetes risk factors, but their combined effects with inflammatory markers remain unclear. The present study systematically investigated and compared the associations between eight non-traditional lipid-inflammatory parameters and new-onset diabetes.
Methods
Data were obtained from the 2011–2020 China Health and Retirement Longitudinal Study. The lipid-inflammatory parameters were created by integrating non-traditional lipid indices and high-sensitivity C-reactive protein (hsCRP). Cox regression and restricted cubic spline models examined the associations of baseline and cumulative lipid-inflammatory parameters with diabetes risk. Time-dependent receiver operating characteristic analyses evaluated the predictive performance of these parameters. Mediation analyses explored the reciprocal links between non-traditional lipid indices, hsCRP, and diabetes risk.
Results
The study enrolled 7,356 participants, with 1,173 (15.95%) developing diabetes over a median follow-up period of 9 years. In the multivariable-adjusted model, individuals with higher baseline and cumulative non-traditional lipid-inflammatory parameter levels had an elevated risk of developing diabetes, especially for lipoprotein combined index (LCI)-hsCRP (hazard ratio [HR] = 2.03, 95% confidence interval [CI]: 1.70–2.41). All lipid-inflammatory parameters were nonlinearly associated with diabetes risk. The area under the curve (AUC) values for the eight lipid-inflammatory parameters ranged from 0.595 to 0.619, with LCI-hsCRP exhibiting the highest predictive value (AUC = 0.619, 95% CI: 0.597–0.640). The results were consistent across sensitivity and subgroup analyses. Mediation analyses showed significant bidirectional mediation between non-traditional lipid indices and hsCRP for diabetes.
Conclusions
Elevated baseline and cumulative non-traditional lipid-inflammatory parameter levels are linked to a higher risk of incident diabetes, with LCI-hsCRP showing the highest predictive value.
Clinical trial number
Not applicable.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12902-026-02219-9.
Keywords: Non-traditional lipid-inflammatory parameters, Non-traditional lipid indices, High-sensitivity C-reactive protein, Diabetes, China Health and Retirement Longitudinal Study
Introduction
Diabetes is one of the most prevalent and serious chronic diseases worldwide. According to the International Diabetes Federation, the global number of adults aged 20–79 years living with diabetes reached about 536.6 million in 2021 and is expected to rise to 783.2 million by 2045 [1]. Adults with diabetes have a two- to fourfold increased incidence of cardiovascular events and a twofold increase in all-cause mortality relative to non-diabetic individuals [2–4]. Consequently, the identification of modifiable determinants of diabetes is critical for reducing its incidence and mitigating the escalating public health burden.
Lipid metabolism disorders and inflammation are two major modifiable risk factors that play crucial roles in the pathogenesis of diabetes [5–8]. These two factors are biologically interdependent and act in concert to impair pancreatic β-cell function [9, 10], exacerbate insulin resistance [11–13], and promote islet inflammation [14], thereby contributing to the initiation and progression of diabetes. In recent years, several non-traditional lipid indices derived from conventional lipid parameters—including the atherogenic index of plasma (AIP), remnant cholesterol (RC), non-high-density lipoprotein cholesterol (Non-HDL-C), RC/HDL-C ratio, atherogenic coefficient (AC), Castelli’s risk index I (CRI-I), Castelli’s risk index II (CRI-II), and the lipoprotein combined index (LCI)—have been shown to be closely linked to insulin resistance [15, 16], prediabetes [17–19], and diabetes risk [8, 20, 21]. In parallel, high-sensitivity C-reactive protein (hsCRP), a commonly used clinical marker of systemic inflammation, has also been found to be independently associated with diabetes risk [22–24]. Although a few studies have demonstrated that the joint assessment of RC or AIP with hsCRP may improve the prediction of diabetes risk [25–29], the combined effects of other non-traditional lipid indices with hsCRP on incident diabetes remain understudied. Moreover, it is unclear which non-traditional lipid-inflammatory composite marker is most effective for predicting diabetes risk.
Thus, the present study aimed to systematically investigate and compare the associations of baseline and cumulative non-traditional lipid-inflammatory parameters with incident diabetes, using data collected in the 2011–2020 waves of the China Health and Retirement Longitudinal Study (CHARLS) [30]. Meanwhile, we conducted mediation analyses to examine the mutual mediating roles of non-traditional lipid indices and hsCRP in relation to diabetes risk.
Methods
Study design and participants
Data were obtained from the CHARLS, a national prospective cohort of middle-aged and older adults in China. Detailed descriptions of the study design and sampling procedures have been reported previously [30]. Briefly, participants in CHARLS were recruited using a multistage probability sampling strategy. The baseline survey (wave 1), carried out from June 2011 to March 2012, included 17,708 participants from 450 villages or urban communities across 28 provinces in China. Follow-up assessments have been conducted every two years, with minimal new recruitment at each wave. To date, four follow-up waves have been completed in 2013, 2015, 2018, and 2020, corresponding to Waves 2–5. At each wave, the social, economic, and health circumstances were collected by trained interviewers.
The current analysis used data from the 2011, 2013, 2015, 2018, and 2020 waves. After exclusion of 7,081 participants without fasting blood tests or with unknown fasting status, 294 participants lacking baseline data on lipids, hsCRP, fasting plasma glucose (FPG), or glycated hemoglobin (HbA1c), 351 participants who were aged < 45 years or had missing age data, 1,639 participants with a history of diabetes, 131 participants who had cancer, and 856 participants lacking complete follow-up data, a final cohort of 7,356 participants was obtained. To investigate the association of cumulative non-traditional lipid-inflammatory parameters with incident diabetes, the cohort was further restricted to 3,544 participants who had fasting blood tests and complete data on lipids, hsCRP, FPG, and HbA1c at both Wave 1 and Wave 3, were free of diabetes or a history of cancer at or before Wave 3, and had complete follow-up data. Figure 1 illustrates the participant selection flowchart.
Fig. 1.
Flowchart of study participants. Abbreviations: hsCRP, high-sensitivity C-reactive protein; FPG, fasting plasma glucose; HbA1c, glycated hemoglobin
Exposure
In accordance with the standard protocol, medical staff from the Chinese Center for Disease Control and Prevention acquired and analyzed participants’ fasting venous blood samples. Blood lipid profiles, including triglycerides (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C) were quantified via enzymatic colorimetric assays. Serum hsCRP levels were measured by an immunoturbidimetric assay. Detailed information on assay methods, coefficients of variation, and detection limits for lipid profiles and hsCRP is provided in Table S1. Based on previous studies [8, 20, 21, 31], the non-traditional lipid indices were calculated using the formulas below: AIP = lg(TG/HDL-C), RC = TC − (HDL-C + LDL-C), Non-HDL-C = TC − HDL-C, RC/HDL-C ratio = RC/HDL-C, AC = Non-HDL-C/HDL-C, CRI-I = TC/HDL-C, CRI-II = LDL-C/HDL-C, and LCI = TC × TG × LDL-C/HDL-C. A lipid-inflammatory parameter was calculated by multiplying each non-traditional lipid index by hsCRP as follows: lipid-inflammatory parameter = non-traditional lipid index (mg/dL) × hsCRP (mg/L) / 10. The scaling factor of 10 was applied to standardize the units between non-traditional lipid indices and hsCRP, ensuring consistency and enabling a more precise analysis of their interaction. This method for calculating non-traditional lipid-inflammatory parameters has been widely used in previous studies [25–27, 31, 32], which have shown that integrating lipid and inflammatory indices into a composite marker can enhance the prediction of diabetes and other chronic diseases. In this study, baseline non-traditional lipid-inflammatory parameters were derived from blood lipid profiles (TC, TG, LDL-C, and HDL-C) and hsCRP measured in Wave 1 (2011–2012). Cumulative non-traditional lipid-inflammatory parameters were calculated based on blood lipid profiles (TC, TG, LDL-C, and HDL-C) and hsCRP from both Wave 1 (2011–2012) and Wave 3 (2015). Specifically, the cumulative lipid-inflammatory parameter was calculated using the trapezoidal rule for the area under the curve (AUC) as follows: cumulative lipid-inflammatory parameter = [(lipid-inflammatory parameter2012 + lipid-inflammatory parameter2015) / 2] × (2015 − 2012). This trapezoidal approximation approach adheres to the methodology used to construct cumulative metabolic indices in the CHARLS cohorts, which is consistent with the approach described in previous studies [31, 32].
Outcome
Incident diabetes was our primary outcome. Based on the American Diabetes Association criteria [33], a participant was considered to have diabetes if any of the following criteria were met: (1) FPG ≥ 126 mg/dL; (2) HbA1c ≥ 6.5%; (3) self-reported diabetes diagnosed by a physician; and (4) use of glucose-lowering drugs or insulin. The timing of diabetes onset was determined as the midpoint between the follow-up visit at which diabetes was first identified and the immediately preceding follow-up visit. Survival time was calculated according to participants’ diabetes status: for participants who remained free of diabetes, it was the interval from baseline to the final follow-up or study end date; for participants with incident diabetes, it was the interval from baseline to the date of diabetes onset.
Covariates
The study covariates comprised age, sex (male or female), place of residence (urban or rural), educational attainment (below high school versus high school or above), marital status (married or partnered versus all other statuses, such as separated, divorced, widowed, or never married), smoking status (never smoker or ever smoker, the latter comprising both past and current smokers), alcohol use (never drinker or ever drinker, the latter comprising both past and current drinkers), body mass index (BMI), hypertension, heart disease, stroke, history of dyslipidemia medication use, and serum uric acid levels. Participants were classified as having hypertension if they met any of the following conditions: (1) mean systolic blood pressure (SBP) ≥ 140 mmHg; (2) mean diastolic blood pressure (DBP) ≥ 90 mmHg; (3) self-reported physician-diagnosed hypertension; or (4) use of blood pressure-lowering medications. Heart disease (comprising myocardial infarction, angina pectoris, congestive heart failure, coronary heart disease, or other cardiac conditions) and stroke were defined based on self-reported physician diagnoses.
Statistical analyses
Participants were stratified according to diabetes status for group comparisons. Normally distributed continuous variables were presented as mean with standard deviation (SD) and compared between groups using two-sample t-tests. Non-normally distributed continuous variables were summarized as median with interquartile range (IQR) and compared between groups using the Wilcoxon rank-sum test. Categorical variables were reported as counts with percentages and compared between groups using Pearson’s chi-squared test. To handle missing data, the multiple imputation by chained equations method was performed with the “mice” package [34]. A total of five imputed datasets were generated. Table S2 summarizes the distribution of missing data and the corresponding imputation methods.
The cumulative incidence of diabetes across quartiles of each lipid-inflammatory parameter was estimated using Kaplan-Meier survival analysis, with intergroup differences examined by the log-rank test. Four Cox proportional hazards regression models were fitted to investigate the associations of baseline and cumulative non-traditional lipid-inflammatory parameters with incident diabetes, with hazard ratios (HR) and 95% confidence intervals (CI) calculated. The unadjusted model did not include any covariates. Model 1 included adjustments for age and sex. Model 2 further controlled for marital status, residence, education level, smoking status, and alcohol consumption. Model 3 additionally accounted for BMI, hypertension, heart disease, stroke, history of dyslipidemia medication use, and uric acid. Multicollinearity among covariates was evaluated using variance inflation factors (VIF). All VIF values were below 5, indicating no significant multicollinearity problems (Table S3). Additionally, the dose-response associations of non-traditional lipid-inflammatory parameters with incident diabetes were examined using restricted cubic splines (RCS) models.
Time-dependent receiver operating characteristic (ROC) analyses were conducted at the 9-year follow-up to assess each non-traditional lipid-inflammatory parameter’s predictive ability for diabetes risk, with the AUC, sensitivity, specificity, and optimal cutoff values estimated. To evaluate the incremental predictive value of these parameters, the net reclassification improvement (NRI) and integrated discrimination improvement (IDI) indices were also calculated.
Several sensitivity analyses were performed to evaluate the robustness of the results. First, participants with incomplete data were excluded. Second, participants with a follow-up duration of under 2 years were removed. Third, the primary analyses were replicated in populations with different glycemic status assessed in wave 1, with normal glycemia defined as FPG < 100 mg/dL and HbA1c < 5.7%, and prediabetes defined as FPG 100–125 mg/dL or HbA1c 5.7–6.4%. Fourth, death was considered a competing risk for diabetes, and Fine-Gray subdistribution hazards regression models were fitted. Additionally, subgroup and interaction analyses were conducted according to baseline characteristics, including age (< 60 or ≥ 60 years), sex (male or female), BMI (< 28 or ≥ 28 kg/m²), smoking status (never smokers or ever smokers), drinking status (never drinkers or ever drinkers), and hypertension status (no or yes).
Finally, mediation analyses were performed with the “CMAverse” package [35] to investigate whether hsCRP mediated the association between non-traditional lipid indices and diabetes risk, and whether non-traditional lipid indices mediated the association between hsCRP and diabetes risk. All statistical analyses were conducted using R version 4.3.1 (The R Foundation). Two-sided P < 0.05 was considered statistically significant.
Ethics approval and consent to participate
Ethical approval for the CHARLS study was granted by the Biomedical Ethics Review Board of Peking University (IRB00001052-11015), and written informed consent was obtained from all participants prior to enrollment. The study was conducted in accordance with the principles of the Declaration of Helsinki.
Results
Baseline characteristics of study participants
The final analysis included 7,356 participants, of whom 1,173 (15.95%) developed diabetes over a median follow-up period of 9.0 years. Table 1 summarizes the participants’ baseline characteristics according to diabetes status. Overall, the mean age of the 7,356 participants was 58.91 (SD 9.34) years, and 52.94% were female. Compared with non-diabetic participants, those who developed diabetes tended to have lower educational attainment and were more likely to have comorbidities, particularly hypertension, heart disease, and stroke (all P < 0.05). They also exhibited significantly elevated hsCRP, AIP, Non-HDL-C, AC, CRI-I, CRI-II, LCI, RC, and RC/HDL-C ratio (all P < 0.001). Moreover, we compared the participants’ baseline characteristics using non-imputed data (Table S4). The results were consistent with those in Table 1.
Table 1.
Baseline characteristics of the study population by diabetes status
| Variables | Total | Non-diabetes | Diabetes | P value |
|---|---|---|---|---|
| No. of participants | 7356 | 6183 | 1173 | |
| Demographics | ||||
| Age, mean (SD), years | 58.91 (9.34) | 58.83 (9.44) | 59.35 (8.79) | 0.067 |
| Sex, n (%) | 0.064 | |||
| Male | 3462 (47.06%) | 2939 (47.53%) | 523 (44.59%) | |
| Female | 3894 (52.94%) | 3244 (52.47%) | 650 (55.41%) | |
| Residence, n (%) | 0.079 | |||
| Rural | 2611 (35.49%) | 2221 (35.92%) | 390 (33.25%) | |
| Urban | 4745 (64.51%) | 3962 (64.08%) | 783 (66.75%) | |
| Education, n (%) | 0.031 | |||
| Below high school | 6573 (89.36%) | 5504 (89.02%) | 1069 (91.13%) | |
| High school or above | 783 (10.64%) | 679 (10.98%) | 104 (8.87%) | |
| Marital status, n (%) | 0.040 | |||
| Married or partnered | 6507 (88.46%) | 5490 (88.79%) | 1017 (86.70%) | |
| Other marital status | 849 (11.54%) | 693 (11.21%) | 156 (13.30%) | |
| Smoking status, n (%) | 0.208 | |||
| Never smokers | 4444 (60.41%) | 3716 (60.10%) | 728 (62.06%) | |
| Ever smokers | 2912 (39.59%) | 2467 (39.90%) | 445 (37.94%) | |
| Drinking status, n (%) | 0.678 | |||
| Never drinkers | 4469 (60.75%) | 3750 (60.65%) | 719 (61.30%) | |
| Ever drinkers | 2887 (39.25%) | 2433 (39.35%) | 454 (38.70%) | |
| BMI, median (IQR), kg/m² | 23.01 (20.71, 25.67) | 22.83 (20.60, 25.42) | 24.10 (21.49, 26.97) | < 0.001 |
| SBP, mean (SD), mmHg | 128.07 (21.01) | 127.49 (20.97) | 131.18 (20.93) | < 0.001 |
| DBP, mean (SD), mmHg | 74.84 (12.19) | 74.51 (12.16) | 76.61 (12.22) | < 0.001 |
| Comorbidities | ||||
| Hypertension, n (%) | < 0.001 | |||
| No | 4611 (62.68%) | 3997 (64.64%) | 614 (52.34%) | |
| Yes | 2745 (37.32%) | 2186 (35.36%) | 559 (47.66%) | |
| Heart disease, n (%) | < 0.001 | |||
| No | 6548 (89.02%) | 5552 (89.79%) | 996 (84.91%) | |
| Yes | 808 (10.98%) | 631 (10.21%) | 177 (15.09%) | |
| Stroke, n (%) | 0.034 | |||
| No | 7192 (97.77%) | 6055 (97.93%) | 1137 (96.93%) | |
| Yes | 164 (2.23%) | 128 (2.07%) | 36 (3.07%) | |
| Lipid-lowering drugs, n (%) | < 0.001 | |||
| No | 7052 (95.87%) | 5950 (96.23%) | 1102 (93.95%) | |
| Yes | 304 (4.13%) | 233 (3.77%) | 71 (6.05%) | |
| Antihypertensive drugs, n (%) | < 0.001 | |||
| No | 6019 (81.82%) | 5157 (83.41%) | 862 (73.49%) | |
| Yes | 1337 (18.18%) | 1026 (16.59%) | 311 (26.51%) | |
| Laboratory values | ||||
| FPG, mean (SD), mg/dL | 100.21 (11.04) | 99.41 (10.82) | 104.42 (11.24) | < 0.001 |
| HbA1c, mean (SD), % | 5.11 (0.39) | 5.09 (0.38) | 5.25 (0.41) | < 0.001 |
| TC, mean (SD), mg/dL | 192.78 (37.54) | 191.83 (37.44) | 197.76 (37.70) | < 0.001 |
| TG, median (IQR), mg/dL | 100.89 (72.57, 144.26) | 98.24 (71.68, 139.83) | 113.28 (80.54, 163.73) | < 0.001 |
| HDL-C, mean (SD), mg/dL | 52.25 (15.10) | 52.72 (15.03) | 49.76 (15.23) | < 0.001 |
| LDL-C, mean (SD), mg/dL | 117.67 (33.91) | 116.96 (33.71) | 121.39 (34.72) | < 0.001 |
| UA, mean (SD), mg/dL | 4.42 (1.22) | 4.40 (1.22) | 4.52 (1.24) | 0.002 |
| hsCRP, median (IQR), mg/L | 0.97 (0.53, 2.04) | 0.93 (0.52, 1.97) | 1.18 (0.62, 2.47) | < 0.001 |
| AIP, mean (SD) | 0.32 (0.30) | 0.30 (0.30) | 0.39 (0.32) | < 0.001 |
| Non-HDL-C, mean (SD) | 140.53 (37.08) | 139.11 (36.75) | 147.99 (37.90) | < 0.001 |
| AC, median (IQR) | 2.73 (2.04, 3.61) | 2.67 (2.02, 3.53) | 3.07 (2.22, 4.14) | < 0.001 |
| CRI-I, median (IQR) | 3.73 (3.04, 4.61) | 3.67 (3.02, 4.53) | 4.07 (3.22, 5.14) | < 0.001 |
| CRI-II, median (IQR) | 2.30 (1.76, 2.94) | 2.26 (1.73, 2.89) | 2.55 (1.91, 3.21) | < 0.001 |
| LCI, median (IQR) | 43611.30 (24550.95, 80665.47) | 41580.75 (23718.45, 76433.14) | 58317.56 (30036.50, 103027.99) | < 0.001 |
| RC, median (IQR) | 18.56 (10.82, 29.38) | 17.78 (10.44, 28.61) | 22.04 (12.76, 34.02) | < 0.001 |
| RC/HDL ratio, median (IQR) | 0.36 (0.19, 0.66) | 0.34 (0.18, 0.63) | 0.45 (0.23, 0.79) | < 0.001 |
Data are presented as number (%), mean (standard deviation [SD]), or median (interquartile range [IQR]). Group comparisons were performed using the Pearson chi-square test, the two-sample t test, or the Wilcoxon rank-sum test, as appropriate
Abbreviations: BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; FPG, fasting plasma glucose; HbA1c, glycated hemoglobin; TC, total cholesterol; TG, triglycerides; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; UA, uric acid; hsCRP, high-sensitivity C-reactive protein; AIP, atherogenic index of plasma; Non-HDL-C, non-high-density lipoprotein cholesterol; AC, atherogenic coefficient; CRI-I, Castelli’s risk index I; CRI-II, Castelli’s risk index II; LCI, lipoprotein combined index; RC, remnant cholesterol; RC/HDL-C, remnant cholesterol/high-density lipoprotein cholesterol
Associations and dose-response relationships between non-traditional lipid-inflammatory parameters and diabetes risk
Figure 2 shows the Kaplan-Meier curves, indicating that the cumulative incidence of diabetes progressively increased from Quartile 1 to Quartile 4 for each non-traditional lipid-inflammatory parameter, with significant differences across quartiles (all log-rank P < 0.0001). Table 2 presents the associations of non-traditional lipid-inflammatory parameters with the risk of incident diabetes. After full adjustment for potential confounders in Model 3, participants in Quartile 4 of each parameter exhibited a significantly higher risk of developing diabetes compared with those in Quartile 1. Among all parameters, LCI-hsCRP showed the strongest association, with a two-fold higher risk of diabetes in the highest quartile (HR = 2.03, 95% CI: 1.70–2.41). Multivariable-adjusted RCS analyses revealed that all lipid-inflammatory parameters were nonlinearly associated with diabetes risk (Fig. 3, all P for nonlinearity < 0.001). The specific risk thresholds identified in the RCS curves were 0.02 for AIP-hsCRP, 13.76 for Non-HDL-C-hsCRP, 0.28 for AC-hsCRP, 0.38 for CRI-I-hsCRP, 0.23 for CRI-II-hsCRP, 4698.93 for LCI-hsCRP, 1.86 for RC-hsCRP, and 0.04 for the RC/HDL-C ratio-hsCRP (Fig. 3).
Fig. 2.
Kaplan-Meier plot of cumulative incidence of new-onset diabetes based on non-traditional lipid-inflammatory parameters levels. Abbreviations: hsCRP, high-sensitivity C-reactive protein; AIP, atherogenic index of plasma; Non-HDL-C, non-high-density lipoprotein cholesterol; AC, atherogenic coefficient; CRI-I, Castelli’s risk index I; CRI-II, Castelli’s risk index II; LCI, lipoprotein combined index; RC, remnant cholesterol; RC/HDL-C, remnant cholesterol/high-density lipoprotein cholesterol
Table 2.
Associations of non-traditional lipid-inflammatory parameters with the risk of incident diabetes
| Incidence rate (%) | Unadjusted | Model 1 | Model 2 | Model 3 | |||||
|---|---|---|---|---|---|---|---|---|---|
| HR (95% CI) | P value | HR (95% CI) | P value | HR (95% CI) | P value | HR (95% CI) | P value | ||
| AIP-hsCRP | |||||||||
| Quartile 1 | 12.6 | Reference | — | Reference | — | Reference | — | Reference | — |
| Quartile 2 | 13.8 | 1.11 (0.92–1.33) | 0.278 | 1.11 (0.92–1.34) | 0.261 | 1.13 (0.94–1.35) | 0.206 | 1.10 (0.92–1.33) | 0.302 |
| Quartile 3 | 18.0 | 1.46 (1.23–1.74) | < 0.001 | 1.47 (1.23–1.75) | < 0.001 | 1.50 (1.26–1.79) | < 0.001 | 1.39 (1.16–1.66) | < 0.001 |
| Quartile 4 | 24.2 | 2.06 (1.75–2.43) | < 0.001 | 2.04 (1.73–2.41) | < 0.001 | 2.10 (1.78–2.48) | < 0.001 | 1.85 (1.56–2.20) | < 0.001 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | < 0.001 | |||||
| Non-HDL-C-hsCRP | |||||||||
| Quartile 1 | 12.8 | Reference | — | Reference | — | Reference | — | Reference | — |
| Quartile 2 | 14.9 | 1.18 (0.98–1.41) | 0.073 | 1.17 (0.97–1.40) | 0.095 | 1.17 (0.98–1.40) | 0.086 | 1.13 (0.94–1.35) | 0.185 |
| Quartile 3 | 18.8 | 1.53 (1.29–1.82) | < 0.001 | 1.51 (1.27–1.79) | < 0.001 | 1.53 (1.29–1.82) | < 0.001 | 1.41 (1.19–1.68) | < 0.001 |
| Quartile 4 | 22.1 | 1.84 (1.55–2.17) | < 0.001 | 1.79 (1.52–2.12) | < 0.001 | 1.81 (1.53–2.14) | < 0.001 | 1.61 (1.35–1.91) | < 0.001 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | < 0.001 | |||||
| AC-hsCRP | |||||||||
| Quartile 1 | 12.2 | Reference | — | Reference | — | Reference | — | Reference | — |
| Quartile 2 | 15.0 | 1.25 (1.04–1.50) | 0.016 | 1.25 (1.04–1.50) | 0.017 | 1.26 (1.05–1.51) | 0.014 | 1.22 (1.01–1.46) | 0.035 |
| Quartile 3 | 18.8 | 1.60 (1.34–1.90) | < 0.001 | 1.59 (1.34–1.89) | < 0.001 | 1.62 (1.36–1.93) | < 0.001 | 1.50 (1.25–1.78) | < 0.001 |
| Quartile 4 | 22.6 | 1.97 (1.67–2.33) | < 0.001 | 1.94 (1.64–2.30) | < 0.001 | 1.97 (1.66–2.34) | < 0.001 | 1.75 (1.47–2.08) | < 0.001 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | < 0.001 | |||||
| CRI-I-hsCRP | |||||||||
| Quartile 1 | 12.5 | Reference | — | Reference | — | Reference | — | Reference | — |
| Quartile 2 | 14.7 | 1.19 (0.99–1.43) | 0.059 | 1.19 (0.99–1.42) | 0.065 | 1.20 (1.00–1.44) | 0.053 | 1.16 (0.96–1.39) | 0.116 |
| Quartile 3 | 19.3 | 1.61 (1.36–1.91) | < 0.001 | 1.60 (1.35–1.90) | < 0.001 | 1.63 (1.37–1.94) | < 0.001 | 1.50 (1.26–1.78) | < 0.001 |
| Quartile 4 | 22.0 | 1.86 (1.58–2.20) | < 0.001 | 1.83 (1.55–2.17) | < 0.001 | 1.86 (1.57–2.20) | < 0.001 | 1.65 (1.39–1.96) | < 0.001 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | < 0.001 | |||||
| CRI-II-hsCRP | |||||||||
| Quartile 1 | 12.6 | Reference | — | Reference | — | Reference | — | Reference | — |
| Quartile 2 | 14.6 | 1.18 (0.98–1.42) | 0.075 | 1.17 (0.98–1.41) | 0.085 | 1.18 (0.99–1.42) | 0.072 | 1.15 (0.96–1.38) | 0.134 |
| Quartile 3 | 18.8 | 1.56 (1.32–1.86) | < 0.001 | 1.55 (1.30–1.84) | < 0.001 | 1.58 (1.33–1.88) | < 0.001 | 1.46 (1.23–1.74) | < 0.001 |
| Quartile 4 | 22.6 | 1.91 (1.62–2.26) | < 0.001 | 1.88 (1.59–2.22) | < 0.001 | 1.91 (1.61–2.26) | < 0.001 | 1.70 (1.43–2.02) | < 0.001 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | < 0.001 | |||||
| LCI-hsCRP | |||||||||
| Quartile 1 | 11.9 | Reference | — | Reference | — | Reference | — | Reference | — |
| Quartile 2 | 14.6 | 1.25 (1.04–1.51) | 0.017 | 1.24 (1.03–1.49) | 0.022 | 1.25 (1.04–1.50) | 0.018 | 1.20 (0.99–1.44) | 0.057 |
| Quartile 3 | 16.8 | 1.46 (1.22–1.75) | < 0.001 | 1.44 (1.20–1.72) | < 0.001 | 1.46 (1.22–1.75) | < 0.001 | 1.35 (1.12–1.62) | 0.001 |
| Quartile 4 | 25.2 | 2.31 (1.96–2.73) | < 0.001 | 2.27 (1.92–2.68) | < 0.001 | 2.31 (1.95–2.73) | < 0.001 | 2.03 (1.70–2.41) | < 0.001 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | < 0.001 | |||||
| RC-hsCRP | |||||||||
| Quartile 1 | 12.4 | Reference | — | Reference | — | Reference | — | Reference | — |
| Quartile 2 | 15.1 | 1.24 (1.03–1.48) | 0.021 | 1.24 (1.03–1.48) | 0.021 | 1.24 (1.04–1.49) | 0.019 | 1.20 (1.00–1.44) | 0.046 |
| Quartile 3 | 18.7 | 1.58 (1.33–1.87) | < 0.001 | 1.57 (1.32–1.86) | < 0.001 | 1.59 (1.33–1.89) | < 0.001 | 1.44 (1.21–1.72) | < 0.001 |
| Quartile 4 | 22.4 | 1.94 (1.64–2.29) | < 0.001 | 1.91 (1.61–2.26) | < 0.001 | 1.94 (1.63–2.29) | < 0.001 | 1.70 (1.43–2.03) | < 0.001 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | < 0.001 | |||||
| RC/HDL-C ratio-hsCRP | |||||||||
| Quartile 1 | 12.0 | Reference | — | Reference | — | Reference | — | Reference | — |
| Quartile 2 | 14.6 | 1.23 (1.02–1.48) | 0.028 | 1.23 (1.03–1.48) | 0.025 | 1.24 (1.03–1.49) | 0.022 | 1.21 (1.01–1.46) | 0.039 |
| Quartile 3 | 18.9 | 1.64 (1.38–1.95) | < 0.001 | 1.64 (1.38–1.95) | < 0.001 | 1.67 (1.40–1.98) | < 0.001 | 1.54 (1.29–1.83) | < 0.001 |
| Quartile 4 | 23.0 | 2.05 (1.73–2.43) | < 0.001 | 2.04 (1.72–2.41) | < 0.001 | 2.08 (1.75–2.46) | < 0.001 | 1.85 (1.55–2.20) | < 0.001 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | < 0.001 | |||||
Model 1 was adjusted for age and sex; Model 2 was additionally adjusted for marital status, residence, education, smoking, and drinking; Model 3 was further adjusted for body mass index, hypertension, heart disease, stroke, history of dyslipidemia medication use, and uric acid
Abbreviations: HR, hazard ratio; CI, confidence interval; hsCRP, high-sensitivity C-reactive protein; AIP, atherogenic index of plasma; Non-HDL-C, non-high-density lipoprotein cholesterol; AC, atherogenic coefficient; CRI-I, Castelli’s risk index I; CRI-II, Castelli’s risk index II; LCI, lipoprotein combined index; RC, remnant cholesterol; RC/HDL-C, remnant cholesterol/high-density lipoprotein cholesterol
Fig. 3.
Dose-response relationship between non-traditional lipid-inflammatory parameters and the risk of incident diabetes. Models were adjusted for age, sex, residence, education level, marital status, smoking status, drinking status, body mass index, hypertension, heart disease, stroke, history of dyslipidemia medication use, and uric acid. Abbreviations: HR, hazard ratio; CI, confidence interval; hsCRP, high-sensitivity C-reactive protein; AIP, atherogenic index of plasma; Non-HDL-C, non-high-density lipoprotein cholesterol; AC, atherogenic coefficient; CRI-I, Castelli’s risk index I; CRI-II, Castelli’s risk index II; LCI, lipoprotein combined index; RC, remnant cholesterol; RC/HDL-C, remnant cholesterol/high-density lipoprotein cholesterol
Using identical analytical methods, the associations of cumulative non-traditional lipid-inflammatory parameters with incident diabetes were also examined (Figure S1 and Table 3). In the multivariable-adjusted model (Model 3), all analyzed cumulative parameters were positively associated with diabetes risk when comparing the Quartile 4 with the Quartile 1, most notably cumulative LCI-hsCRP (HR = 2.75, 95% CI: 1.84–4.12). RCS analyses further demonstrated that all cumulative lipid-inflammatory parameters were nonlinearly associated with diabetes risk (Figure S2, all P for nonlinearity < 0.05). The specific risk thresholds identified in the RCS curves were 0.11 for cumulative AIP-hsCRP, 50.69 for cumulative Non-HDL-C-hsCRP, 1.02 for cumulative AC-hsCRP, 1.41 for cumulative CRI-I-hsCRP, 0.81 for cumulative CRI-II-hsCRP, 16978.74 for cumulative LCI-hsCRP, 8.73 for cumulative RC-hsCRP, and 0.17 for cumulative RC/HDL-C ratio-hsCRP (Figure S2).
Table 3.
Associations of cumulative non-traditional lipid-inflammatory parameters with the risk of incident diabetes
| Incidence rate (%) | Unadjusted | Model 1 | Model 2 | Model 3 | |||||
|---|---|---|---|---|---|---|---|---|---|
| HR (95% CI) | P value | HR (95% CI) | P value | HR (95% CI) | P value | HR (95% CI) | P value | ||
| Cumulative AIP-hsCRP | |||||||||
| Quartile 1 | 4.6 | Reference | — | Reference | — | Reference | — | Reference | — |
| Quartile 2 | 6.6 | 1.46 (0.97–2.19) | 0.070 | 1.47 (0.98–2.21) | 0.064 | 1.52 (1.01–2.28) | 0.045 | 1.46 (0.97–2.20) | 0.066 |
| Quartile 3 | 9.9 | 2.24 (1.53–3.26) | < 0.001 | 2.26 (1.55–3.30) | < 0.001 | 2.35 (1.61–3.43) | < 0.001 | 2.15 (1.46–3.15) | < 0.001 |
| Quartile 4 | 10.2 | 2.29 (1.57–3.33) | < 0.001 | 2.30 (1.58–3.35) | < 0.001 | 2.42 (1.66–3.53) | < 0.001 | 2.10 (1.43–3.09) | < 0.001 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | < 0.001 | |||||
| Cumulative Non-HDL-C-hsCRP | |||||||||
| Quartile 1 | 4.9 | Reference | — | Reference | — | Reference | — | Reference | — |
| Quartile 2 | 7.8 | 1.63 (1.11–2.39) | 0.013 | 1.63 (1.11–2.39) | 0.013 | 1.63 (1.11–2.39) | 0.013 | 1.55 (1.06–2.28) | 0.025 |
| Quartile 3 | 8.1 | 1.69 (1.15–2.47) | 0.007 | 1.69 (1.15–2.47) | 0.007 | 1.73 (1.18–2.53) | 0.005 | 1.59 (1.08–2.33) | 0.019 |
| Quartile 4 | 10.3 | 2.16 (1.50–3.12) | < 0.001 | 2.16 (1.50–3.12) | < 0.001 | 2.17 (1.50–3.13) | < 0.001 | 1.89 (1.30–2.76) | < 0.001 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | 0.005 | |||||
| Cumulative AC-hsCRP | |||||||||
| Quartile 1 | 4.7 | Reference | — | Reference | — | Reference | — | Reference | — |
| Quartile 2 | 6.9 | 1.49 (1.00–2.23) | 0.048 | 1.50 (1.00–2.23) | 0.048 | 1.50 (1.01–2.24) | 0.046 | 1.46 (0.98–2.17) | 0.066 |
| Quartile 3 | 9.3 | 2.04 (1.40–2.97) | < 0.001 | 2.03 (1.39–2.97) | < 0.001 | 2.09 (1.43–3.05) | < 0.001 | 1.90 (1.30–2.79) | 0.001 |
| Quartile 4 | 10.3 | 2.26 (1.56–3.28) | < 0.001 | 2.26 (1.55–3.27) | < 0.001 | 2.30 (1.58–3.34) | < 0.001 | 2.03 (1.38–2.97) | < 0.001 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | 0.001 | |||||
| Cumulative CRI-I-hsCRP | |||||||||
| Quartile 1 | 5.1 | Reference | — | Reference | — | Reference | — | Reference | — |
| Quartile 2 | 7.0 | 1.38 (0.94–2.04) | 0.099 | 1.38 (0.94–2.04) | 0.101 | 1.39 (0.94–2.04) | 0.098 | 1.33 (0.90–1.95) | 0.154 |
| Quartile 3 | 9.3 | 1.85 (1.28–2.67) | 0.001 | 1.85 (1.28–2.67) | 0.001 | 1.89 (1.31–2.73) | < 0.001 | 1.71 (1.18–2.48) | 0.005 |
| Quartile 4 | 9.7 | 1.93 (1.34–2.78) | < 0.001 | 1.92 (1.34–2.77) | < 0.001 | 1.95 (1.35–2.81) | < 0.001 | 1.70 (1.17–2.48) | 0.005 |
| P for trend | < 0.001 | 0.001 | < 0.001 | 0.013 | |||||
| Cumulative CRI-II-hsCRP | |||||||||
| Quartile 1 | 5.1 | Reference | — | Reference | — | Reference | — | Reference | — |
| Quartile 2 | 6.4 | 1.26 (0.85–1.87) | 0.243 | 1.26 (0.85–1.87) | 0.243 | 1.26 (0.85–1.87) | 0.244 | 1.21 (0.82–1.80) | 0.338 |
| Quartile 3 | 9.6 | 1.93 (1.34–2.77) | < 0.001 | 1.93 (1.34–2.77) | < 0.001 | 1.98 (1.38–2.86) | < 0.001 | 1.83 (1.26–2.64) | 0.001 |
| Quartile 4 | 9.9 | 1.98 (1.38–2.84) | < 0.001 | 1.97 (1.37–2.83) | < 0.001 | 2.00 (1.39–2.88) | < 0.001 | 1.76 (1.21–2.55) | 0.003 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | 0.004 | |||||
| Cumulative LCI-hsCRP | |||||||||
| Quartile 1 | 3.9 | Reference | — | Reference | — | Reference | — | Reference | — |
| Quartile 2 | 7.5 | 1.97 (1.30–2.99) | 0.002 | 1.98 (1.31–3.00) | 0.001 | 1.99 (1.31–3.02) | 0.001 | 1.91 (1.26–2.91) | 0.002 |
| Quartile 3 | 8.6 | 2.28 (1.52–3.43) | < 0.001 | 2.30 (1.53–3.46) | < 0.001 | 2.35 (1.56–3.53) | < 0.001 | 2.15 (1.42–3.24) | < 0.001 |
| Quartile 4 | 11.2 | 3.01 (2.03–4.46) | < 0.001 | 3.03 (2.05–4.50) | < 0.001 | 3.14 (2.12–4.66) | < 0.001 | 2.75 (1.84–4.12) | < 0.001 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | < 0.001 | |||||
| Cumulative RC-hsCRP | |||||||||
| Quartile 1 | 4.8 | Reference | — | Reference | — | Reference | — | Reference | — |
| Quartile 2 | 7.5 | 1.57 (1.07–2.32) | 0.023 | 1.57 (1.06–2.32) | 0.023 | 1.61 (1.09–2.38) | 0.017 | 1.54 (1.04–2.28) | 0.029 |
| Quartile 3 | 9.2 | 1.97 (1.35–2.86) | < 0.001 | 1.98 (1.36–2.88) | < 0.001 | 2.03 (1.39–2.95) | < 0.001 | 1.83 (1.25–2.67) | 0.002 |
| Quartile 4 | 9.7 | 2.07 (1.42–3.00) | < 0.001 | 2.07 (1.42–3.00) | < 0.001 | 2.13 (1.46–3.09) | < 0.001 | 1.85 (1.26–2.71) | 0.002 |
| P for trend | 0.001 | 0.001 | < 0.001 | 0.016 | |||||
| Cumulative RC/HDL-C ratio-hsCRP | |||||||||
| Quartile 1 | 4.7 | Reference | — | Reference | — | Reference | — | Reference | — |
| Quartile 2 | 7.1 | 1.52 (1.02–2.26) | 0.040 | 1.52 (1.02–2.26) | 0.040 | 1.55 (1.04–2.30) | 0.032 | 1.49 (1.00–2.22) | 0.049 |
| Quartile 3 | 9.3 | 2.05 (1.40–2.98) | < 0.001 | 2.05 (1.41–2.99) | < 0.001 | 2.11 (1.45–3.09) | < 0.001 | 1.92 (1.31–2.81) | < 0.001 |
| Quartile 4 | 10.1 | 2.22 (1.53–3.23) | < 0.001 | 2.22 (1.53–3.22) | < 0.001 | 2.29 (1.58–3.33) | < 0.001 | 1.99 (1.36–2.92) | < 0.001 |
| P for trend | < 0.001 | < 0.001 | < 0.001 | 0.004 | |||||
Model 1 was adjusted for age and sex; Model 2 was additionally adjusted for marital status, residence, education, smoking, and drinking; Model 3 was further adjusted for body mass index, hypertension, heart disease, stroke, history of dyslipidemia medication use, and uric acid
Abbreviations: HR, hazard ratio; CI, confidence interval; hsCRP, high-sensitivity C-reactive protein; AIP, atherogenic index of plasma; Non-HDL-C, non-high-density lipoprotein cholesterol; AC, atherogenic coefficient; CRI-I, Castelli’s risk index I; CRI-II, Castelli’s risk index II; LCI, lipoprotein combined index; RC, remnant cholesterol; RC/HDL-C, remnant cholesterol/high-density lipoprotein cholesterol
Performance of non-traditional lipid-inflammatory parameters in predicting diabetes
The time-dependent ROC analysis revealed that the AUC values for the eight non-traditional lipid-inflammatory parameters in predicting diabetes risk ranged from 0.595 to 0.619 (Fig. 4 and Table S5). Among these parameters, LCI-hsCRP demonstrated the highest individual predictive value, with an AUC of 0.619 (95% CI: 0.597–0.640). Consistent with these findings, Table S6 quantifies the incremental predictive value of non-traditional lipid-inflammatory parameters using the NRI and IDI indices. Incorporation of these parameters into the basic model significantly improved both NRI and IDI values (all P < 0.001). LCI-hsCRP achieved the greatest incremental value, with an NRI of 0.150 (95% CI: 0.119–0.181) and an IDI of 0.011 (95% CI: 0.006–0.017).
Fig. 4.
ROC curve analysis of non-traditional lipid-inflammatory parameters in predicting diabetes. Abbreviations: AUC, area under the curve; hsCRP, high-sensitivity C-reactive protein; AIP, atherogenic index of plasma; Non-HDL-C, non-high-density lipoprotein cholesterol; AC, atherogenic coefficient; CRI-I, Castelli’s risk index I; CRI-II, Castelli’s risk index II; LCI, lipoprotein combined index; RC, remnant cholesterol; RC/HDL-C, remnant cholesterol/high-density lipoprotein cholesterol
Sensitivity and subgroup analysis
Sensitivity analyses, as detailed in Tables S7–S16, confirmed the robustness of the findings. Following the exclusion of individuals with missing data, the associations of baseline and cumulative non-traditional lipid-inflammatory parameters with incident diabetes persisted and were comparable to the primary results (Tables S7 and S12). These associations were also unchanged after removing participants followed for under two years (Tables S8 and S13). Moreover, the results were consistent across participants with different baseline glycemic statuses (Tables S9–S10 and S14–S15). The findings were similarly robust when death was considered a competing risk for diabetes (Tables S11 and S16). Stratified analyses showed that baseline non-traditional lipid-inflammatory parameters were more strongly associated with diabetes among females, individuals without a history of smoking or alcohol consumption, and hypertensive individuals (P for interaction < 0.05) (Table S17). By contrast, the associations of cumulative non-traditional lipid-inflammatory parameters with diabetes risk were consistent across subgroups, with no significant interaction effects observed (all P for interaction > 0.05) (Table S18).
Mediation analysis
As shown in Fig. 5, a reciprocal relationship was observed among non-traditional lipid indices, hsCRP, and diabetes risk. Specifically, hsCRP significantly mediated the associations of most non-traditional lipid indices with incident diabetes, except for Non-HDL-C and AC (P > 0.05). The mediation proportions were 2.2% for AIP, 2.7% for CRI-I, 3.6% for CRI-II, 2.4% for LCI, 2.1% for RC, and 2.9% for RC/HDL-C. Conversely, non-traditional lipid indices significantly mediated the associations of hsCRP with incident diabetes, excluding Non-HDL-C (P = 0.148). The mediation proportions were 14.2% for AIP, 22.3% for AC, 24.3% for CRI-I, 18.9% for CRI-II, 12.9% for LCI, 5.7% for RC, and 18.5% for RC/HDL-C.
Fig. 5.
Mediation analysis of the associations between non-traditional lipid indices and hsCRP with incident diabetes. Models were adjusted for age, sex, residence, education level, marital status, smoking status, drinking status, body mass index, hypertension, heart disease, stroke, history of dyslipidemia medication use, and uric acid. The non-traditional lipid indices and hsCRP were categorized into two groups based on the 75% cut-off. Abbreviations: HR, hazard ratio; hsCRP, high-sensitivity C-reactive protein; AIP, atherogenic index of plasma; Non-HDL-C, non-high-density lipoprotein cholesterol; AC, atherogenic coefficient; CRI-I, Castelli’s risk index I; CRI-II, Castelli’s risk index II; LCI, lipoprotein combined index; RC, remnant cholesterol; RC/HDL-C, remnant cholesterol/high-density lipoprotein cholesterol
Discussion
Utilizing data from a nationwide prospective cohort, this study systematically investigated and compared the associations and predictive value of eight non-traditional lipid-inflammatory parameters for diabetes in middle-aged and older Chinese adults. The findings indicate that elevated baseline and cumulative non-traditional lipid-inflammatory parameter levels are linked to a higher risk of incident diabetes. Among these parameters, LCI-hsCRP exhibited the highest predictive value. The associations of non-traditional lipid-inflammatory parameters and incident diabetes followed a nonlinear dose-response pattern. These findings remained consistent across sensitivity and subgroup analyses. Moreover, most non-traditional lipid indices and hsCRP were found to reciprocally mediate each other in relation to diabetes. These findings indicate that integrating non-traditional lipid indices with inflammatory markers may aid in identifying individuals at elevated risk of diabetes.
Our study demonstrated that non-traditional lipid-inflammatory parameters were positively correlated with the risk of diabetes, with LCI-hsCRP exhibiting superior predictive value. In recent years, emerging evidence has suggested that lipid dysregulation, along with concurrent systemic inflammation, may jointly contribute to the development of diabetes [25–29]. A longitudinal study involving populations from China and the United Kingdom reported that high levels of the remnant cholesterol inflammatory index (RCII), which integrates RC and hsCRP, were linked to an elevated risk of diabetes [25]. Similarly, another prospective cohort study in China found that elevated levels of AIP and hsCRP jointly increased the risk of new-onset diabetes [28]. However, most prior research has concentrated exclusively on the combined effects of individual non-traditional lipid indices and hsCRP, lacking comprehensive evaluations and comparisons across multiple lipid indices. The optimal combination of non-traditional lipid indices with hsCRP, as well as their critical thresholds for identifying diabetes, remains unclear. By systematically evaluating eight non-traditional lipid-inflammatory parameters concerning diabetes occurrence, our study addresses this gap and further highlights the predictive value of these markers, notably LCI-hsCRP. Additionally, consistent with previous studies [25, 27], we observed that non-traditional lipid-inflammatory parameters were nonlinearly associated with diabetes risk, suggesting that clinical diabetes risk assessment should consider the specific thresholds of these parameters for a more accurate evaluation.
In our study, LCI-hsCRP demonstrated the strongest association with incident diabetes and the highest predictive value among all examined non-traditional lipid-inflammatory parameters. This superior performance may be attributed to the unique composition of LCI. Unlike other non-traditional lipid indices that incorporate only two or three conventional lipid components, LCI (TC × TG × LDL-C/HDL-C) integrates all four conventional lipid components. This comprehensive integration may enable LCI to capture a more complete profile of lipid metabolism disturbances, reflecting both pro-atherogenic lipid overload (TC, TG, LDL-C) and reverse cholesterol transport (HDL-C) [36, 37]. Additionally, it is worth noting that incorporating the non-traditional lipid-inflammatory parameters into the basic model significantly improved NRI and IDI, but the absolute AUC values for these parameters remained relatively low. This phenomenon may be attributable to the multifactorial nature of diabetes. As a complex metabolic disorder, diabetes is influenced by a myriad of interconnected risk factors such as epigenetics, genetics, gut microbiome, environmental factors, organelle stress, and dietary habits [38], which may limit the discriminatory capacity of any single biomarker or composite index when used alone.
The underlying biological pathways linking non-traditional lipid indices and hsCRP to diabetes development remain unknown. In this study, we found that most non-traditional lipid indices partially mediated the effect of hsCRP on diabetes. It has been suggested that inflammation largely mediates lipid metabolism, influencing the composition of the blood lipid profile [39]. Excessive lipid deposition in pancreatic β-cells may lead to decreased insulin secretion and even trigger apoptosis in these cells [40, 41]. Abnormal lipid accumulation in non-adipose tissues can also induce oxidative stress [42], which may further compromise insulin sensitivity and contribute to abnormal glucose metabolism [10]. Conversely, our findings also indicated that the effect of most non-traditional lipid indices on diabetes was partially mediated by hsCRP. This aligns with existing evidence suggesting that dyslipidemia may aggravate the low-grade inflammatory state through lipotoxic effects [43]. Chronic low-grade inflammation may exacerbate glucose metabolism disorders by damaging pancreatic β-cells [10, 14], intensifying insulin resistance [44, 45], and promoting islet inflammation [14]. These findings suggest that non-traditional lipid indices and hsCRP are interrelated and may act synergistically to promote pancreatic β-cell dysfunction and insulin resistance, thereby contributing to the development of diabetes. Furthermore, prior research has identified a bidirectional mediating relationship between elevated RC or AIP and hsCRP in the onset of diabetes [25, 28]. Our study further extends this evidence, indicating that such bidirectional mediation may also apply to a broader range of non-traditional lipid indices, such as CRI-I, CRI-II, LCI, and RC/HDL-C. A comprehensive assessment of non-traditional lipid indices and hsCRP levels is crucial for detecting and classifying individuals at high risk for diabetes. Moreover, our findings align with prior research revealing a mutual association between non-traditional lipid indices and low-grade inflammation in other chronic diseases [31, 32], further emphasizing the clinical relevance of such comprehensive assessments. However, it is worth noting that the reciprocal mediation proportions between non-traditional lipid indices and hsCRP were relatively modest, indicating that these two factors serve as partial mediators for one another rather than dominant pathways. This implies that there may be other biological mechanisms or pathways underlying the effects of non-traditional lipid indices and hsCRP on diabetes risk, beyond the bidirectional mediation observed in this study.
Our findings have several important clinical implications. The predictive utility of non-traditional lipid-inflammatory parameters for diabetes highlights the need for comprehensive assessments of both non-traditional lipid indices and inflammatory markers when evaluating diabetes risk. Integrating these markers into routine clinical evaluations may help identify high-risk individuals, facilitating personalized lifestyle and pharmacological interventions, as well as more precise long-term monitoring. The combined effect of dyslipidemia and low-grade inflammation, as reflected by the non-traditional lipid-inflammatory parameters, indicates that interventions should simultaneously target both lipid and inflammatory pathways. Moreover, stratified analyses revealed that the relationships of non-traditional lipid-inflammatory parameters with diabetes varied across subgroups, underscoring the need to consider population differences when conducting early risk assessments and developing targeted prevention strategies.
Strengths and limitations
Our study had several strengths. First, the longitudinal associations between non-traditional lipid-inflammatory parameters and diabetes were examined in a nationally representative sample of Chinese adults aged ≥ 45 years, with the large sample size enhancing the reliability of the findings. Second, a comprehensive comparison of eight non-traditional lipid-inflammatory parameters was conducted, and the AUCs and optimal thresholds for these markers in diabetes detection were determined. Finally, the robustness of the results was supported by multiple sensitivity analyses.
However, the present study also had several limitations. First, due to data constraints, our study was unable to distinguish between type 1 and type 2 diabetes. Since the pathogenesis, risk factors, and roles of lipid metabolism and inflammation differ substantially between these two conditions, pooling all diabetes cases may weaken the clinical relevance and biological plausibility of the findings. Additionally, given the high prevalence of type 2 diabetes among middle-aged and older adults, the observed associations are likely driven primarily by type 2 diabetes. Second, due to the observational design, causal relationships between non-traditional lipid-inflammatory parameters and diabetes cannot be definitively established. Mediation analysis results should also be interpreted with extreme caution and not as definitive evidence of a causal pathway, as the assumption of no unmeasured confounding between the exposure, mediator, and outcome is unlikely to hold in observational data. Third, although multiple confounders were adjusted for, unmeasured or inadequately controlled factors, such as diet, physical activity, family history, and medication use (e.g., anti-inflammatory drugs), may still influence the accuracy and reliability of the findings. Fourth, this study focused on middle-aged and older Chinese adults, which may limit the generalizability of the findings to younger populations or non-Chinese populations with different ethnicities, genetic backgrounds, and lifestyles. Additionally, the absence of an external validation cohort further restricts the generalizability of the findings. Future prospective studies using cohorts such as the Health and Retirement Study (HRS) and the English Longitudinal Study of Ageing (ELSA) are warranted to validate these findings in Western populations. Fifth, all non-traditional lipid-inflammatory parameters demonstrated only modest discriminatory power, and the clinical utility of adding these parameters to existing risk prediction tools (e.g., those based on BMI, FPG, and HbA1c) [46–48] remains unclear and requires further validation. Finally, CHARLS blood biomarker data were measured only in Wave 1 (2011–2012) and Wave 3 (2015), with a relatively long interval between the two waves. Consequently, the cumulative non-traditional lipid-inflammatory parameters, derived by averaging these two measurements and multiplying by the time interval, cannot capture fluctuations between these time points and may not adequately reflect true long-term cumulative exposure. Future studies should include multiple repeated measurements and apply the AUC method to achieve a more accurate assessment of cumulative exposure.
Conclusion
In summary, our study found that higher baseline and cumulative non-traditional lipid-inflammatory parameter levels were significantly linked to a higher risk of incident diabetes. Among these parameters, LCI-hsCRP exhibited the strongest predictive ability. Non-traditional lipid indices and hsCRP may exert bidirectional mediating effects on the incidence of diabetes. These findings highlight the importance of combining non-traditional lipid indices and inflammatory markers for a thorough assessment of diabetes risk. Integrating these parameters into clinical and therapeutic practice could enhance early detection of individuals at elevated risk for diabetes and support more personalized preventive measures.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
The authors express their sincere gratitude to the CHARLS research team for their invaluable efforts, as well as to all study participants for generously sharing their data.
Abbreviations
- AC
Atherogenic coefficient
- AIP
Atherogenic index of plasma
- AUC
Area under the curve
- BMI
Body mass index
- CHARLS
China Health and Retirement Longitudinal Study
- CI
Confidence interval
- CRI-I
Castelli’s risk index I
- CRI-II
Castelli’s risk index II
- DBP
Diastolic blood pressure
- FPG
Fasting plasma glucose
- HbA1c
Glycated hemoglobin
- HDL-C
High-density lipoprotein cholesterol
- HR
Hazard ratio
- hsCRP
High-sensitivity C-reactive protein
- IDI
Integrated discrimination improvement
- IQR
Interquartile range
- LCI
Lipoprotein combined index
- LDL-C
Low-density lipoprotein cholesterol
- Non-HDL-C
Non-high-density lipoprotein cholesterol
- NRI
Net reclassification improvement
- RC
Remnant cholesterol
- RC/HDL-C
Remnant cholesterol/high-density lipoprotein cholesterol
- RCII
Remnant cholesterol inflammatory index
- RCS
Restricted cubic spline
- ROC
Receiver operating characteristic
- SBP
Systolic blood pressure
- SD
Standard deviation
- TC
Total cholesterol
- TG
Triglycerides
- VIF
Variance inflation factors
Author contributions
Jie Hua, Qingqing Jiang, and Furong Wang conceived and designed the study. Jie Hua and Qingqing Jiang conducted the data analysis and drafted the manuscript. Jie Hua, Qingqing Jiang, and Shiyi Cao interpreted the results. Shiyi Cao and Furong Wang supervised the study. All authors have read and approved the final manuscript.
Funding
This work was supported by the Key Research and Development Program of Hubei Province (Grant No. 2020BCA089).
Data availability
This study utilized data from the CHARLS. The data are publicly accessible upon registration and formal application via the official CHARLS website (http://charls.pku.edu.cn).
Declarations
Ethics approval and consent to participate
Ethical approval for the CHARLS study was granted by the Biomedical Ethics Review Board of Peking University (IRB00001052-11015), and written informed consent was obtained from all participants prior to enrollment. The study was conducted in accordance with the principles of the Declaration of Helsinki.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Jie Hua and Qingqing Jiang contributed equally to this work.
Contributor Information
Shiyi Cao, Email: caoshiyi@hust.edu.cn.
Furong Wang, Email: wangfurong.china@163.com.
References
- 1.Sun H, Saeedi P, Karuranga S, Pinkepank M, Ogurtsova K, Duncan BB, et al. IDF Diabetes Atlas: Global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045. Diabetes Res Clin Pract. 2022;183:109119. 10.1016/j.diabres.2021.109119. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Dal Canto E, Ceriello A, Rydén L, Ferrini M, Hansen TB, Schnell O, et al. Diabetes as a cardiovascular risk factor: An overview of global trends of macro and micro vascular complications. Eur J Prev Cardiol. 2019;26 2suppl:25–32. 10.1177/2047487319878371. [DOI] [PubMed] [Google Scholar]
- 3.Bragg F, Holmes MV, Iona A, Guo Y, Du H, Chen Y, et al. Association Between Diabetes and Cause-Specific Mortality in Rural and Urban Areas of China. JAMA. 2017;317:280–9. 10.1001/jama.2016.19720. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Emerging Risk Factors Collaboration. Life expectancy associated with different ages at diagnosis of type 2 diabetes in high-income countries: 23 million person-years of observation. Lancet Diabetes Endocrinol. 2023;11:731–42. 10.1016/S2213-8587(23)00223-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Bertoni AG, Burke GL, Owusu JA, Carnethon MR, Vaidya D, Barr RG, et al. Inflammation and the incidence of type 2 diabetes: the Multi-Ethnic Study of Atherosclerosis (MESA). Diabetes Care. 2010;33:804–10. 10.2337/dc09-1679. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Donath MY, Shoelson SE. Type 2 diabetes as an inflammatory disease. Nat Rev Immunol. 2011;11:98–107. 10.1038/nri2925. [DOI] [PubMed] [Google Scholar]
- 7.Peng J, Zhao F, Yang X, Pan X, Xin J, Wu M, et al. Association between dyslipidemia and risk of type 2 diabetes mellitus in middle-aged and older Chinese adults: a secondary analysis of a nationwide cohort. BMJ Open. 2021;11:e042821. 10.1136/bmjopen-2020-042821. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Sheng G, Kuang M, Yang R, Zhong Y, Zhang S, Zou Y. Evaluation of the value of conventional and unconventional lipid parameters for predicting the risk of diabetes in a non-diabetic population. J Transl Med. 2022;20:266. 10.1186/s12967-022-03470-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Sokooti S, Flores-Guerrero JL, Heerspink HJL, Connelly MA, Bakker SJL, Dullaart RPF. Triglyceride-rich lipoprotein and LDL particle subfractions and their association with incident type 2 diabetes: the PREVEND study. Cardiovasc Diabetol. 2021;20:156. 10.1186/s12933-021-01348-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Dludla PV, Mabhida SE, Ziqubu K, Nkambule BB, Mazibuko-Mbeje SE, Hanser S, et al. Pancreatic β-cell dysfunction in type 2 diabetes: Implications of inflammation and oxidative stress. World J Diabetes. 2023;14:130–46. 10.4239/wjd.v14.i3.130. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Elkanawati RY, Sumiwi SA, Levita J. Impact of Lipids on Insulin Resistance: Insights from Human and Animal Studies. Drug Des Devel Ther. 2024;18:3337–60. 10.2147/DDDT.S468147. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Olefsky JM, Glass CK. Macrophages, Inflammation, and Insulin Resistance. Annu Rev Physiol. 2010;72:219–46. 10.1146/annurev-physiol-021909-135846. [DOI] [PubMed] [Google Scholar]
- 13.Glass CK, Olefsky JM. Inflammation and lipid signaling in the etiology of insulin resistance. Cell Metab. 2012;15:635–45. 10.1016/j.cmet.2012.04.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Eguchi K, Nagai R. Islet inflammation in type 2 diabetes and physiology. J Clin Invest. 2017;127:14–23. 10.1172/JCI88877. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Lin D, Qi Y, Huang C, Wu M, Wang C, Li F, et al. Associations of lipid parameters with insulin resistance and diabetes: A population-based study. Clin Nutr. 2018;37:1423–9. 10.1016/j.clnu.2017.06.018. [DOI] [PubMed] [Google Scholar]
- 16.Wang K, Yu G, Yan L, Lai Y, Zhang L. Association of non-traditional lipid indices with diabetes and insulin resistance in US adults: mediating effects of HOMA-IR and evidence from a national cohort. Clin Exp Med. 2025;25:281. 10.1007/s10238-025-01819-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Li M, Zhang W, Zhang M, Li L, Wang D, Yan G, et al. Nonlinear relationship between untraditional lipid parameters and the risk of prediabetes: a large retrospective study based on Chinese adults. Cardiovasc Diabetol. 2024;23:12. 10.1186/s12933-023-02103-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Qiu X, Han Y, Cao C, Liao Y, Hu H. Association between atherogenicity indices and prediabetes: a 5-year retrospective cohort study in a general Chinese physical examination population. Cardiovasc Diabetol. 2025;24:220. 10.1186/s12933-025-02768-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Zou Y, Lu S, Li D, Huang X, Wang C, Xie G, et al. Exposure of cumulative atherogenic index of plasma and the development of prediabetes in middle-aged and elderly individuals: evidence from the CHARLS cohort study. Cardiovasc Diabetol. 2024;23:355. 10.1186/s12933-024-02449-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Wu X, Gao Y, Wang M, Peng H, Zhang D, Qin B, et al. Atherosclerosis indexes and incident T2DM in middle-aged and older adults: evidence from a cohort study. Diabetol Metab Syndr. 2023;15:23. 10.1186/s13098-023-00992-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Song Q, Liu X, Wang A, Wang Y, Zhou Y, Zhou W, et al. Associations between non-traditional lipid measures and risk for type 2 diabetes mellitus in a Chinese community population: a cross-sectional study. Lipids Health Dis. 2016;15:70. 10.1186/s12944-016-0239-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Yang X, Tao S, Peng J, Zhao J, Li S, Wu N, et al. High-sensitivity C-reactive protein and risk of type 2 diabetes: A nationwide cohort study and updated meta-analysis. Diabetes Metab Res Rev. 2021;37:e3446. 10.1002/dmrr.3446. [DOI] [PubMed] [Google Scholar]
- 23.Tong KI, Hopstock LA, Cook S. Association of C-reactive protein with future development of diabetes: a population-based 7-year cohort study among Norwegian adults aged 30 and older in the Tromsø Study 2007–2016. BMJ Open. 2023;13:e070284. 10.1136/bmjopen-2022-070284. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Pradhan AD, Manson JE, Rifai N, Buring JE, Ridker PM. C-reactive protein, interleukin 6, and risk of developing type 2 diabetes mellitus. JAMA. 2001;286:327–34. 10.1001/jama.286.3.327. [DOI] [PubMed] [Google Scholar]
- 25.Shao Y, Li Z, Wu Q, Ye L, Shi H. Effects of combined high-sensitivity C-reactive protein and residual cholesterol levels on new-onset diabetes: evidence from two prospective cohort studies. J Transl Med. 2025. 10.1186/s12967-025-07544-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Chen N, Gong A, Huang X, Wang N, Pan T, Pan X. Remnant Cholesterol Inflammatory Index and New-Onset Diabetes in Middle-Aged and Older Adults: Evidence From Prospective Surveys of Chinese and UK Populations. J Diabetes Res. 2025;2025:1579313. 10.1155/jdr/1579313. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Li Q, Gong L, Li H. Longitudinal correlation between cumulative remnant cholesterol inflammatory index and incident diabetes. Diabetol Metab Syndr. 2025;17:445. 10.1186/s13098-025-02018-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Wang T, Zhang M, Shi W, Li Y, Zhang T, Shi W. Atherogenic index of plasma, high sensitivity C-reactive protein and incident diabetes among middle-aged and elderly adults in China: a national cohort study. Cardiovasc Diabetol. 2025;24:103. 10.1186/s12933-025-02653-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Lan Y, Chen G, Wu D, Ding X, Huang Z, Wang X, et al. Temporal relationship between atherogenic dyslipidemia and inflammation and their joint cumulative effect on type 2 diabetes onset: a longitudinal cohort study. BMC Med. 2023;21:31. 10.1186/s12916-023-02729-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Zhao Y, Hu Y, Smith JP, Strauss J, Yang G. Cohort profile: the China Health and Retirement Longitudinal Study (CHARLS). Int J Epidemiol. 2014;43:61–8. 10.1093/ije/dys203. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Huang X, Li C, Zeng P, Ling Y, Tan S, Bai Z, et al. Non-traditional lipid-inflammatory parameters estimate the risk of stroke in middle-aged and older Chinese adults: a nationwide prospective cohort study. J Adv Res. 2025. 10.1016/j.jare.2025.11.029. [DOI] [PubMed] [Google Scholar]
- 32.Chen J, Wu Q, Liu H, Hu W, Zhu J, Ji Z, et al. Predictive value of remnant cholesterol inflammatory index for stroke risk: Evidence from the China health and Retirement Longitudinal study. J Adv Res. 2025;76:543–52. 10.1016/j.jare.2024.12.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.American Diabetes Association Professional Practice Committee. 2. Diagnosis and Classification of Diabetes: Standards of Care in Diabetes-2025. Diabetes Care. 2025;48(1 Suppl 1):S27–49. 10.2337/dc25-S002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.van Buuren S, Groothuis-Oudshoorn K. mice: Multivariate Imputation by Chained Equations in R. J Stat Softw. 2011;45:1–67. 10.18637/jss.v045.i03. [Google Scholar]
- 35.Shi B, Choirat C, Coull BA, VanderWeele TJ, Valeri L, CMAverse:. A Suite of Functions for Reproducible Causal Mediation Analyses. Epidemiology. 2021;32:e20–2. 10.1097/EDE.0000000000001378. [DOI] [PubMed] [Google Scholar]
- 36.Taskinen M-R, Borén J. New insights into the pathophysiology of dyslipidemia in type 2 diabetes. Atherosclerosis. 2015;239:483–95. 10.1016/j.atherosclerosis.2015.01.039. [DOI] [PubMed] [Google Scholar]
- 37.Ouimet M, Barrett TJ, Fisher EA. HDL and Reverse Cholesterol Transport. Circ Res. 2019;124:1505–18. 10.1161/CIRCRESAHA.119.312617. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Singh S, Kriti M, K S A, Sarma DK, Verma V, Nagpal R, et al. Deciphering the complex interplay of risk factors in type 2 diabetes mellitus: A comprehensive review. Metabol Open. 2024;22:100287. 10.1016/j.metop.2024.100287. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.McGillicuddy FC, de la Llera Moya M, Hinkle CC, Joshi MR, Chiquoine EH, Billheimer JT, et al. Inflammation impairs reverse cholesterol transport in vivo. Circulation. 2009;119:1135–45. 10.1161/CIRCULATIONAHA.108.810721. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Hao M, Head WS, Gunawardana SC, Hasty AH, Piston DW. Direct effect of cholesterol on insulin secretion: a novel mechanism for pancreatic beta-cell dysfunction. Diabetes. 2007;56:2328–38. 10.2337/db07-0056. [DOI] [PubMed] [Google Scholar]
- 41.Bogan JS, Xu Y, Hao M. Cholesterol accumulation increases insulin granule size and impairs membrane trafficking. Traffic. 2012;13:1466–80. 10.1111/j.1600-0854.2012.01407.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Hajri T, Gharib M, Fungwe T, M’Koma A. Very low-density lipoprotein receptor mediates triglyceride-rich lipoprotein-induced oxidative stress and insulin resistance. Am J Physiol Heart Circ Physiol. 2024;327:H733–48. 10.1152/ajpheart.00425.2023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Kraaijenhof JM, Hovingh GK, Stroes ESG, Kroon J. The iterative lipid impact on inflammation in atherosclerosis. Curr Opin Lipidol. 2021;32:286–92. 10.1097/MOL.0000000000000779. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Li H, Meng Y, He S, Tan X, Zhang Y, Zhang X, et al. Macrophages, chronic inflammation, and insulin resistance. Cells. 2022;11. 10.3390/cells11193001. [DOI] [PMC free article] [PubMed]
- 45.Wu H, Ballantyne CM. Metabolic Inflammation and Insulin Resistance in Obesity. Circ Res. 2020;126:1549–64. 10.1161/CIRCRESAHA.119.315896. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Cai X, Wang M, Liu S, Yuan Y, Hu J, Zhu Q, et al. Establishment and validation of a nomogram that predicts the risk of type 2 diabetes in obese patients with non-alcoholic fatty liver disease: a longitudinal observational study. Am J Transl Res. 2022;14:4505–14. [PMC free article] [PubMed] [Google Scholar]
- 47.Cai X, Zhu Q, Cao Y, Liu S, Wang M, Wu T, et al. A Prediction Model Based on Noninvasive Indicators to Predict the 8-Year Incidence of Type 2 Diabetes in Patients with Nonalcoholic Fatty Liver Disease: A Population-Based Retrospective Cohort Study. Biomed Res Int. 2021;2021:5527460. 10.1155/2021/5527460. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Cai X, Zhu Q, Wu T, Zhu B, Aierken X, Ahmat A, et al. Development and Validation of a Novel Model for Predicting the 5-Year Risk of Type 2 Diabetes in Patients with Hypertension: A Retrospective Cohort Study. Biomed Res Int. 2020;2020:9108216. 10.1155/2020/9108216. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
This study utilized data from the CHARLS. The data are publicly accessible upon registration and formal application via the official CHARLS website (http://charls.pku.edu.cn).